TW201143455A - Image encoding device, image decoding device, image encoding method, image decoding method - Google Patents

Abstract

A similarity calculation part 42 which calculates an evaluation value for spatial direct mode using spatial direct vectors and calculates an evaluation value for temporal direct mode using temporal direct vectors, and a direct vector selection part 43 which compares the evaluation value for spatial direct mode and the evaluation value for temporal direct mode and selects either the spatial direct vector or the temporal direct vector are provided in an image encoding device of the present invention. A motion compensation prediction processing part 23 generates a prediction image by performing a motion compensation prediction process using the direct vector selected by the direct vector selection part 43.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding device, a video decoding device, an image encoding method, and a video decoding method used in image compression coding technology, compressed image data transmission technology, and the like. [Prior Art] For example, the international standard video coding method such as MPEG (Moving Picture Experts Group) and "πυ-Τ Η 26x" is used to combine the luminance signal 16x16 pixels and the luminance signal 16χ16 The block data (bl〇ckdata) (hereinafter referred to as "macroblock") formed by the color difference signal 8x8 is a unit, and the number of the image according to the shift technique and the orthogonal transform/transformation The method of compression by technology. The motion compensation processing in the video encoding device and the video decoding device refers to the front or rear picture, and performs motion vector detection and prediction image generation in units of large blocks. This temple's reference to a facet and the inter-plane predictive coding is called P-picture (P picture). At the same time, the two-faced reference is used for inter-plane prediction. The coder is called B picture. In the international standard coding method, AVC/H.264CISO/IEC 14496-10 I ITU-TH. 264) 'When performing B-plane coding, it can be called direct mode (direct coding mode (see for example Non-Patent Document 1) 〇, that is, you can choose: the large block of the object to be edited (the large block to be coded) does not retain the data of the moving vector, but by using the coded 322763 4 201143455 The motion vector of the large block of his picture, or the predetermined calculation process of the motion vector of the surrounding large block, to generate the coding mode of the motion vector of the large block of the coding object. The direct mode has time Direct mode and spatial direct mode. The time direct mode refers to the motion vector of other encoded faces, and adjusts the ratio of the moving vector according to the time difference between the encoded face and the face of the coded object (sea 1 ing). To generate a motion vector of a large block of the coding object. The spatial direct mode refers to a motion vector of at least one of the coded large blocks located around the large block of the coding object, from which The motion vector is used to generate a motion vector of the large block of the coding object. The direct mode can be obtained by using the flag "direct_spatial_mv_pred_flag" set in the slice header (s1 i ce header) and the slice (si ice) The unit selects either the time direct mode or the spatial direct mode. Fig. 9 is a schematic diagram showing a method of generating a motion vector in a time direct mode. In Fig. 9, "P" represents a P plane, and "B" represents a B node. In addition, the numbers 0-3 indicate the order in which the faces are displayed, representing the display images of the times TO, T1, T2, and T3. The encoding processing of the screens is performed in the order of P0, P3, B1, and B2. The case where the large block MB 1 in the face B2 is encoded in the time direct mode. In this case, the closest picture to the rear of the edited 5 322763 201143455 coded face located behind the face B2 on the time axis is used. The motion vector of the face P3 of B2, that is, the motion vector of the large block MB2 spatially located at the same position as the large block MB1 in the facet P3. The motion vector MV is referred to the face p〇, Large area To use the occasion of encoding the motion vectors MVLO, MVL1, to obtain the following system ⑴. Su formula λ Τ2-Τ0

MVL0---xMV Τ3-Γ0 λ/η/τ 1 — T3

Mm=~ ^MV T3-TQ Therefore, to obtain the motion vector of the large block of the coding object in the direct time mode, the motion vector My of the coded face of one picture is required, so the memory that holds the motion vector is needed. (mem〇ry). Figure 10 shows the intent of the method of generating a motion vector in spatial direct mode. In Fig. 10, current MB indicates a large block of the encoding object. At this time, it is assumed that the motion vector of the coded large block A on the left side of the large block of the coding object is MVa, and the coded object above the large block is encoded: the motion vector of the large block B is MVb, and the coding target area is large. If the motion vector of the coded large block C on the upper right of the block is MVc, the median value of the motion vectors MVa, MVb, and MVc is determined as shown in the following equation (2). The motion vector MV of the large block of the coding object is extracted. MV = ine d i a η (Μ V a, MVb, MVc) (2) In the spatial direct mode, the motion vector is obtained for each of the front and the rear, but either one can be obtained by the above method. 322763 6 201143455 In AVC/H·264, which side of the spatial direct mode or the time direct mode is to be selected, as described above with reference to the flag "direct one spatial_mv_pred_flag" set in the fragment header, thus, it is necessary to The slice is selected in units of time direct mode or spatial direct mode, so it is not possible to switch and use the most suitable direct mode for each large block one by one. [Prior Art Document] [Non-Patent Document] Non-Patent Document 1: MPEG-4 AVC (IS0/IEC 14496-10)/ITU-T H.264 Specifications [Summary of the Invention] [Problems to be Solved by the Invention] Conventional Images The coding apparatus is constructed as described above, and the time direct mode and the spatial direct mode can be switched in units of segments (siice) by referring to the flag "direct_spatial_mv_pred_flag" provided in the slice header. However, since the direct direct mode and the spatial direct mode cannot be switched in units of large blocks, the most suitable direct mode for a large block belonging to a certain segment is, for example, a spatial direct mode 'but corresponding to The direct mode of the segment is determined to be the time direct mode, and the time direct mode must be used for the large block, and the most suitable direct mode cannot be selected. In such a case, since the most suitable direct mode cannot be selected, it is necessary to encode an unnecessary motion vector, and there is a problem that the amount of coding increases. The present invention has been made in order to solve the above problems, and an object thereof is to obtain an image encoding device, a video decoding device, a video encoding method, and a video decoding method capable of selecting an optimum direct mode in units of predetermined blocks in 322763 7 201143455. . [Means for Solving the Problem] The video encoding device of the present invention is provided with a direct vector generating means for generating a spatial direct vector of a spatial direct mode from a motion vector of an encoded block located around a coding target block, and The time vector of the encoded direct face is generated in time near the coding target block to generate a temporal direct vector of the temporal direct mode; the evaluation value calculation means calculates the spatial direct using the spatial direct vector generated by the direct vector generation means The evaluation value of the mode, and using the time direct vector to calculate the evaluation value of the temporal direct mode; and the direct vector selection means, the evaluation value of the spatial direct mode calculated by the evaluation value calculation means and the evaluation value of the temporal direct mode For comparison, one of the spatial direct vector or the temporal direct vector is selected, and the predicted image generating means performs the motion compensated prediction process using the direct vector selected by the direct vector selecting means to generate the predicted image. [Effect of the Invention] According to the present invention, since the direct vector generation means is provided, a spatial direct vector of the spatial direct mode is generated from the motion vector of the coded block located around the coding target block, and from the time The motion vector of the coded facet near the coding target block is used to generate a temporal direct vector of the temporal direct mode; the evaluation value calculation means calculates the spatial direct mode using the spatial direct vector generated by the direct vector generation means 322763 201143455 The evaluation value is sufficient to use the time direct vector to calculate the time direct mode = estimate; R and the direct vector selection means to calculate the value and time of the spatial direct mode calculated by calculating the hand & The ratio of the flat mode of the direct mode & the selection of the spatial direct vector or the temporal direct vector and the predicted image generation means implement the motion compensation prediction process using the direct vector selected by the direct vector selection means to generate the prediction The block that is reserved by the image department selects the most suitable direct mode for the unit. As a result, the result is that the encoding of unnecessary moving vectors can be avoided, and the effect of increasing the amount of encoding is prevented. [Embodiment] Hereinafter, the present invention will be described in order to explain the present invention in more detail. (Embodiment 1) The video coding apparatus of the first embodiment of the video coding apparatus of the first embodiment is described as an example of the coding method used in the AVC/H.264 specification. In Fig. 1, the motion vector memory 1 is a recording medium such as a RAM for storing a motion vector of a coded large block (or a sub-block (sub_macr〇b 1〇ck) in which a large block is divided). . The motion compensation prediction unit 2 performs a process of selecting a reference image of a frame from a motion compensation prediction reference image stored in one of the frames of the frame memory (frame mem〇ry) 9, and then constructing The large block of the input image (or the sub-block formed by dividing the large block) is the unit performing/shifting 322763 9 201143455 motion compensation prediction processing, and generating the motion vector of the large block (coding block large block) to generate The image is predicted, and an identification number, a motion vector, a predicted image, and the like of the reference image selected for each of the large blocks are output. However, for convenience of explanation, it is assumed that the motion vector is generated in units of large blocks to generate a predicted image. However, when generating the motion vector of the large block to be encoded and generating the predicted video, the motion compensation prediction unit 2 performs the following processing for each of the large blocks constituting the input video: from the coded around the large block a large block motion vector (moving vector stored in the motion vector memory 1) to generate a spatial direct vector of the spatial direct mode, and a motion vector from the encoded picture that is temporally near the large block (stored in The motion vector in the vector memory 1 is moved to generate a temporal direct vector of the temporal direct mode. Then, the motion compensation prediction unit 2 performs a process of calculating an evaluation value of the spatial direct mode using the spatial direct vector, and calculating an evaluation value of the temporal direct mode using the time direct vector. Then, the motion compensation prediction section 2 performs a process of comparing the evaluation value of the spatial direct mode with the evaluation value of the temporal direct mode, and selecting either one of the spatial direct vector or the temporal direct vector. Then, the motion compensation prediction unit 2 performs a process of generating a prediction video by performing motion compensation prediction processing using the selected direct vector. The subtracter 3 performs the following processing: the difference image of the predicted image generated by the motion compensation prediction unit 2 and the input image is given to 10 322763 201143455 •=2= indicates that the difference difference signal of the difference image is output to the code=nuclear type The determination unit 4 performs a process of evaluating the prediction efficiency of the difference signal of the round (four) difference of the subtractor 3, and selecting the prediction efficiency most = two difference signals from among the more than one prediction difference signals of the subtractor 3, and then The prediction difference signal relates to a motion vector and a sub-block pattern used in the generation of the predicted image of the prediction unit 2 (including, for example, the -L: The output code of the inter mode or the direct mode refers to the code of the image as the coding mode information, and the signal is output to the compression unit 5 and the prediction difference compression unit 5 with the highest efficiency is implemented as follows: The difference is calculated by the DCT (four) T (discrete cosine transform) processed by the code pattern determining unit 4 to process the coefficient of the temple, and the (10) system is processed to treat the DCT coefficient (a pressure) The quantization unit is used to quantize the quantization unit 1 to μ. The local decoding unit 6 and the variable length, the subtractor 3, and the coding mode are quantized. The 卩4 and the compression unit 5 are localized. The decoding unit 6 performs the following processing. The reduced data is inversely quantized to obtain the inverse DCT (inverse discrete cosine transform) processing of the DCT output by the DCT ZHAO 邛 邛 5 "the tube ^ the 实 difference of the DCT coefficient real part 4 (four) 阙 difference The following equations are performed: The pre-determination adder 7 performs the following processing: The pre-circle 322763 11 201143455 calculated by the local decoding unit 6: the difference signal and the predicted video image generated by the motion compensation unit 2: =r By adding 'to generate a local decoding shadow solution: the chopper (1_filter) 8 performs the following processing: for the addition, the local image signal complemented by the coding distortion contained in the locally decoded video signal The F-wool image is output to the frame memory 9 » ^ memory 9 is used to store a recording medium such as a RAM of the loop chopper image. The τ and output degree encoding unit 1G performs the following processing: From _ section 5 number) into the 弁, 弋 move inward, ginseng The identification bit string 'Ά. entRDPy' of the test image encodes 'and generates a round π reamx_data indicating the entropy coding result'' and then flows the bit stream to the length coding unit 1Q to form a variable length coding The figure shows a configuration diagram of the motion compensation unit 2 in the embodiment J of the present invention. The image correction unit 21 receives the representation inter-surface mode = (real: outer 2 , motion vector, and 'quantity input: = ^ ^ direct vector generation unit 22 in the pile of 5 丨丨 local-station information, the implementation of the non-flat code mode for direct mode implementation: for each stone The large block of the horse object ' 322763 12 201143455 generates a spatial direct vector of the spatial direct mode and a temporal direct vector of the temporal direct mode, and outputs any one of the spatial direct vector or the temporal direct vector as a motion vector to the motion compensation process Part 23. The motion compensation processing unit 23 performs a process of performing motion compensation prediction processing using the motion vector output from the motion vector search unit 21 or the direct vector generation unit 22 and the reference image of a frame stored in the frame memory 9. The process of generating a predicted image. Further, the motion compensation processing unit 23 constitutes a predictive video generating means. Fig. 3 is a view showing the configuration of the direct vector generation unit 22 constituting the motion compensation prediction unit 2. In Fig. 3, the spatial direct vector generation unit 31 performs a process of encoding a large block located around the large block of the coding target from among the motion vectors of the coded large block stored in the motion vector memory 1. The motion vector is read out, and a spatial direct vector of the spatial direct mode is generated from the motion vector. The temporal direct vector generation unit 32 performs a process of shifting the motion vector of the encoded face temporally in the vicinity of the large block of the coding target from among the motion vectors of the coded large block stored in the motion vector memory 1, That is, the motion vector of the large block in the coded picture that is spatially located at the same position as the large block of the coding object is read, and the time direct vector of the temporal direct mode is generated from the motion vector. The spatial direct vector generation unit 31 and the temporal direct vector generation unit 32 constitute a direct vector generation means. The direct vector determination unit 33 performs a process of calculating the evaluation value of the spatial direct mode and the temporal direct vector generated by the temporal direct vector generation unit 32 using the spatial direct vector generated by the spatial direct vector 13 322763 201143455. To calculate the evaluation value of the time direct mode, and then compare the evaluation value of the spatial direct mode with the evaluation value of the temporal direct mode, and select either one of the spatial direct vector or the temporal direct vector. Fig. 4 is a view showing the configuration of the direct vector determining unit 33 constituting the direct vector generating unit 22. In the fourth diagram, the motion compensation unit 41 performs a process of generating a forward prediction video and a backward prediction video in the spatial direct mode using the spatial direct vector generated by the spatial direct vector generation unit 31, and the use time direct vector generation unit 32. The generated time direct vector is used to generate a forward predicted image and a backward predicted image in a temporal direct mode. The similarity calculation unit 42 performs a process of calculating the similarity between the forward predicted image and the backward predicted image in the spatial direct mode as the evaluation value of the spatial direct mode, and calculating the forward predicted image of the temporal direct mode and the backward predicted image. Degree, which is used as an evaluation value of the time direct mode. The motion compensation unit 41 and the similarity calculation unit 42 constitute an evaluation value calculation means. The direct vector selection unit 43 performs a process of performing the similarity between the forward predicted image and the backward predicted image in the spatial direct mode calculated by the similarity calculation unit 42 and the similarity between the forward predicted image and the backward predicted image in the temporal direct mode. In contrast, a direct vector of 14 322763 201143455 ίΠΐ with a higher degree of similarity between the forward-predicted image and the rear-predicted image is selected in the spatial direct vector or the temporal direct vector. The direct vector selecting unit 43 constitutes an example of a coding method employed by the video decoding device of Fig. 5 of the video decoding device in the present embodiment. H 264 specification

