JP2011130265A - Motion vector detection apparatus, motion vector detection method and program - Google Patents

Abstract

When a motion vector detection apparatus performs a motion search, the motion vector detection apparatus is less susceptible to disturbances and reduces the amount of motion search computation.
A primary searcher 6 performs a primary search in units of macroblocks, and a neighborhood searcher 8 performs a neighborhood search around a primary motion vector and a zero vector obtained as a result of the primary search, for blocks smaller than the macroblock size. Do it in units. The block integrator 11 combines adjacent blocks according to the motion vector for each block, thereby determining the block size to be applied to the macroblock and determining the motion vector of each block.
[Selection] Figure 1

Description

  The present invention relates to a motion vector detection device, a motion vector detection method, and a program.

  As a moving picture coding method for compressing and coding a moving picture, MPEG (Moving Picture Experts Group) coding method, The H.264 / AVC encoding method and the like are known. As an MPEG encoding method, the MPEG-2 encoding method is used for high image quality applications such as digital broadcasting and DVD, and the MPEG-4 encoding method is widely used for low bit rate applications such as mobile videophones. . H. The H.264 / AVC encoding system is an encoding system jointly standardized by ITU-T (The International Telecommunication Union Telecommunication Standardization Sector) and MPEG, and is a new system that achieves both high image quality and high compression rate. It attracts attention.

In these moving image encoding methods, generally, motion prediction (inter prediction) is performed to reduce the code amount. In motion prediction, when a single image (a macroblock (MB) that is a part of a frame) is encoded, an image (reference image) positioned in the past or future in terms of time. Search for a macroblock to be encoded, or a region similar to the image obtained by further dividing the macroblock, and obtain a difference between the image of the region detected by the search and the image to be encoded. This is a technique for encoding, and can greatly reduce the amount of code indicating a subject that exists continuously in time. This operation of searching for an encoding target macroblock or an area similar to an image of a block is called “motion search”.
MPEG system, H.264, etc. In the H.264 / AVC format, images to be encoded are classified into three types of I pictures (Intra-coded Pictures), P pictures (Predicated Pictures), and B pictures (Bi-directed Pictures). The I picture is an image for which motion prediction is not performed, and is referred to as a P picture or a B picture. A P picture is an image in which motion prediction in one direction is performed using an I picture or P picture located in the past in time. A B picture is an image that performs bidirectional motion prediction using an I picture or P picture located in the past in time and an I picture or P picture located in the future.

12 shows MPEG-2 and H.264. 2 is a diagram illustrating an example of an image referred to in H.264 / AVC motion prediction.
In the motion prediction of MPEG-2, an image referred to by a P picture or a B picture is uniquely determined as an image located immediately before or after. The picture referred to in the motion prediction of the P picture is the immediately preceding I picture or P picture. In the example of FIG. 12A, five images P1215 to P1211 are shown in time order, P1215, P1213, and P1211 are P pictures, and P1214 and P1212 are B pictures. Among these, the P picture P1213 is referred to immediately before the motion prediction of the P picture P1211. In addition, the images referred to in the motion prediction of the B picture are the immediately preceding I picture or P picture, and the immediately following I picture or P picture. In the example of FIG. 12A, when the motion prediction of the B picture P1212 is performed, the immediately preceding P picture P1213 and the immediately following P picture P1211 are referred to.

On the other hand, H.H. In the H.264 / AVC encoding scheme, a reference image used for motion prediction can be selected from a large number of candidates in order to increase encoding efficiency.
In the example of FIG. 12B, five images P1225 to P1221 are shown in time order, P1225, P1223, and P1221 are P pictures, and P1224 and P1222 are B pictures. Among these, when performing motion prediction of P picture P25, P pictures P1225 and P1223 are selected as reference pictures from temporally forward I and P pictures (and not deleted from the reference picture memory). ing. In the example of FIG. 12B, when performing motion prediction of the B picture P24, temporally forward P pictures P1225 and P1223 and temporally backward P picture P1221 are selected. .