In W, the 'moving vector note 51 is used to store the J-recording medium such as the decoded large area: the sub-block formed by the block larger than the block. For example, the RAM image: =:::== 1st (four) bit The input of the meta-string U code 枓), for the vertical string machine to decode to decode the compressed data and coding mode information number) U-type / sub-block type, motion vector, reference image recognition and processing The Yanxian decoding unit 53 performs the second motion compensation prediction unit 54. The variable length decoding unit 52 constitutes a variable length decoding means. The prediction error decoding unit 53 performs inverse quantization on the compressed data output from the variable length demodulation unit 52 in the following processing 1 to obtain (4) coefficients, and performs inverse Da processing on the DCT coefficients to calculate a prediction error signal indicating the difference image (and The preamble outputted by the coding mode portion 4 of Fig. 1 =:: prediction error signal). The prediction error decoding unit 53 performs a reference image indicated by the identification number outputted by the == 322763 15 201143455 unit 52 among the reference pictures on the following processing_the following: * from the variable length The large block type two output block outputted by the decoding unit 52 indicates that the material used in the inter-plane mode is used, and the motion vector output from the 256 and the reference picture, *, are used. The measurement process is performed to generate a predicted image. On the other hand, in the case where the large-length block type money (4) of the variable length material unit (10) is in the direct mode, the motion compensation prediction unit 2 in the video encoding apparatus is provided. Processing: Generate spatial direct vector and time straight, 'implement the following direct vector or time direct vector 任°1 and then select the null vector, and the identification indicated by the identification number only use the selected direct measurement processing This produces a predicted image. The 'f target movement compensation pre-adder 55 performs the following processing: the difference image indicated by the signal generated by the pre-clearing and Na's solution (4) 彳 54 is added, and the prediction error of the 彳 wheel is shown in the first figure. The image code is loaded with the image of the additive image of the + (the image is equivalent). ° The local solution of the mother's shadow wave chopper 56 implements the following processing: the coded distortion included in the decoded image signal is generated by the complemented by the decoded image signal, and then the coded image is decoded. Store it in the frame memory image as a reference external. And the decoded image is rotated out to two plus= and the loop chopper 56 is formed into an image plus a 匡 匡 己 己 己 己 己 己 己 57 57 57 57 57 57 57 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 322 media. Fig. 6 is a view showing the configuration of the motion compensation prediction unit 54 in the first embodiment of the present invention. In the sixth image of the image decoding apparatus, the direct vector generation unit 61 performs the following processing in the case of the large block type/sub-block type four-length decoding unit mode outputted by 52: for each solid use疋 Directly generate spatial direct vector space direct vector code object large block, direct vector, and output the space direct vector or time =: as the motion vector to the mobile compensation β - mobile The compensation processing unit 62 performs the following process two. Among the reference images above the frame of the body 57, the reference $ indicated by the identification number outputted in the frame memory 52 will be expressed from the variable length decoding unit large block type/sub-block type system. For example, the motion vector of the motion compensated prediction sound and the reference block type/sub-block are implemented using the emotion image of the inter-plane mode using the variable length decoding unit 52. The pattern indicates that the predicted image is generated by the large direct vector generation unit 61. When the pattern is output from the large direct vector generation unit 61, the surface amount and the reference image are subjected to the motion compensation gamma processing to form the predicted image. Deconstruction section. (4) Imagery. Movement compensation processing unit Fig. 7 is a view showing a configuration of the configuration shifting unit 61. The direct vector of the member prediction unit 54 generates a picture _, and the space is directly stored in the motion vector memory 51. The generation unit performs the following processing: from the order of 'the decoding object large area=the decoded large block's motion vector The Dragon's Zhou Yuanzhi has solved the movement of the big block 322763 17 201143455 The vector is read out from the moving vector to generate a spatial direct vector of the spatial direct mode. The temporal direct vector generation unit 72 performs a process of shifting the motion vector of the decoded face that is temporally in the vicinity of the large block of the decoding object from among the motion vectors stored in the decoded large block of the motion vector memory 51. And reading a motion vector of the large block in the decoded decoding plane that is spatially located at the same position as the decoding target large block, and generating a temporal direct vector of the temporal direct mode from the motion vector. The spatial direct vector generation unit 71 and the temporal direct vector generation unit 72 constitute a direct vector generation means. The direct vector determination unit 73 performs a process of calculating the evaluation value of the spatial direct mode using the spatial direct vector generated by the spatial direct vector generation unit 71, and using the temporal direct vector generated by the temporal direct vector generation unit 72. The evaluation value of the time direct mode is calculated, and then the evaluation value of the spatial direct mode and the evaluation value of the direct time mode are compared, and either one of the spatial direct vector or the temporal direct vector is selected. Fig. 8 is a view showing the configuration of the direct vector determining unit 73 constituting the direct vector generating unit 61. In the eighth diagram, the motion compensation unit 81 performs a process of generating a forward prediction video and a backward prediction video in the spatial direct mode using the spatial direct vector generated by the spatial direct vector generation unit 71, and using the temporal direct vector generation unit 72. The generated temporal direct vector generates a forward predicted image and a backward predicted image in a temporal direct mode. The similarity calculation unit 82 performs a process of calculating the similarity between the forward predicted image and the backward predicted image in the spatial direct mode to calculate the similarity between the forward mode and the backward predicted image, and the similarity between the forward predicted image and the backward predicted image in the temporal direct mode. Use it as the evaluation of the time direct mode. Means movement compensation unit 81 and similarity calculation (4) 82_ into evaluation value calculation The direct vector selection unit 83 calculates the following virtual, Tian., 82: forward prediction in the spatial direct mode: ==

The degree of similarity of the image, and the comparison of the similarity of the M predicted images in the direct mode of time, and the selection of the front image of the image and the rear vector, and the following: = = ίΠί direct inward. The direct vector selection unit 83 constitutes a direct vector in the first drawing, and the video encoding device is subjected to a motion compensation prediction unit 2, a subtractor 3, a bat# component, that is, a local gamma unit 6, an adder 7, and a ^^^ Fixed part 4, _Miyazaki's wave device 8 and variable i code part. It is assumed that the hardware is dedicated (for example, a guided volume circuit or a single-chip microcomputer), but the image coding device is constructed by a computer. The processing contents of the motion compensation prediction unit 2, the adder/pattern determination (4) 4, the compression unit 5, the local decoding unit 6, the process (4) wave device 8 and the variable length coding unit 1G are stored in the read 3 == record ^ 'By the CPU of the brain, 322763 19 201143455 Process==Fig. 5 of the video encoding device according to the first embodiment of the present invention _ 'The video decoding device variable length decoding unit decoding unit is an element, that is, 5[Adder 55 and back (four) wave device % set to shift _ compensation pre-clear part = such as: installed m; semiconductor integrated circuit; ^ by a dedicated hardware 荨), but the child is strong or The early wafer microcomputer has the following form: if it is composed of a computer, it can be formed as a process of 53, a motion compensation pre-, a =I decoding unit 52, and a lion error solution; == adder 55 Loop filter

The operation of the video decoding device according to the first embodiment of the present invention, which is stored in the memory, is described below. The processing content of the image coding device of Fig. 1. When the wheel is in, the player is °卩β2, and the dynamic image sub-block 2 divides the motion image signal into a large area in units of pivots (or sub-blocks) of the motion picture signal indicating the input image. Block (or mobile compensation _ use - above the frame and then select a frame of the reference image in the large area ^ ^ image, ^ Cheng Xin two for the second): =, processing 'pre-rigid image. a motion vector of the human block, and further a 322763 20 201143455 motion compensation prediction unit 2 generates a motion vector of a coding target large block (or a secondary block) and generates a predicted image, and outputs the predicted image to the subtractor 3, and The motion vector used in the generation of the predicted image, the large block type/sub-block pattern (including, for example, the coding mode used in the large block (or the secondary block) is an inter-picture mode or a direct mode. The information of the reference image and the identification number of the reference image are output to the encoding mode determining unit 4. Hereinafter, the processing content of the motion compensation prediction unit 2 will be specifically described. However, for convenience of explanation, it is assumed that the motion vector is generated in units of large blocks to generate a predicted image. When the motion vector search unit 21 of the motion compensation prediction unit 2 receives the information indicating that the coding mode is the inter-plane mode (for example, receiving information indicating that the inter-plane mode is used from the outside), the motion search mode is the most searched by the inter-plane mode. A suitable motion vector is output to the motion compensation processing section 23. The process of searching for the most suitable motion vector in the inter-plane mode is a well-known technique, and a detailed description thereof will be omitted. When receiving the information indicating that the encoding mode is the direct mode, the direct vector generating unit 22 of the motion compensation predicting unit 2 generates the spatial direct vector of the spatial direct mode and the time of the temporal direct mode for each large block of the encoding target. The vector is output to the motion compensation processing unit 23 as any one of the spatial direct vector or the temporal direct vector as a motion vector. That is, the spatial direct vector generation unit 31 of the direct vector generation unit 22, from among the motion vectors stored in the coded large block of the motion vector memory 1, the coded large block located around the large block of the coding target. The motion vector is read out, and the space of the spatial direct mode is generated from the motion vector. 21 322763 201143455 Direct vector (step ST1 in the flute, 12). From the time of the storage, the direct vector generation unit 32 of the direct vector generation unit 22: • the motion vector of the coded large block of the motion vector memory 1 must be &: in the vicinity of the large block of the coding object in time The motion vector of the encoded face, 0, and the motion vector of the large block in the encoded face that is spatially located at the same position as the coded block is read, from which the time direct mode is generated. Time direct vector (step ST2). Figure 9 is a schematic diagram showing a method of generating a motion vector (time direct vector) in a temporal direct mode. The false μ is, for example, a case where the large block ΜΒ 1 in the face Β 2 is the coding target large block ’ and the large block ΜΒ 1 is to be coded in the time direct mode. In this case, the coded 昼 located at the rear of the facet Β 2 on the time axis is used, and the motion vector closest to the face Ρ 3 of the face Β 2 is located, and the motion vector is spatially located in the face Ρ 3 The motion vector mv of the large block ΜΒ2 at the same position as the large block leg. The motion vector MV is referred to the pupil plane ρ〇, and the motion vectors MVLO, MVL1 to be used when encoding the large block Μβ1 are obtained by the following equation (3). MVL0 = MVL\^ T2rT0 Π~Γ3 Τ^ΤΟ

XMV

The MV 3 time directly calculates the motion vector MVL 〇, MVU from the 篁 generating unit 32, and outputs the motion vector MVLO, MVL1 as the temporal direct vector of the temporal direct mode to the direct vector determining unit 33. 322763 22 201143455 The time direct vector generation unit 32 can be used as described above in the ninth method, and other methods can be used: no. 11. 264 mode, but not limited to direct = map A schematic diagram showing a method of generating a mobile vector gate direct vector in a spatial direct mode. In the two (two: two pictures, current MB represents the large block of the encoding object. At this point, 'Assume that the moving vector of the large block A of the stone object is MV : = the edited area of the side has been compiled, ah, The upper right side of the coding object block, and the motion vector of the block C is MVc, then the median value of the media MVa, MVb, MVc (M) is as follows in the following equation (4). Calculate the motion vector MV of the large block of the encoding target. med 1 a Π (Mv a, MVb, MVc) (4) In the spatial direct mode, the motion vector is obtained for the front and the rear, respectively, but either side The inter-I direct vector generation unit 31 calculates the forward and backward motion vectors MV' as the spatial direct vectors of the spatial direct mode and outputs them to the direct vector determination as described above. Part 33. The generation method of the spatial direct vector in the spatial direct vector generation unit 31 may be as described above with reference to the 264 method as shown in Fig. 1, but is not limited thereto. Other methods may be employed. Direct vector determination unit 3 of vector generation unit 22 3. When the spatial direct-quantity generation unit 31 generates the spatial direct vector, the spatial direct mode evaluation value is calculated directly from the space. 23 322763 201143455 Further, the direct vector determination unit 33 generates the time in the temporal direct vector generation unit 32. When the direct vector is used, the time direct vector is used to calculate the evaluation value of the temporal direct mode. The direct vector determining unit 33 then performs the comparison of the evaluation value of the spatial direct mode with the evaluation value of the temporal direct mode, and directly vector or from the space. Among the time direct vectors, the direct vector of the direct mode is determined by the determination means described below and output to the motion compensation processing unit 23. Hereinafter, the processing content of the direct vector determining unit 33 will be specifically described. When the spatial direct vector generation unit 31 generates the spatial direct vectors MVLO and MVL1, the spatial direct vector generation unit 31 generates the forward prediction image fspatial of the spatial direct mode using the spatial direct vector MVL0 as shown in FIG. And use the spatial direct vector MVL1 to generate the back of the spatial direct mode The image gSPatial (step ST3). When the temporal direct vector generation unit 32 generates the temporal direct vector (i.e., the front and rear motion vectors MV), the motion compensation unit 41 uses the forward motion as shown in FIG. The vector MV generates the forward predicted image f^poral in the temporal direct mode, and generates the backward predicted image gte^ral in the temporal direct mode using the rear motion vector MV (step ST4). The similarity calculating unit of the direct vector determining unit 33 42. When the motion compensation unit 41 generates the forward prediction video fSPatial and the backward prediction image gspatial in the spatial direct mode, the similarity between the forward prediction video f spatial and the backward prediction image gspatial is calculated and used as the evaluation value of the spatial direct mode SADspatial (Step ST5) 0 24 322763 201143455 SADSpatial — If spatial δ spatial I (5) Further, the similarity calculation unit 42 generates a forward predicted image f时间 in the temporal direct mode in the motion compensation unit 41. ... and the rear predicted image gtemp (3ra丨, the similarity between the forward predicted image f temp〇ral and the backward predicted image gtemp() ral is calculated and used as the evaluation value SADte^ra! of the time direct mode (step ST6) SADlenlporal — Iftcmporal Stemporal I (6) The greater the difference between the forward predicted image and the backward predicted image, the lower the similarity between the two images (indicating that the absolute value of the difference between the two images and the evaluation value SAD become larger) The temporal correlation becomes lower. Conversely, the smaller the difference between the forward predicted image and the backward predicted image, the higher the similarity between the two images becomes higher (indicating the absolute value of the difference between the two images and the evaluation value SAD becomes The direct vector selection unit 43 of the direct vector determination unit 33 calculates the evaluation value SADspatial of the spatial direct mode and the evaluation value SADte^radf of the temporal direct mode by the similarity calculation unit 42. Comparing the evaluation value SADSpatial with the evaluation value SADtemporal' to compare the similarity between the forward prediction image f spatial and the backward prediction image gspatial in the spatial direct mode, and the direct mode of time The degree of similarity of the forward predicted image fteDlpc)ral&back predicted image gtelnpc)ral in the equation (step ST7). The direct vector selection unit 43 has a higher degree of similarity between the forward predicted image f SPatial and the backward predicted image gspatial in the spatial direct mode than the forward predicted image f temporal and the backward predicted image gtemp 〇raI in the temporal direct mode ( The three persons 0_131$540_-) select the spatial direct vector generated by the spatial direct vector generating unit 25 322763 201143455 3i, and output the spatial direct vector as the motion vector to the motion compensation processing unit 23 (step ST8). On the other hand, in the time direct mode, the forward predicted image f_ai and the rear predicted image like gteinp (the similarity of the 3I_al similarity in the tb spatial direct mode is higher in the forward predicted image with 1 and the rearward image gs_ ( SADSPatial>SADtempQral) selects the time direct vector generated by the temporal direct vector generation unit and outputs the temporal direct vector to the motion compensation processing unit 23 as a motion vector (step ST9). The mobile complement processing unit 23 encodes When the mode is not the direct mode (step stio), if the motion vector transmitted from the motion vector search unit 21 is received, the movement is performed using the reference image of the _ frame stored in the frame memory 9 using 4 movements. On the other hand, when the coding mode is the direct mode (step ST10), the motion vector transmitted from the direct vector generation unit 22 (direct vector selection unit/3) is generated. The selected direct vector is used to implement the motion compensation prediction process using the motion vector and the reference image stored in the frame of the image Test image (step ST12). The motion compensation prediction process of the motion compensation processing unit 23 is a well-known technique, and detailed description thereof will be omitted. Here, the similarity calculation unit 42 calculates the absolute value of the difference as the three-level estimate SAD, and the direct vector selection unit (5) performs the evaluation value (10): 1 匕 compared to the form 'but the similarity calculation unit may be used. 4 2 Calculate the sum of the squares of the differences between the predicted images of the forward predictive influences as the evaluation values, and 322763 26 201143455 The vector selection unit 43 compares the squares of the differences and the SSE. Although the use of SSE increases the throughput, the similarity can be calculated more correctly. When the motion compensation prediction unit 2 generates the predicted video, the subtracter 3 calculates the difference video between the predicted video and the input video, and outputs the predicted difference signal indicating the difference image to the encoding mode determining unit 4 (step ST13). Each time the coding mode determination unit 4 receives the prediction difference signal transmitted from the subtractor 3, the prediction efficiency of the prediction difference signal is evaluated, and among the at least one prediction difference signal output by the subtractor 3, The predicted difference signal with the highest prediction efficiency is selected. The process of evaluating the prediction efficiency of the prediction difference signal in the coding mode determination unit 4 is a well-known technique, and detailed description thereof will be omitted. The coding mode determination unit 4 selects the most significant prediction difference signal, and the coding mode information includes the motion vector and the region used in the generation of the prediction image of the motion compensation prediction unit 2 related to the prediction difference signal. The block type/sub-block type (including, for example, information indicating whether the coding mode used in the large block is the inter-picture mode or the direct mode), and the identification number of the reference picture are output to the variable length coding. Department 10. Further, the coding mode determination unit 4 outputs the prediction difference signal having the highest prediction efficiency to the compression unit 5 (step ST14). The coding mode determination unit 4 includes the motion vector used in the generation of the predicted video in the coding mode information, and outputs the coding mode information including the motion vector to the case where the coding mode is the inter mode. The variable length coding unit 10, but the case where the coding mode is the direct mode, 27 322763 201143455, does not include the generation of the predicted image in the coding mode information, but does not include the mobile vector information of the current use. Output to the variable length coding unit 1〇. The coding mode compression unit 5 of the motion vector receives the signal from the coding mode determining unit 4, and performs the prediction difference of the DCT at the prediction difference signal.