FIG. 2 is a diagram illustrating the size of a block serving as a unit for performing motion prediction in the H.264 / AVC encoding scheme. FIG.
In the MPEG-2 encoding method, the unit for motion prediction is limited to a macro block unit of 16 horizontal x 16 vertical pixels (or 16 horizontal x 8 pixel block units depending on the encoding mode). On the other hand, H.H. In the H.264 / AVC encoding method, the horizontal 16 × vertical 16 pixel block B1311, the horizontal 16 × vertical 8 pixel block B1321 and B1322, the horizontal 8 × vertical 16 pixel block B1331 and B1332, and the horizontal 8 × vertical 8 pixel block shown in FIG. B1341 to B1344, subblocks B1351 and B1352 of horizontal 8 × vertical 4 pixels obtained by further dividing each block of horizontal 8 × vertical 8 pixels, subblocks B1361 and B1362 of horizontal 4 × vertical 8 pixels, horizontal 4 × vertical 4 One of the division methods is selected from the pixel sub-blocks B1371 to B1374. Then, for each block or subblock, a motion search is performed to search for a region on the reference image similar to the block or subblock of the image to be encoded, and a motion vector (MV) indicating this region is obtained. It is possible to calculate and perform motion prediction for each block or each sub-block. The “motion vector” here is a vector indicating a region on the reference image searched by motion search, and is indicated by a relative coordinate from the coordinates of the block or sub-block to be encoded to the coordinates of the searched region.

  In the MPEG-2 encoding method, an area on a reference image is searched with 0.5 pixel (half-pel) accuracy in motion search. In the H.264 / AVC encoding method, a search can be performed with an accuracy of 0.25 pixels (Quarter-Pel). As a result, H.C. With the H.264 / AVC encoding method, more accurate motion prediction can be performed.

  When the reference image and the encoding target block size can be selected from a large number of candidates as in the H.264 / AVC encoding scheme, all reference images and block sizes are ideally selected in order to improve the encoding efficiency. It is desirable to perform a motion search using as candidates and determine an optimal reference image and block size. However, if motion search is performed for all reference images and block sizes, the amount of computation becomes enormous. Therefore, in a general encoder, the amount of calculation is reduced by performing motion search in two stages. For example, in the method of Patent Literature 1, first, candidates for a reference image and a block size are narrowed down by a primary search that performs a wide range of motion searches for a plurality of reference images with a limited block size. Based on the result of the primary search, a secondary search, which is a detailed motion search including other block sizes, is performed, a final reference image and a block size are selected, and a motion vector for each block is detected.

FIG. 14 is a diagram illustrating an example of a primary search performed by the method of Patent Document 1.
As shown in the figure, the motion vector search device performs a primary search only for a limited block size, for example, horizontal 8 × vertical 8 pixel blocks B1511 to B1514. In the figure, three reference images P1411 to P1413 are shown, and the motion vector search device performs a motion search with integer pixel accuracy using a horizontal 8 × vertical 8 pixel block for these reference images. Then, the motion vector search device determines a reference pixel candidate as one of the reference images P1411 to P1413 for each block, and detects an integer pixel precision motion vector for each block.

FIG. 15 is a diagram illustrating an example of a secondary search performed by the method of Patent Document 1.
As shown in the figure, the motion vector search device performs a motion search with various block sizes based on the result of the primary search.
First, for the reference image (reference image (1)) determined in the upper left block B1511 of the primary search for the horizontal 16 × vertical 16 pixel block B1521, the motion vector search device uses the motion vector (MV (1)) of this block. A motion search with an accuracy of 0.25 pixel is performed in a range of ± 0.75 pixels centering on the coordinates indicated by ().
Similarly for the other blocks B1512 to B1514 of the primary search, a motion search with an accuracy of 0.25 pixel is performed in the range of ± 0.75 pixels with respect to the determined reference image, centering on the coordinates indicated by the motion vector of the block. Do.