The coefficient, and the DCT coefficient is /κ processed to calculate the DCT to the singulation [the step compression unit 5 will be the quantized DCT coefficient =). Part decoding unit 6 and variable length coding unit 10 ^

When the local decoding unit Θ receives the compressed data from the compression unit 5, the compressed data is inverse quantized to obtain d, and the coefficient is subjected to inverse DCT processing to calculate the tilt and the predicted difference signal of the DCT. Prediction error signal. And the determination unit 4 rotates the adder 7 to decode the = error signal in the local decoding unit 6, and adds the predicted signal of the shaved image to the image indicating that the motion compensation_part, ==, to generate the p 2 The generated pre-local decoded video signal. The loop filter 8 that does not locally decode the image is used for the following: the output material is decoded by the image signal (4), and the adder 7 = compensates for the local &amplitude distortion after the compensation of the code, and the image of the drink is used as the reference material =^ The variable length length of the _ indicated by the _ is stored in the figure. The material is 'compressed drinking mode information transmitted to the constricted data contraction section 5 (the block type "the bat type output by the dynamic compensation prediction unit 2 is the inter-plane mode"), the second block type, the motion vector (the stone) Mamo/), the identification number of the reference image] is encoded 322763 28 201143455 code, and the generation indicates that the encoding result _ stream is rotated (step sti6). The bit stream is streamed, and then the bit string is followed by 'description number The shadow of the figure 5 is used by the variable length decoding unit 52 to decode the processing contents of the device. When the bit stream of the output bit is rounded up! From the scene of the first picture > the encoding device code to decode the compressed data and edit Then, the entropy deblocking pattern and the motion vector are performed on the bit stream (the bat code mode 1 type information [large block type/sub-region image identification number], and then the 'minute, 4-face mode mode'), reference The code portion 53 transmits the attack code mode information (four) material to the step π? 1) in the prediction error solution 13; §; to the mobile complement prediction unit 54 (the compression received by the first prediction error decoding unit 53) When the data is obtained, J is transmitted from the variable length decoding unit 52.

Then, the DCT of the DCT system and the number of 4 are obtained, and the DCT shows the difference image of the Guanxian shout (with the HDGT processing, the pre-image mode determination unit 4 that calculates the _ difference of the output of the table is calculated. The compensation prediction unit 54^=1^step coffee). When the sigh is lost, the sigh is transmitted from the sizable length decoding unit 52: the identification number of the image is displayed from the reference image stored in the frame memory Μ: read out. In addition, when the motion compensation prediction unit 54 receives the large block type μ block type handle transmitted from the variable length debunking unit 52, the financial phase large block type/subblock type, and the image of the first picture is discriminated. The encoding device uses the inter-panel mode as the editing mode or the direct mode as the encoding mode (step ST23). , 322763 29 201143455

The motion compensation predicting unit 54 generates the predicted video G as the motion vector outputted in the encoding mode and the above-described processing in the flute 1-interface mode. On the other hand, when the video encoding apparatus of the first image is used as the encoding mode, the motion compensation unit 54 moves. The compensation prediction unit 2 generates a spatial direct vector and a time direct quantity, and then selects the spatial direct vector or the time direct vector.

The motion compensation prediction process is performed to generate a predicted image (step ST25). The processing contents of the motion compensation prediction unit 54 will be specifically described below. The direct vector generation unit 61 of the motion compensation prediction unit 54 generates a spatial direct space for each decoding target large block when the large block type/sub-block pattern output by the variable-length decoding unit 52 indicates the used direct mode. The spatial direct vector of the pattern and the temporal direct vector of the temporal direct mode are output to the motion compensation processing unit 62 as either of the spatial direct vector or the temporal direct vector. °, from the movement of the decoded large block stored in the vector memory 51, that is, the spatial direct vector generation unit of the direct vector generation unit 61? 1 Quantity: Read out the large block of decontamination located around the large block of the decoding object ^ Move to # Generate a spatial direct mode ancient vector according to the motion vector. However, the space in the 'spatial direct vector generation unit 71 is directly directed to the method, and the space in the spatial direct vector generation unit 31 of the fourth generation of the Wanxing diagram is generated directly to the 322763 30 201143455. The method is the same, so the detailed description will be omitted. The temporal direct vector generation unit 72 of the direct vector generation unit 61 is from the motion vector stored in the decoded large block of the motion vector memory 51, and is decoded in the vicinity of the decoding target large block in time. a motion vector, and for reading a motion vector of a large block spatially located at the same position as the decoding object large block in the decoded plane, a time direct vector of the temporal direct mode is generated from the motion vector. However, the method of generating the temporal direct vector in the temporal direct vector generating unit 72 is the same as the method of generating the temporal direct vector in the temporal direct vector generating unit 32 of Fig. 3, and therefore detailed description thereof will be omitted. When the spatial direct vector generation unit 71 generates the spatial direct vector, the direct vector determination unit 73 of the direct vector generation unit 61 calculates the evaluation value of the spatial direct mode using the spatial direct vector. Further, when the time direct vector generation unit 72 generates the time direct vector, the direct vector determination unit 73 calculates the evaluation value of the time direct mode using the time direct vector. Then, the direct vector determining unit 73 compares the evaluation value of the spatial direct mode with the evaluation value of the temporal direct mode, and selects a direct vector of the appropriate direct mode among the spatial direct vector or the temporal direct vector and outputs it. To the motion compensation processing unit 62. Hereinafter, the processing content of the direct vector determining unit 73 will be specifically described. When the spatial direct vector generation unit 71 generates the spatial direct vectors MVLO and MVL1, the spatial direct vector generation unit 71 generates the first 31 of the spatial direct mode using the spatial direct vector MVL0 as shown in FIG. 11 . The square predicts the image fspatial 'and uses the spatial direct vector MVL1 to generate the spatially direct mode rear prediction image gspatiai. Further, when the temporal direct vector generation unit 72 generates the temporal direct vector of the motion vector MV of the front and the rear, the motion compensation unit 81 generates the temporal direct mode using the forward motion vector MV as shown in FIG. The forward-predicted video fteinp (3ral' and the rear-predicted video using the rear motion vector MV to generate the temporal direct mode gtemporal ° The similarity calculating unit 82 of the direct vector determining unit 73 generates the forward-predicted video of the spatial direct mode in the motion compensating unit 81. In the fspatial and the backward prediction image gspatial, the degree of similarity between the forward predicted image f spatial and the backward predicted image gspatial is calculated and used as the evaluation value SADspatU1 of the spatial direct mode, similarly to the similarity calculating unit 42 of Fig. 4 . When the motion compensation unit 81 generates the forward predicted video f temporal and the backward predicted video gtemporal in the temporal direct mode, the similarity calculating unit 82 calculates the forward predicted video f teap〇 as in the similarity calculating unit 42 of the fourth figure. Ral and rear prediction image gtemp (3ral similarity and use it as time direct mode Valuation SADtenp(>fal. The larger the difference between the forward predicted image and the backward predicted image, the lower the similarity between the two images (indicating the absolute value of the difference between the two images and the evaluation value SAD 1_ is large) The temporal correlation is lower. Conversely, the smaller the difference between the W images predicted by the latter prediction image, the higher the similarity between the two images (representing the two; the absolute value of the difference between the image and the evaluation) When the value (10) is small, the temporal correlation is increased. The direct vector selection unit 83 of the direct orientation determination unit 73 calculates the evaluation value SADspatial of the spatial direct mode and the temporal direct mode in the similarity 322763 32 201143455 calculation unit 82. When evaluating the value SADtemporal, it is compared with the direct vector selection unit 43 of Fig. 4, and the forward predicted image in the spatial direct mode is compared by comparing the evaluation value SADspatial with the evaluation value SADtemporal '; f spatial and the rear predicted image gspatial Similarity, similarity to the forward predicted image f temporal and the backward predicted image gtemp() ral in the temporal direct mode. When the forward predicted image fspatial and after in the spatial direct mode When the similarity degree of the predicted image gspatial is higher than the similarity between the forward predicted image f tenoral and the backward predicted image gtempcral in the temporal direct mode (SADspatial $ SADtemporal), the direct vector selecting unit 83 selects the generated by the spatial direct vector generating unit 71. The spatial direct vector is output to the motion compensation processing unit 62 as a motion vector. On the other hand, 'the similarity between the forward predicted image ftempc> rai and the backward predicted image gtemp() ral in the temporal direct mode is higher than the similarity between the temporal predicted image f spatial and the rear predicted image gspatiai in the spatial direct mode. When (SADspatiai > SADte, ral) ', the time direct vector generated by the time direct vector generating unit 72 is selected, and the time direct vector is output as the motion vector to the motion compensation processing unit 62. The motion compensation processing unit 62 uses the motion vector output from the variable length decoding unit 52 when the large block type/secondary block pattern output from the variable length decoding unit 52 indicates that the use of the presentation format is used. The reference image of the frame in the frame memory 9 (the reference image indicated by the identification number output by the variable length decoding unit 52) performs motion compensation prediction processing to generate a predicted image. 322763 33 201143455 大 巴 另 在 在 在 在 在 在 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变 可变Fig. ==? The identification number output by the length decoding unit 52 is performed), and the motion compensation prediction process is performed to generate a prediction shadow technique. The ==::: compensation prediction process is a well-known sum S evaluation irsrtr calculation unit 82 Calculate the absolute value of the difference == The square of the difference between the measured images and the 8 value. Using the difference image indicated by the evaluation number of the similarity calculation unit 4M, the decoded image signal of the undecoded image (corresponding image) is generated by adding the image from the image of the first image by adding the u_error. ^: The local decoding loop filter 56 outputted by the device 7 compensates for the distortion of the decoded distortion image after the encoding of the encoded image signal contained in the decoded image signal t, and then compensates the image. The decoded image stored in the picture:::: is taken as the parameter 57, and the received (four) material image is rotated 322763 34 201143455 to the outside (step ST27). As can be understood from the above description, it is constructed according to the first embodiment. In the form of the following components: the direct vector generation unit 22 generates a spatial direct vector of the spatial direct mode according to the motion vector of the coded large block located around the large block of the coding object, and is based on the time of the coding object. A time direct vector of the temporal direct mode is generated by the motion vector of the coded face near the block; the similarity calculation unit 42 uses the space generated by the direct vector generation unit 22 The evaluation value of the spatial direct mode is calculated by the direct vector, and the evaluation value of the temporal direct mode is calculated using the direct vector of the time; and the direct vector selection unit 43 compares the evaluation value of the spatial direct mode calculated by the similarity calculation unit 42 and With respect to the evaluation value of the time direct mode, one of the spatial direct vector or the temporal direct vector is selected, and the motion compensation processing unit 23 performs motion compensation prediction processing using the direct vector selected by the direct vector selecting unit 43 to generate a prediction. Since the image is imaged, it is possible to select the most suitable direct mode in units of large blocks, thereby producing an image encoding device capable of obtaining an encoding which can avoid unnecessary moving vectors and preventing an increase in the amount of encoding. According to the first embodiment, the configuration is such that a spatial direct vector of the spatial direct mode is generated from the motion vector of the decoded large block located around the large block of the decoding target, and the decoding target is temporally decoded. When the motion vector of the decoded face near the large block is generated The temporal direct vector of the direct mode; the similarity calculating unit 82 calculates the evaluation value of the spatial direct mode using the spatial direct vector generated by the direct vector generating unit 61, and uses the time direct vector to calculate 35 322763 201143455 The evaluation value of the mode; and the direct vector selection unit 83 selects the spatial direct vector or the time straight vector by the evaluation value of the spatial direct mode calculated by the similarity calculation unit 82 and the evaluation value of the direct mode. The motion compensation processing unit 62 performs the motion compensation prediction process using the direct vector selected by the direct vector selection unit 83 to predict the video, so that it is possible to obtain the most suitable one that can be selected in units of large blocks. The effect of the coded data solution device in the direct mode mode. Embodiment 2 In the first embodiment described above, the similarity calculation unit 4 converts the similarity between the front paste image (5) and the second image: g_al in the direct space mode and uses it as a space. The direct mode SADsPatia1, ^ external calculation (four) direct mode front prediction ancient value and the rear prediction image g - similarity and used as the time = the sweat valuation SAD - the form, however, can also be set to Work: the dispersion value of the motion vector of the coded large block =::: block around the large block of the coding object (decoding object) = the evaluation value of the direct mode, and additionally calculate the image in the opposite direction Block (decoded large block);,, flat code alignment, spatially located and edited: 2: face (decoded face) The dispersion value of the vector ·::=== In the prefecture, the W(10) canning/stiffing type, that is, the similarity calculating unit 4U2 is as shown in Fig. 14 (4) two 322763 36 201143455. The encoded large block located around the large block of the encoding target (decoding target) (decoded large area) The dispersion value σ (spat ia 1) of the motion vector of the block is used as the evaluation value of the spatial direct mode Referring to the following formula (7)), in this way, the similarity between the forward predicted image fspatial and the backward predicted image gspatial of the spatial direct mode is calculated and used as the evaluation value of the spatial direct mode SADspatial. As shown in FIG. 14(b), the calculation units 42 and 82 calculate the spatially encoded side (decoded face) in the vicinity of the large block to be encoded (the decoded large block). The dispersion value σ (temporal) of the motion vector of the coded large block (decoded large block) around the large block at the same position as the coding target large block (decoded large block) is used as the temporal direct mode The evaluation value (refer to the following equation (7)) is used in place of the similarity between the forward predicted image f bporai and the backward predicted image gtemporal of the temporal direct mode and used as the evaluation value SADtemporal of the temporal direct mode. (w)=Go Σ One Two-)2,#= 4 (7)