In the case of a horizontal 16 × vertical 8 pixel block, this block divides a horizontal 16 × vertical 16 pixel macroblock vertically into two, and the upper block B1531 includes a horizontal 8 × vertical 8 pixel block B1511 and B1512. It is in. For the upper block B1531, the motion vector search device uses the coordinates indicated by the motion vector (MV (1)) of this block for the reference image (reference image (1)) determined in the primary search block B1511 included in this block. The motion search with an accuracy of 0.25 pixel is performed in a range of ± 0.75 pixels. Further, for the other block B1512 of the primary search included in the block B1531, the determined reference image (reference image (2)) is centered on the coordinates indicated by the motion vector (MV (2)) of this block. , 0.25 pixel accuracy motion search is performed in the range of ± 0.75 pixels.
Also for the lower block B1532, the motion search is performed based on the reference image and the motion vector determined in the primary search for the primary search blocks B1513 and B1514 included in this block.

Similarly, in the case of a horizontal 8 × vertical 16 pixel block, a motion search is performed on the primary search blocks B1511 and B1513 included in the left block B1541 based on the reference image and the motion vector determined by the primary search. Also for the right block B1542, motion search is performed based on the reference image and the motion vector determined in the primary search for the included primary search blocks B1512 and B1514.
Further, for the horizontal 8 × vertical 8 pixel blocks B1551 to B1554, a motion search with an accuracy of 0.25 pixel is further performed within a range of ± 0.75 pixels based on the result of an integer pixel accuracy motion search performed by the primary search. To do.
After performing a motion search for each block size, the motion vector search device determines a combination of a block size and a motion vector that minimizes the evaluation value, based on the evaluation value obtained by each motion search.

As described above, for example, the primary search is performed only on the horizontal 8 × vertical 8 pixel block, and the search of other block sizes is performed by the secondary search, so that the calculation amount of the primary search can be reduced. However, in the primary search, in order to detect the part to be encoded and the part having higher similarity from the reference image, it is necessary to perform a motion search over a wide range on the reference image, and the block used for the primary search Even if the size is limited, the calculation amount becomes very large. Therefore, a typical encoder, an image and a reference image to be encoded, such as one of the 1 or 4 minutes 2 minutes vertical and horizontal, 1 2 ⁿ fraction (n is an integer of 0 ≦ n ≦ 4) to Reduce and perform a primary search. For example, by reducing the image to be encoded and the reference image by a factor of two both vertically and horizontally, the number of pixels to be compared can be reduced to a quarter, and a wide range is searched for motion with a smaller amount of computation. be able to. For example, when a primary search is performed using a 1 / ⁿ reduced image and a secondary search is performed using a non-reduced image, the secondary search usually includes at least 2 ^n-1 pixels around the primary search result. Perform a search.

  For example, when a primary search is performed with a reduced image that is half of both vertical and horizontal, if a search is performed in units of horizontal 8 × vertical 8 pixel blocks, the area of 4 × horizontal 4 pixel blocks (16 pixels) is displayed on the reduced image. The image to be encoded and the reference image are matched. Alternatively, in the case of performing a primary search with a reduced image of 1/4 in both length and width, if the search is performed in units of 8 × 8 pixel blocks in the horizontal direction, the area of 2 × 2 pixel blocks (4 pixels) on the reduced image. Matching will be taken. In motion search, generally, a value indicating a difference between an encoding target image and a reference image (generally, an absolute value sum of differences between pixels in a block is used) is used as a motion vector or an encoding mode. An evaluation value obtained by adding a value indicating the bit amount necessary to describe is calculated, and a motion vector that minimizes the evaluation value is selected. At this time, if a motion search is performed with a block having a small number of pixels as described above, the influence of noise or the like in the image is large, and it is difficult to accurately track the motion vector. Further, due to disturbance such as noise, a small block is easily selected, and a motion vector in a different direction for each block is easily selected. As a result, the coding cost of the motion vector increases and the overall code amount increases.

  Furthermore, when performing a motion search of an adjacent macroblock, the motion vector prediction value (predicted motion vector; PMV) does not reflect the motion of the entire screen, and moves around the motion vector prediction value. Even if it searches, an appropriate motion search cannot be performed. H. In the H.264 / AVC encoding method and the like, the difference between the motion vector prediction value and the motion vector is taken to reduce the amount of code, but the motion vector prediction value is an inappropriate value due to the influence of the disturbance. However, the amount of codes cannot be reduced, and the coding efficiency is further reduced. In particular, a macroblock of a relatively flat image is easily affected by disturbance such as noise, so that a small block is easily selected as described above, and a motion vector in a different direction is easily selected for each block. As a result, the encoding efficiency is lowered, and further, an image error is likely to be included in the decoded image at the time of encoding / decoding, resulting in a decrease in subjective quality.

Note that the “motion vector prediction value” here is a value obtained from the motion vector of each block that has already detected a motion vector adjacent to the top, top right, top left, or left of the macroblock to be encoded. It is a vector calculated based on a procedure defined by an encoding standard such as H.264 / AVC. For example, in the case of the blocks B1321 and B1322 of 16 × 8 pixels described in FIG. 13, the upper block B1321 calculates a motion vector prediction value from the motion vector of the block adjacent on the block B1321. In the lower block B1322, the motion vector prediction value is calculated from the motion vector of the block adjacent to the left of the block B1322.
The motion vector prediction value is a vector indicating an average value of the motion amount of the image of the adjacent block that has already been encoded, and the image of the block to be encoded has the same motion as the image of the surrounding blocks. , It is expected that the motion vector of the encoding target block will be approximately the same value as the motion vector prediction value. For example, H.M. When a motion vector is encoded according to the H.264 / AVC encoding standard, encoding is performed by taking the difference between the motion vector and the motion vector prediction value. For this reason, the closer the motion vector value is to the motion vector prediction value, the smaller the amount of code for expressing the motion vector.

Therefore, it is conceivable to perform the primary search in units of macroblocks. By not dividing the macroblock into blocks, the number of pixels from which the evaluation value is calculated increases, so that it is less susceptible to noise and other disturbances.
The primary search for each macroblock can be performed using, for example, a motion search method called a telescopic search method shown in paragraphs 0028 to 0032 of Patent Document 2. The telescopic search method is a method that can follow a large movement of an object on a reference image by repeating a search in a relatively small range, and can be used for a macroblock unit motion search or a block unit motion search. it can.

JP 2008-236096 A JP 2000-242554 A

  However, when the primary search is performed in units of macroblocks, only one motion vector is obtained for the entire macroblock. When performing a secondary search around the motion vector, if the search range of the secondary search is narrowed, for example, when the macro block includes a moving object image and a stationary background image, If the motions are different, there is a possibility that an appropriate motion vector cannot be detected. On the other hand, if the search range of the secondary search is widened, the possibility of detecting an appropriate motion vector increases, but the calculation amount increases. Also, regarding the selection of the block size, if a motion search is performed for various block sizes in the secondary search, the amount of calculation increases.

  The present invention has been made in consideration of such circumstances, and an object of the present invention is to provide a motion vector detection device, a motion vector detection method, and a program that are less susceptible to disturbance and that can reduce the amount of motion search operations. Is to provide.

[1] The present invention has been made to solve the above-described problems, and a motion vector detection device according to an aspect of the present invention provides a reference image for each first block obtained by dividing a macroblock of an input moving image. A motion vector detecting device for detecting a motion vector indicating a position of a region corresponding to the first block, wherein a primary motion search is performed on the macroblock to detect a primary motion vector; and A search center indicator that outputs a search center coordinate indicating a center coordinate of a range in which a secondary motion search is performed, and the macroblock is divided for each coordinate between the coordinate indicated by the primary motion vector and the search center coordinate. For each second block, a neighborhood searcher that performs a secondary motion search to detect a motion vector, and a motion vector detected by the neighborhood searcher, the second search A motion vector selector that selects a motion vector for each lock, and a block obtained by combining the second block or the second blocks adjacent to each other based on the motion vector selected by the motion vector selector; And a block integrator that determines the motion vector of the first block based on the motion vector of the second block.
The “primary motion vector” here is a motion vector obtained by a primary search, which is a motion search performed prior to a neighborhood search (secondary search), and indicates the center of the secondary search.