^ ieR where MVm,i represents the surrounding motion vector m representing the average of the surrounding motion vectors. In addition, m is a symbol indicating spatial or temporal. The direct vector selecting sections 43, 83 perform a comparison between the dispersion value σ (spatial) of the motion vector and the dispersion value σ (temporal) of the motion vector, and the dispersion value σ (spatial) of the motion vector is larger than the dispersion value σ of the motion vector. When (temporal) is large, it is judged that the motion vector of the spatial direct mode (space straight 37 322763 201143455) is less reliable, and the motion vector of the direct direct mode (time direct vector) is selected. On the other hand, when the dispersion value σ (temporal) of the motion vector is larger than the dispersion value σ (spatial) of the motion vector, it is determined that the reliability of the motion vector (time direct vector) of the temporal direct mode is low, and the selection is low. The motion vector (spatial direct vector) of the spatial direct mode. Embodiment 3. In the first embodiment described above, when the coding mode is the direct mode, a spatial direct vector and a temporal direct vector are generated in units of large blocks, and a spatial direct vector or a temporal direct vector is selected, and then The selected direct vector is used to generate the predicted image. However, it can be set to, for example, only the direct mode switching flag "direct_spatial_mv_pred_flag" contained in the slice header is expressed as "meaningless" (for example, " When 0"), the same prediction image generation processing as in the first embodiment is performed, and when the direct mode switching flag indicates "meaningful" (for example, "Γ or "2"), the direct mode switching flag is selected. The direct vector of the indicated direct mode (for example, when the flag = 1, the spatial direct vector of the spatial direct mode is selected, and when the flag = 2, the time direct vector of the direct direct mode is selected). The processing content of the third embodiment. Here, for convenience of explanation, it is assumed that if the fragment header contains The mode switching flag is "0", and the same prediction image generation process as that of the above-described first embodiment is performed (the spatial direct vector or the temporal direct vector is selected in units of large blocks). 38 322763 201143455 If the direct mode switching flag is “Γ, then a spatial direct vector of spatial direct mode is selected for all large blocks in the segment. If the direct mode switch flag is "2", the time direct vector of the time direct mode is selected for all large blocks in the segment. In the third embodiment, in the case where the direct mode switching flag is "Γ or "2", the segment is set as one unit, and the direct vector is selected as the selection space or the time direct vector is selected, but it is not limited thereto, and may be set to For example, the picture is taken as a unit or the sequence is a unit, and the mode is switched to the selection of the direct vector of the space or the form of the direct vector of the time. Furthermore, in the third embodiment, the direct mode switching flag is used. The three states ("0", "Γ, "2") are described, but are not limited thereto, and may be set, for example, when the direct mode switching flag indicates only 0N (meaningful) or OFF (meaningless), Then enter the form of additional flags (such as profile information, mandatory flags (constraint_set_flag), etc.). That is, in the case where the direct mode switching flag is OFF, the same predicted image generation processing as in the first embodiment described above is performed. On the other hand, when the direct mode switching flag is 0N, for example, when the extra flag is "0", the spatial direct vector of the spatial direct mode is selected, and when the extra flag is "Γ, the time direct mode is selected. Fig. 15 is a view showing a configuration of a video encoding apparatus according to a third embodiment of the present invention. In the figure, the same reference numerals as in the first embodiment denote the same or equivalent parts of 39 322763 201143455. The motion compensation prediction unit 11 performs a process similar to the motion compensation prediction unit 2 of Fig. 1 in the direct mode switching flag "^ is Gk' included in the slice header. 'The motion compensation prediction unit n is The direct mode switching flag included in the fragment header is θ'Ί" 4, and the following processing is performed: the spatial direct mode space is generated directly inward, and the spatial direct vector is used to implement the motion compensation prediction process to generate a predicted image. Further, when the direct mode switching flag included in the slice header is "2", the motion compensation prediction unit 11 performs a process of generating the time direct mode time directly to 1, and implementing the motion compensation using the time direct vector. The prediction process is generated by the prediction process. The motion compensation prediction unit 11 is composed of the motion vector search unit 21, the direct vector generation unit 22, and the motion compensation processing unit 23 as in the motion compensation prediction unit 2 of Fig. 1 (see Fig. 2). The variable length coding unit 12 performs the following processing: the compressed data output from the compression unit 5 and the coding mode output by the motion compensation prediction unit U = (large block type/sub-block type, motion vector, reference picture recognition) Encoding =), and the direct mode switching flag is entropy encoded, and a bit stream (encoded data) indicating the result of the compilation is generated, and then the bit string machine is input to the iB variable length coding unit. I constituting a variable length coding hand Fig. 16 is a configuration diagram showing a straight portion 22 of a configuration of a motion compensation pre-u. In the figure, the same symbols as in the third diagram represent the same 322763 40 201143455 or equivalent. Partially, the description will be omitted. When the direct mode switching flag included in the slice header is "0", the direct vector determining unit 34 performs the following processing in the same manner as the direct vector determining unit 33 of the third drawing: the generated by the spatial direct vector generating unit 31 The spatial direct vector calculates the evaluation value of the spatial direct mode, and uses the temporal direct vector generated by the temporal direct vector generating unit 32 to calculate the evaluation value of the temporal direct mode, and then compares the evaluation value of the spatial direct mode with the temporal direct mode. To compare the evaluation values, select either the direct vector of the space or the direct vector of the time. When the direct mode switching flag included in the slice header is "Γ", the direct vector determining unit 34 performs the following processing: the spatial direct vector generated by the spatial direct vector generating unit 31 is selected, and the spatial direct vector is used as the motion vector. In addition, when the direct mode switching flag included in the slice header is "2", the direct vector determining unit 34 performs a process of selecting the time generated by the temporal direct vector generating unit 32. The direct vector is output to the motion compensation processing unit 23 as a motion vector. Fig. 17 is a view showing a configuration of the direct vector determining unit 34 constituting the direct vector generating unit 22. In Fig. 17, the motion compensating unit When the direct mode switching flag included in the slice header is "〇", the following processing is performed in the same manner as the motion compensation unit 41 of Fig. 4: the spatial direct vector generated by the spatial direct vector generating unit 31 is used to generate The forward predicted image and the backward predicted image of the spatial direct mode and the time generated by the use time direct vector generating unit 32 are straight 41 3227 63 201143455 The vector is used to generate the forward predicted image and the backward predicted image in the temporal direct mode. Further, the motion compensating unit 44 performs a process of generating the space generated by the spatial direct vector generating unit 31 when the direct mode switching flag is "Γ". The direct vector is output to the similarity calculation unit 45, and when the direct mode switching flag is "2", the temporal direct vector generated by the temporal direct vector generation unit 32 is output to the similarity calculation unit 45. When the direct mode switching flag is "0", the similarity calculating unit 45 performs the following processing similarly to the similarity calculating unit 42 of the fourth drawing: calculating the similarity between the forward predicted image and the backward predicted image in the spatial direct mode. It is used as an evaluation value of the spatial direct mode, and the similarity between the forward predicted image and the backward predicted image of the temporal direct mode is calculated as the evaluation value of the temporal direct mode. Further, the similarity calculation unit 45 performs a process of outputting the spatial direct vector output from the motion compensation unit 44 to the direct vector selection unit 46 when the direct mode switching flag is "Γ", and switching the flag to "2" in the direct mode. The time direct vector output from the motion compensation unit 44 is output to the direct vector selection unit 46. The motion compensation unit 44 and the similarity calculation unit 45 constitute an evaluation value calculation means. The direct vector selection unit 46 switches the flag in the direct mode. When it is "0", the following process is performed in the same manner as the direct vector selection unit 43 of Fig. 4: the similarity and time of the forward predicted image and the backward predicted image in the spatial direct mode calculated by the similarity calculation unit 45 are compared. In the direct mode, the front 42 322763 201143455 predicts the similarity between the image and the rear predicted image, and selects the direct vector of the direct mode in front of the * or time direct vector and the direct vector with higher splicing vector degree. The direct vector selection unit 46 performs the following processing: When the switching flag is "Γ", the similarity calculation unit 4 is selected to be connected to the mode direct vector. The outputted to the movement compensation processing unit 23 of the space for a flag ^ selected similarity calculating unit connected milk cut vector 2 and m of formula movement compensation processing unit 23 is "2" helmet ". = time straight 46 series constitute a direct vector selection means. Fig. 18 is a view showing a configuration of a shadow in the third embodiment of the present invention. The same symbol as in Fig. 5 indicates a part of %1, and therefore the description thereof will be omitted. The gate or equivalent 1-bit length decoding unit 58 performs a process of causing the bit stream (encoded data) output from the image encoding device to be transcoded and transcoded to output compressed data, edited, and Bit block type/sub-block type, motion vector, reference scene ". The area and direct mode switching flags are then outputted to the decoding unit 53 by the compressed data, and the coding mode subtraction direct mode is switched to the standard motion compensation prediction unit 59. The variable length decoding unit 58 constitutes a decoding means. The movement compensation processing unit 59 performs a process of reading out the reference image indicated by the identification number output from the variable length decoding unit 58 from the reference image stored above the frame of the frame memory 57, and then ^ The large block type/sub-block type output by the variable length decoding unit 58 indicates 322763 43 201143455 In the inter-face mode, the predicted image of the variable length solution is used. The motion compensation prediction unit 59 performs the motion compensation prediction processing, and the block type/sub-block pattern output from the variable length decoding unit 58 is used in the direct mode, and the variable-length decoding unit 58 is used. When the direct mode switching flag of the output is 'τ, the motion compensation prediction unit in the video encoding apparatus performs the direct vector and the time direct vector, and then selects any vector of the vector or time direct vector. The predicted image is generated by performing motion compensation measurement processing using the selected reference image and the reference image indicated by the identification code. Further, the 'movement compensation unit' 59 uses the direct mode switching flag output by the direct mode intersection length decoding unit 58 in the variable length solution=block type/subblock type pattern| The spatial direct vector is generated. Then, the reference image represented by the space direct == number is used to perform the motion to generate the predicted image. Further, the motion compensation prediction unit 59 displays the large block type/sub-block type in the variable length decoding unit. The direct mode is used, and: the direct mode switching flag outputted by the variable length decoding unit 58 υ °, real, to the child. Generate a time direct vector, and then use the time straight vector and the job code table (4) reference f彡The predicted image is generated by the motion of the wire. The motion compensation prediction unit 59 is composed of the direct vector generation unit 61 and the motion compensation processing unit 62 as in the motion compensation prediction unit 322763 201143455 54 of the sixth figure ( Fig. 6 is a view showing a configuration of a direct vector generation unit 61 constituting the motion compensation prediction unit 59. In the figure, the same reference numerals as in Fig. 7 indicate the same or equivalent. The direct vector determination unit 74 performs the same as the direct vector determination unit 73 of Fig. 7 when the direct mode switching flag outputted by the variable length decoding unit 58 is "0". The following processing: calculating the evaluation value of the spatial direct mode using the spatial direct vector generated by the spatial direct vector generating unit 71, and calculating the evaluation value of the temporal direct mode using the temporal direct vector generated by the temporal direct vector generating unit 72, Then, the evaluation value of the spatial direct mode and the evaluation value of the temporal direct mode are compared, and either one of the spatial direct vector or the temporal direct vector is selected. The direct vector determining unit 74 implements the following when the direct mode switching flag is “Γ” Processing: The spatial direct vector generated by the spatial direct vector generation unit 71 is selected, and the spatial direct vector is output as a motion vector to the motion compensation processing unit 62. Further, when the direct mode switching flag is "2", the direct vector determining unit 74 performs a process of selecting the time direct vector generated by the temporal direct vector generating unit 72 and outputting the time direct vector as a motion vector to the mobile. The compensation processing unit 62. Fig. 20 is a view showing the configuration of the direct vector determining unit 74 constituting the direct vector generating unit 61. In the 20th figure, when the direct mode switching flag of the variable length decoding unit 58 is 45 322763 201143455, the motion compensation unit 84 performs the following processing as in the motion compensation unit 81 of Fig. 8: The forward direct video and the backward predicted video of the spatial direct mode are generated using the spatial direct vector generated by the spatial direct vector generating unit 71, and the temporal direct vector generated by the temporal direct vector generating unit 72 is used to generate the front of the temporal direct mode. Predicted images and rear predicted images. Further, the motion compensating unit 84 performs a process of outputting the spatial direct vector generated by the spatial direct vector generating unit 71 to the similarity calculating unit 85 when the direct mode switching flag is "Γ", and the direct mode switching flag is In the case of "2", the time direct vector generated by the time direct vector generating unit 72 is output to the similarity calculating unit 85. The similarity calculating unit 85 similarly to the eighth figure when the direct mode switching flag is "0". The calculation unit 82 performs the following processing: calculating the similarity between the forward predicted image and the backward predicted image in the spatial direct mode as the evaluation value of the spatial direct mode, and calculating the forward predicted image and the backward predicted image in the temporal direct mode. The similarity degree is used as the evaluation value of the time direct mode. The similarity calculation unit 85 performs a process of outputting the spatial direct vector output from the motion compensation unit 84 to the direct vector when the direct mode switching flag is "Γ". The selection unit 86 outputs the direct vector of the time outputted by the motion compensation unit 84 to the straight when the direct mode switching flag is "2" The amount of selection unit 86 is connected. The motion compensation unit 84 and the similarity calculation unit 85 constitute an evaluation value calculation means. 46 322763 201143455 The direct vector selection unit 86 performs the following processing in the spatial direct mode calculated by the similarity calculation unit 85 in the same manner as the direct vector selection unit 83 of the eighth mode when the direct mode switching flag is "〇". The forward prediction ¥ image and the rear Lai image, and the prediction in the direct subtraction, like the comparison with the backward prediction image, and select the space straight = vector or time direct vector, the front prediction image and the rear In addition, the direct vector selection unit 86 performs the following processing: when the switching flag is "Γ, the similarity is calculated, and the second::? is J',:' The similarity calculation_selection is turned on and output to the motion compensation processing unit 62. The 86 series constitutes a direct vector selection means. Pick up the choice. Each of the components of the p-transformation code device, that is, the unit 5, the local decoding unit 6, and the code-making unit 12 are respectively dedicated hardware (the device 8 and the variable-length volume circuit, or Single s # # . The 裴 裴 CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU CPU Exhaustion 6, addition °. The code mode determination unit 4 and the program contents of the refinement units 5 and 12 are stored in the channel unit, and the variable length coding unit is stored in the unit. The components of the image defrosting device, that is, 322763 47 201143455 ^ variable length decoding unit 58, prediction error decoding unit 53, shift 9, adder 55, and loop filter 56 are sub-prediction units such as a semiconductor product with a CPU installed. Body circuit, or single = ^ more body (for example, the video decoding device is a computer, computer, etc.). The following form: The variable length decoding unit will be described.