  [2] A motion vector detection device according to an aspect of the present invention is the motion vector detection device described above, wherein the search center coordinates are a zero vector indicating the position of the macroblock or the second block. Including a motion vector prediction value calculated from a motion vector of a block adjacent to.

  [3] A motion vector detection device according to an aspect of the present invention is the motion vector detection device described above, wherein the search center coordinates are a motion vector of a left adjacent block adjacent to the left of the macroblock, or The primary motion vector of the left adjacent macroblock adjacent to the left of the macroblock, the motion vector of the upper adjacent block adjacent to the macroblock, or the upper adjacent macroblock adjacent to the macroblock The primary motion vector is included.

  [4] A motion vector detection device according to an aspect of the present invention is the motion vector detection device described above, wherein the search center coordinate is the primary motion vector of a right adjacent macroblock adjacent to the right of the macroblock. Or including the primary motion vector of a lower adjacent macroblock adjacent below the macroblock.

  [5] A motion vector detection device according to an aspect of the present invention is the motion vector detection device described above, and calculates an index indicating a variation in pixel value based on a pixel value of the macroblock, and the macro It further includes an image feature quantity computing unit that determines whether or not the pixel value variation of the block is small, and the secondary searcher is configured such that the image feature quantity computing unit has a small variation in pixel value of the macroblock. If it is determined, the macro block is the second block.

  [6] The motion vector detection method according to an aspect of the present invention indicates the position of the region corresponding to the first block on the reference image for each first block obtained by dividing the macroblock of the input moving image. A motion vector detection method of a motion vector detection apparatus for detecting a motion vector, comprising: a primary search step for detecting a primary motion vector by performing a primary motion search for the macroblock; and a range for performing a secondary motion search. A search center instruction step for outputting a search center coordinate indicating a center coordinate; a secondary for each coordinate between the coordinate indicated by the primary motion vector and the search center coordinate; and for each second block obtained by dividing the macroblock A neighborhood search step of detecting a motion vector by performing a motion search of the first and second motion vectors detected from the motion vectors detected in the neighborhood search step A motion vector selection step for selecting a motion vector for each lock, and a block obtained by combining the second block or the second blocks adjacent to each other based on the motion vector selected in the motion vector selection step. And a block integration step of determining the motion vector of the first block based on the motion vector of the second block.

  [7] A program according to an embodiment of the present invention indicates a position of an area corresponding to a first block on a reference image for each first block obtained by dividing a macro block of an input moving image. A program for detecting a motion vector, comprising: a primary search step for performing a primary motion search on the macroblock to detect a primary motion vector; and a search center indicating a center coordinate of a range for performing a secondary motion search A search center instruction step for outputting coordinates, a secondary motion search is performed for each coordinate between the coordinates indicated by the primary motion vector and the search center coordinates, and for each second block obtained by dividing the macroblock. A neighborhood search step for detecting a motion vector and a motion vector for each second block out of the motion vectors detected in the neighborhood search step. A motion vector selection step for selecting a toll and a block obtained by combining the second block or the second blocks adjacent to each other based on the motion vector selected in the motion vector selection step. And a block integration step of determining the motion vector of the first block based on the motion vector of the second block, and causing the computer to execute the block integration step.

  According to the present invention, it is difficult to be affected by disturbance, and the amount of calculation for motion search can be reduced.