53. The motion compensation prediction unit 59, the adder 55, and the program of the back and the solution are stored in the memory of the computer, and then the program stored in the memory is executed. The CPU of °Λ will then explain the action. Mingdi Chen image coding agricultural processing content. In addition to the variable length material portion 12, the processing of the motion compensation prediction unit 11 and the variable length production phase will be described. Only the length coding unit 12 compensates for the pre-clear 11 and the dynamic motion compensation prediction unit u transmits the motion picture to the secondary block in units of the wheel person image::::box=block (or sub-block). After the unit, the motion vector is selected from the block of the memory W of the frame (or the motion compensation prediction reference image of the selection - the uplink motion compensation prediction process of the frame, thereby blocking the block for each color component) Further, the pseudo-image area or the sub-area motion compensation prediction unit 11 generates a motion vector of the block "block U sub-block" 322763 48 201143455 and generates a predicted image, then the prediction picturer 3, and The edit mode used in the generation of the predicted image generation to the reduced block type/sub-block type (including, for example, indicating the large area and the large block) is an inter-panel mode or a direct mode, (or The information of the sub-area) and the identification number of the reference image are used to determine which part of the round and X is fixed. 4. The encoding mode is determined below, and the processing of the motion compensation prediction unit 具体地说 is specifically described. Convenient, suppose it is a big product The motion vector search unit 21 of the single-moving supplement predictor unit 11 displays the information that the coding mode is the inter-plane mode (from, for example, ° = received table ^ is 昼). In the case of the information of the inter-surface mode, the direct vector t-being processing unit 23 indicating the motion compensation prediction 以 is displayed between the pictures. When the coding mode is the information of the direct mode, / when the large block is received, the generation space is directly generated. The pattern parent has a bat code object, a time direct vector, and the empty f:=, and the time-directed I-side is regarded as a motion vector. I or time directly to the direct vector generation unit 2 dynamic compensation processing unit 23. In the above embodiment, the vector generation unit 31 encodes the motion vector of the coded large block of the motion vector memory of the large block; the surrounding area of the large block of the flat code object; The spatial direct vector in the spatial direct mode is generated from the motion vector. The direct vector generation unit 32 322763 49 201143455 is the same as the first embodiment described above, and is stored in the moving direction. Among the motion vectors of the coded large block of the quantity memory 1, 'the motion vector of the coded face that will be temporally near the large block of the coding object, and is spatially located in the coded facet The motion vector of the large block at the same position of the coding target large block is read, and the temporal direct vector of the temporal direct mode is generated from the motion vector. The direct vector decision unit 34 of the direct vector generation unit 22 includes the slice header. When the direct mode switching flag is "0", when the spatial direct vector generating unit 31 generates the spatial direct vector, the spatial direct vector is used to calculate the evaluation value of the spatial direct mode, and is generated by the temporal direct vector generating unit 32. When the time is a direct vector, the time direct vector is used to calculate the evaluation value of the time direct mode. Then, the direct vector determining unit 34 compares the evaluation value of the spatial direct mode with the evaluation value of the temporal direct mode, and selects the spatial direct vector or the temporal direct vector, and selects the same as the direct vector determining unit 33 of the third figure. The direct vector is output to the motion compensation processing unit 23 as a motion vector. When the direct mode switching flag is "Γ", the direct vector determining unit 34 selects the spatial direct vector generated by the spatial direct vector generating unit 31, and outputs the spatial direct vector to the motion compensation processing unit 23 as a motion vector. When the direct mode switching flag is "2", the direct vector determining unit 34 selects the time direct vector generated by the temporal direct vector generating unit 32 and outputs the time direct vector as a motion vector to the motion compensation processing unit 23. 322763 201143455 For the setting of the direct mode switching flag, for example, if the image as a whole is an input image such as panning, the direct mode switching flag is set to indicate that the time direct mode is selected as "2". If the input image is changed only in the movement direction of the face, the direct mode switching flag is set to indicate "1" which is the spatial direct mode selected. Hereinafter, the processing content of the direct vector determining unit 34 will be specifically described. When the direct mode switching flag is "0", the motion compensating unit 44 of the direct vector determining unit 34 and FIG. 4 Similarly to the motion compensating unit 41, the forward direct video and the backward predicted video in the spatial direct mode are generated using the spatial direct vector generated by the spatial direct vector generating unit 31, and the time direct vector generated by the temporal direct vector generating unit 32 is used to generate the time. When the direct mode switching flag is "Γ", the motion compensation unit 44 outputs the spatial direct vector generated by the spatial direct vector generating unit 31 to the similarity calculating unit 45, in the direct mode switching flag. When the direct mode switching flag is "2", the time direct vector generated by the temporal direct vector generating unit 32 is output to the similarity calculating unit 45. When the direct mode switching flag is "0", the similarity calculating unit 45 of the direct vector determining unit 34 calculates the similarity between the forward predicted image and the backward predicted image in the spatial direct mode, similarly to the similarity calculating unit 42 of Fig. 4 . The degree is used as the evaluation value of the spatial direct mode, and the similarity between the forward predicted image and the backward predicted image of the temporal direct mode is calculated as the evaluation value of the temporal direct mode. Further, the similarity calculation unit 45 outputs the spatial direct vector output from the motion compensation unit 44 to the direct vector selection unit 46 when the direct mode switching flag is "Γ 51 322763 201143455", and the direct mode switching flag is "2". The time vector directly rotated by the motion compensating unit 44 is output to the direct vector selecting unit 46. The direct vector selecting unit 46 of the direct vector determining unit 34 when the direct mode switching flag is "0", and the fourth figure The direct vector selection unit 43 compares the similarity between the forward predicted image and the backward predicted image in the spatial direct mode calculated by the similarity calculation unit 45, and the similarity between the forward predicted image and the backward predicted image in the temporal direct mode. A direct vector of a direct mode in which the forward predicted image and the backward predicted image are highly similar is selected, and the direct vector is output as a motion vector to the motion compensation processing unit 23. In addition, the direct vector selection unit 46 receives the received from the similarity calculation unit 45 when the direct mode switching flag is "1". The spatial direct vector input is output, and the spatial direct vector is output as a motion vector to the shift=compensation processing unit 23. The similarity calculation unit is executed by pressing the key selection unit P and restoring the flag. The time direct vector outputted by 45 is output to the motion compensation processing unit 23 as a motion vector. Material::== The corpse receives the compression block type/sub-block measurement transmitted from the compression unit 5. The coding mode information transmitted by the part U (largely for the compressed data, the encoded motion vector, the identification number of the reference image), the entropy coding, and the generation of the bit stream indicating the result of the mode switching flag entering the bat code Then, the 322763 52 201143455 bit stream is rotated. Next, the processing content of the video decoding device of the 坌 0 ^ ^ ^ 罘 18 picture is explained. The processing content complementation prediction unit 59 and the variable length resolution unit 58 of the compensation unit 59 are the variable length deciphering unit 58 that is output when receiving the image of Fig. 15 and & When the code is loaded, for this bit stream Line network solution type, mobile ^ _ type information [large block type /: under block ° 1 (encoding mode is the case of the inter-plane mode), participation = identification number] and direct mode switching flag, and then The compressed data 3 is output to the prediction error decoding unit 53, and the encoding mode information and the direct switching flag are input & rotation compensation pre-clear 59. "The motion compensation prediction unit 59 receives the reference image transmitted from the variable length decoding unit 58. The identification number ' is read from the reference image stored above the frame of one of the frame memories 57, and the reference shadow indicated by the identification number is read. Further, the 'movement compensation prediction unit 59 receives the large block type/sub-block type from the variable length decoding unit, and refers to the large block type/person block type to determine the image coding device used in FIG. The inter-plane mode is used as the coding mode or the direct mode is used as the coding mode. When the video coding apparatus of FIG. 15 uses the inter-plane mode as the coding mode, the motion compensation prediction unit 59 performs the motion compensation prediction process 322763 by using the motion vector and the reference video rotated by the variable length decoding unit 58. 53 201143455 Generates a predicted image. On the other hand, in the case where the motion compensation prediction unit 59 uses the direct mode wire as the coding mode in the device of Fig. 15, if the county, decoding: P 58 outputs the direct mode switching flag $ "〇", then: The motion compensation prediction unit 11 in the image coding apparatus of the Gth map generates the gate direct vector and the temporal direct vector in the same manner, and then selects any of the spatial direct vectors, and then uses the selected direct vector and the identification: The motion compensation_processing is performed on the non-reference image to generate a predicted shirt image. When the video encoding apparatus of Fig. 15 uses the splicing mode as the encoding mode, the smear compensation prediction unit 59 directly generates a space directly from the direct mode switching flag of the variable length material unit 58. The vector, = then uses the spatial direct vector and the reference image indicated by the identification number to perform motion compensation prediction processing to generate a predicted image. Further, the motion compensation prediction unit 59 uses the direct mode as the coding in the video coding apparatus of FIG. In the mode, if the direct mode switching flag rotated from the variable length depletion unit 58 is "2", the generation time is directly white 1 and then the reference vector represented by the direct vector and the identification number is used to implement the motion compensation prediction. The processing is performed to generate a predicted image. The details of the processing by the motion compensation prediction unit 59 will be described below. The direct vector generation unit 61 of the motion compensation prediction unit 59 outputs the large block type/sub-block in the variable length solution 4 58 . When the pattern indicates that the direct mode is used, 'the space direct mode is generated for each decoding target large block as in the above-described first embodiment. The direct vector, and the time direct vector of the direct type 54 322763 201143455 f, and one of the spatial direct vector or the temporal direct vector is output as a motion vector to the motion compensation processing unit 62. That is, the direct vector generation unit 61 The spatial direct vector generation unit 71 is the same as the above-described embodiment, from the motion vector stored in the decoded large block of the motion vector memory 51, and the large block of the bit object has been depleted. The motion vector of the block is read, and a spatial direct vector of the spatial direct mode is generated based on the motion vector. The temporal direct vector generation unit 72 of the direct vector generation unit 61 is stored in the motion vector from the above-described embodiment 1 Among the motion vectors of the decoded large block of the memory 51, 'the motion vector of the decoded face that will be temporally near the large block of the decoding object, and is the same in the decoded facet. The movement of the large block at the same position of the large block of the decoding object is read inward, and the time direct vector of the time direct mode is generated according to the fresh motion vector. When the direct mode switching flag output by the variable length decoding unit 58 is "〇", the direct vector determining unit 74 of the quantity generating unit 61 generates a spatial direct vector in the space f vector generation unit 71, and directly uses the space to the space. The estimated value of the direct space out mode is generated, and the time direct vector generating unit 72 generates a time direct vector, and uses the time direct vector to calculate an evaluation value of the time direct mode. Then, the direct vector determining unit 74 and the seventh figure The direct vector determining unit 73 compares the evaluation value of the spatial direct mode with the 砰 estimate of the temporal direct mode, selects the spatial direct vector or the temporal direct vector, and outputs the selected direct vector as a motion vector to Motion Compensation Department 322763 55 201143455 Department 62. The direct vector determining unit 74 selects the spatial direct vector generated by the spatial direct vector generating unit 71 when the direct mode switching flag is "Γ", and outputs the spatial direct vector as a motion vector to the motion compensation processing unit 62. When the direct mode switching flag is "2", the direct vector determining unit 74 selects the time direct vector generated by the temporal direct vector generating unit 72, and outputs the time direct vector as a motion vector to the motion compensation processing unit 62. Hereinafter, the processing content of the direct vector determining unit 74 will be specifically described. When the direct mode switching flag output by the variable length decoding unit 58 is "0", the motion compensation unit 84 of the direct vector determining unit 74 is the same as Fig. 8 The motion compensation unit 81 similarly generates the forward prediction video and the backward prediction video in the spatial direct mode using the spatial direct vector generated by the spatial direct vector generation unit 71, and uses the temporal direct vector generated by the temporal direct vector generation unit 72. A forward prediction image and a backward prediction image in the time direct mode are generated. Further, the motion compensation unit 8 4, when the direct mode switching flag is "", the spatial direct vector generated by the spatial direct vector generating unit 71 is output to the similarity calculating unit 85, and when the direct mode switching flag is "2", the temporal direct vector is generated. The time direct vector generated by the unit 72 is output to the similarity degree calculation unit 85. When the direct mode switching flag is "0", the similarity calculation unit 85 calculates the similarity between the forward image 562763 201143455 predicted image and the backward predicted image in the spatial direct mode, similarly to the similarity calculating unit 82 of Fig. 8 It is used as an evaluation value of the spatial direct mode, and the similarity between the forward predicted image and the backward predicted image of the temporal direct mode is calculated and used as the evaluation value of the temporal direct mode. Further, when the direct mode switching flag is "", the similarity calculation unit 85 outputs the spatial direct vector output from the motion compensation unit 84 to the direct vector selection unit 86, and when the direct mode switching flag is "2", The time direct vector output from the motion compensation unit 84 is output to the direct vector selection unit 86. The direct vector selection unit 86 performs the same as the direct vector selection unit 83 of Fig. 8 when the direct mode switching flag is "0". The similarity between the forward predicted image and the backward predicted image in the spatial direct mode calculated by the similarity calculation unit 85, and the similarity between the forward predicted image and the backward predicted image in the temporal direct mode, and the spatial direct vector or Among the temporal direct vectors, the direct vector of the direct mode in which the forward predicted image and the backward predicted image are highly similar, and the direct vector is output as a motion vector to the motion compensation processing section 62. The direct vector selecting section 86 is in the direct mode When the switching flag is "Γ", the spatial direct vector output by the similarity calculating unit 85 is selected, and the spatial direct vector is used as The motion vector is output to the motion compensation processing unit 62, and when the direct mode switching flag is "2", the time direct vector output by the similarity calculating unit 85 is selected, and the time direct vector is output as a motion vector to the motion compensation. Processing unit 62. As can be understood from the above description, according to the third embodiment, the direct mode switching flag "direct_spatial_mv_pred_flag" included in the slice header indicates "meaningless" 57 322763 201143455 (for example, 'Implementation of the predicted image generation process of the above-described embodiment, in the case where the direct mode switching flag table "1" 々 "〇,, 々法.〇有忍义" (for example, 1 or 2), Select the direct vector of the direct mode of the direct mode switching flag (for example, ' 曰 not U j such as 褀 titanium - 1 case, select the space ji joint mode " direct vector, flag = 2, select the time direct mode Time direct vector), so it can generate the effect of selecting the appropriate mode according to the calculation amount of the single line or the amount of memory that can be used, and can absorb the variation of the image coding device and the image, and encode with appropriate processing amount. And decoding., [Industrial use possibility] ... The present invention is suitable for the sale of the party to prevent the increase in the amount of coding. The present invention is applicable to a method of generating a code device for decoding a stone-like material as described above. [Simplified description of the drawings] The figure shows a method of an embodiment of the present invention. And the video decoding device (4) in the image encoding device of FIG. 1 is a block diagram showing the configuration of the silly (four) h moving__ portion 2 in the eleventh embodiment of the present invention. The direct vector generation map showing the motion compensation (fourth) section 2 shows the composition of the direct vector generation 99^ section 33. Wang Cheng. The direct vector determination of 卩22 322763 58 201143455 Fig. 5 shows the first embodiment of the present invention. Fig. 6 is a block diagram showing a configuration of a motion compensation prediction unit 54 of the video decoding device according to the first embodiment of the present invention. Fig. 7 is a diagram showing a direct vector generation unit 61 constituting the motion compensation prediction unit 54. Fig. 8 is a view showing a configuration of a direct vector determining unit 73 constituting the direct vector generating unit 61. Fig. 9 is a view showing a method of generating a motion vector in a time direct mode. Fig. 10 is a schematic diagram showing a method of generating a motion vector in a spatial direct mode. Fig. 11 is an explanatory diagram showing an example of calculation of an evaluation value based on the similarity between a forward predicted image and a backward predicted image. The figure shows a flow chart of the processing contents of the video encoding apparatus according to the first embodiment of the present invention. Fig. 13 is a flow chart showing the processing contents of the video decoding apparatus according to the first embodiment of the present invention. Fig. 14(a) and ( b) is an explanatory diagram showing an evaluation formula using the dispersion value of the motion vector. Fig. 15 is a view showing the configuration of the video encoding apparatus in the third embodiment of the present invention. Fig. 16 is a view showing the configuration of the direct vector generation unit 22 constituting the motion compensation prediction unit 11. 59 322763 201143455 The first diagram shows the configuration of the direct vector decision unit 34 constituting the direct vector generation unit 22. Figure 18 is a block diagram showing the configuration of a video decoding device in accordance with a third embodiment of the present invention. Fig. 19 is a view showing the configuration of the direct vector generation unit 61 constituting the motion compensation prediction unit 59. Fig. 20 is a view showing the configuration of the direct vector decision unit 74 constituting the direct vector generating unit 61. [Description of main component symbols] Motion vector memory: 11, Motion compensation prediction unit 1 Subtractor (quantization means) Encoding mode determination unit (quantization means) 1 Compression unit (quantization means) 1 Local decoding section 'Adder 8 9 10, 12 21 22 23 31 32 Loop filter frame memory variable length coding unit (variable length coding means) Motion vector search unit direct vector generation unit motion compensation processing unit (predictive video generation means) Spatial direct vector generation Part (direct vector generation means) Time direct vector generation unit (direct vector generation means) 322763 60 201143455 33, 34 41 ' 44 42, 45 43, 46 51 52, 58 53 54, 59 55 56 57 61 62 71 72 73, 74 81 ' 84 82, 85 83, 86 Direct vector judgment unit motion compensation unit (evaluation value calculation means) Similarity calculation unit (evaluation value calculation means) Direct vector selection unit (direct vector selection means) Moving vector memory variable length Decoding unit (variable length decoding means) prediction error decoding unit (inverse quantization means) motion compensation prediction unit adder (image phase Addition means) Loop filter (image addition means) Frame memory direct vector generation unit motion compensation processing unit (predictive video generation means) Spatial direct vector generation unit (direct vector generation means) Time direct vector generation unit (direct vector generation Means) Direct vector determination unit motion compensation unit (evaluation value calculation means) Similarity calculation unit (evaluation value calculation means) Direct vector selection unit (direct vector selection means) 61 322763

Claims (1)

201143455 VII. Patent application scope: 1. An image coding device, comprising: a direct vector generation means for generating a spatial direct space for each block constituting the input image according to a motion vector of the coded block located around the block a spatial direct vector of the pattern, and generating a temporal direct vector of the temporal direct mode based on the motion vector of the encoded face temporally near the block; the evaluation value calculation means using the direct vector generation means described above a spatial direct vector to calculate an evaluation value of the spatial direct mode, and calculate an evaluation value of the temporal direct mode using a temporal direct vector generated by the direct vector generating means; the direct vector selection means performs the above evaluation Selecting either the spatial direct vector or the temporal direct vector by comparing the evaluation value of the spatial direct mode calculated by the value calculation means with the evaluation value of the temporal direct mode; and predicting the image generating means using the direct vector selecting means Selected Performing a motion compensation prediction process to generate a predicted image; a quantization means quantizing the difference image between the predicted image generated by the predicted image generating means and the input image, and then dividing the difference image The quantization coefficient is output; and the variable length coding means performs variable length coding on the quantized coefficient output by the quantization means, and outputs the coded data of the quantization coefficient. 1 322763 201143455 2. If the image coding device of the patent application Fanyuan is used, the Yin evaluation value calculation means is to generate the spatial direct mode by using the direct vector generated by the direct vector generation means. The forward predicted image and the backward predicted image are similar to the calculated forward predicted image and the backward predicted image, and are used as evaluation values of the spatial direct mode. On the other hand, the generated vector is generated by using the direct vector. The time direct vector' is used to generate the forward predicted image and the rear pre-image of the temporal direct mode, and the similarity between the forward predicted image and the backward predicted image is calculated and used as a time-based evaluation value. 3. According to the image coding apparatus described in the patent application scope, the method has been developed to calculate the dispersion value of the motion vector of the block located around the coding target block, and the two are in the same position of the block = block The image coding apparatus described above, where the coded object 4 is dispersed and used as the value movement vector of the time direct mode area evaluation, such as ====1, ", the mode receives the representation When selecting the direct valuation method of the object and the flag of the person, the interview is controlled by the direct mode switching flag: == Straight and select the direct vector from the upper variable length bat code to the direct mode And the quantization mode is switched by the quantization means to perform the variable length coding 322763 2 201143455 code, and the above-mentioned direct mode switching flag and the coding data of the above quantization coefficient are output. 5. - Image decoding device The method includes: a variable length decoding means for decoding the quantized coefficients from the encoded data; and an inverse quantization means for inverting the quantized coefficients decoded by the variable length decoding means a direct vector generation means for generating a spatial direct vector of the spatial direct mode according to a motion vector of the decoded block located around the decoding target block, and according to the decoded face in the vicinity of the decoding target block in time Transmitting a vector to generate a temporal direct vector of the temporal direct mode; the evaluation value calculating means calculates the estimated value of the spatial direct mode using the spatial direct vector generated by the direct vector generating means, and generates the direct vector by using the above-described direct vector Calculating the evaluation value of the temporal direct mode by the time direct vector generated by the means; and comparing the evaluation value of the spatial direct mode and the evaluation value of the temporal direct mode calculated by the evaluation value calculation means by the direct vector selection means, And selecting the above-mentioned spatial direct vector or any of the above-mentioned temporal direct vectors; the predicted image generating means performs motion compensation prediction processing using the direct vector selected by the direct vector selecting means to generate a predicted image; and image addition hand By using the predicted image generating means, the predicted image formed by 3 322763 201143455 and the difference image represented by the inverse quantization result of the inverse decoding means are added to obtain a wheel image of the image encoding device. 6. The image decoding apparatus described in claim 5, wherein the evaluation value calculation means uses the spatial direct vector generated by the direct vector generation means to generate the direct space of the lecture space. Calculating the similarity between the forward predicted image and the rear predicted image, and calculating the similarity between the forward predicted image and the backward predicted image, and using the direct vector generation hand ^ The generated time direct vector is used to generate the square predicted image and the backward predicted image of the temporal direct mode and the similarity between the forward predicted image and the backward predicted image is calculated, and used as the evaluation value of the time straight mode. 7. The video decoding device according to claim 5, wherein the evaluation value calculation means calculates a dispersion value of a motion vector of the decoded block located around the decoding target block, and uses it as a spatial direct mode. The evaluation value, on the other hand, calculates the dispersion of the motion vector of the decoded block in the block that is spatially located at the same position as the decoding target block in the decoded picture that is temporally near the decoding target block. Value, and use it as an evaluation of the time direct mode. 8. The image decoding apparatus according to claim 5, wherein the variable length decoding means decodes the quantization coefficient and the direct mode switching flag from the encoded data; the direct vector selection means does not depend on the evaluation value The calculation means 322763 4 201143455 is used to calculate the result of the comparison of the kernel values, and the direct vector of the ^ mode represented by the genre of the lion is decoded by the variable length and decoding means. The image encoding method comprises: a direct vector generation processing step, and each of the blocks constituting the wheeled image is obliquely paired by a direct vector generation means, and a space is generated according to a motion vector of the coded block located around the block. The spatial direct vector of the direct mode generates a time direct vector of the temporal direct mode according to the motion vector of the coded face that is temporally near the secret block; the evaluation value calculation processing step, and the evaluation value calculation means is used in the above direct vector Generating a spatial direct vector generated in the processing step to calculate an evaluation value of the spatial direct mode, and calculating an evaluation value of the temporal direct mode using the temporal direct vector generated in the direct vector generation processing step; direct vector selection processing Step, the direct vector selection means compares the evaluation value of the spatial direct mode and the evaluation value of the temporal direct mode calculated in the evaluation value calculation processing step, and selects either the spatial direct vector or the temporal direct vector; Image generation processing The predicted image generating means performs the movement using the direct vector selected in the direct vector selection processing step, and performs the prediction processing to generate the predicted image. The quantization processing step, the quantization means is performed in the step of generating the predicted image. The difference image between the generated predicted image and the input image is quantized, and then the quantized image of the difference image is outputted by 5 322763 201143455; and the variable length encoding processing step, the variable length encoding means for the above quantum The quantized coefficients outputted in the processing step are subjected to variable length coding, and the encoded data of the above quantized coefficients are output. 10. An image decoding method comprising: a variable length decoding processing step, the variable length decoding means decoding the quantized coefficients from the encoded data; an inverse quantization processing step, and the inverse quantization means is to perform the variable length decoding processing The quantized coefficients decoded in the step are inversely quantized; the direct vector generation processing step, the direct vector generation means generates a spatial direct vector spatial direct vector according to the motion vector of the decoded block located around the decoding target block, and according to A temporal direct vector of a temporal direct mode is generated by temporally moving the decoded motion vector near the decoding target block; an evaluation value calculation processing step of using the space generated in the direct vector generation processing step directly a vector for calculating an evaluation value of the spatial direct mode, and calculating an evaluation value of the temporal direct mode using a temporal direct vector generated in the direct vector generation processing step; a direct vector selection processing step, and a direct vector selection method Comparing the evaluation value of the spatial direct mode calculated in the evaluation value calculation processing step with the evaluation value of the temporal direct mode, and selecting either the spatial direct vector or the temporal direct vector; 6 322763 201143455 Predictive image generation processing step, prediction The image generation means performs the motion compensation prediction process using the direct vector selected in the direct vector selection processing step to generate a predicted image; and the image addition processing step, the image addition means generates the generated image in the predicted image generation processing step The difference image represented by the predicted image and the inverse quantization result of the inverse quantization process is added to obtain a decoded image corresponding to the input image of the video encoding device. 7 322763