It is a block diagram which shows schematic structure of the motion vector detection apparatus in one Embodiment of this invention. In the same embodiment, it is a figure which shows the example in which the primary searcher 6 performs a primary search. In the same embodiment, it is a figure which shows the example in which the primary searcher 6 performs a primary search using a telescopic search method. In the same embodiment, it is a figure which shows the example of the motion vector obtained as a result of a primary search. In the same embodiment, it is a figure which shows the example of the search range of the secondary search which the neighborhood searcher 8 performs. FIG. 6 is a diagram illustrating an example in which the neighborhood searcher 8 obtains an optimal motion vector for each horizontal 8 × vertical 8 pixel block in the embodiment. It is a figure which shows a motion vector and a zero vector as a search center coordinate in the embodiment. It is a figure which shows the motion vector of the left adjacent block or the primary motion vector of the left adjacent macroblock, and the motion vector of the upper adjacent block or the primary motion vector of the upper adjacent macroblock as search center coordinates in the same embodiment. It is a figure which shows the primary motion vector of a lower adjacent macroblock, and the primary motion vector of a right adjacent macroblock as a search center coordinate in the embodiment. In the embodiment, an example of a motion vector and a motion vector evaluation value that the neighborhood searcher 8 outputs to the motion vector selector 10 and an example in which the motion vector selector 10 selects a motion vector and a motion vector evaluation value FIG. In the embodiment, the motion vector detection device 100 is a flowchart showing a processing procedure for detecting a motion vector for a macroblock C121 to be encoded. MPEG-2 and H.264 2 is a diagram illustrating an example of an image referred to in H.264 / AVC motion prediction. H. 2 is a diagram illustrating the size of a block serving as a unit for performing motion prediction in the H.264 / AVC encoding scheme. FIG. It is a figure which shows the example of the primary search performed with the method of patent document 1. FIG. It is a figure which shows the example of the secondary search performed with the method of patent document 1. FIG.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a configuration diagram showing a schematic configuration of a motion vector detection device according to an embodiment of the present invention.
In the figure, a motion vector detecting device 100 includes an image memory 1, a search range segmenter 2, image reducers 3 and 12, a reduced image cache memory 4, an image cache memory 5, a primary searcher 6, A search center indicator 7, a neighborhood searcher 8, an image feature quantity calculator 9, a motion vector selector 10, and a block integrator 11 are provided.
The image reducer 12 receives an input macroblock image C121 in which the input moving image is divided into macroblock units, and the encoding target macroblock image C121 is divided into half the number of pixels both vertically and horizontally. The reduced image C122 is output to the primary searcher 6.

The image memory 1 receives the input of the local decoded image C11 and accumulates the local decoded image C11. The local decoded image stored in the image memory 1 is an original image from which a reference image used for motion search is cut out. Here, the “local decoded image” is an image obtained by encoding and further decoding an input image by a moving image encoding device including the motion vector detecting device 100. The local decoded image is an image generated also by the video decoding device, and by performing motion search using the local decoded image, image errors due to encoding and decoding can be suppressed. Note that the image memory 1 may store the input image instead of the local decoded image. When the motion search is performed using the input image, the motion search can be performed without waiting for the generation and transfer of the local decoded image, so that the degree of freedom of the configuration of the moving image coding apparatus is increased.
Although an image error occurs due to encoding and decoding, it is not necessary to generate a decoded image on the moving image encoding device side, and the configuration of the moving image encoding device can be simplified.

The search range extractor 2 reads the image C12 stored in the image memory 1, and a range set around the encoding target macroblock as a range for which the primary searcher 6 and the neighborhood searcher 8 are subject to motion search. Are cut out and output to the image reducer 3 and the image cache memory 5. For example, when the encoding target macroblock is located at the upper left of the image C12, the range that the primary searcher 6 and the neighborhood searcher 8 are subject to motion search is centered on the encoding target macroblock. The search range clipper 2 cuts out an image in this range and outputs it to the image reducer 3 and the image cache memory 5.
The image reducer 3 reduces the image C21 output from the search range extractor 2 to half the number of pixels both vertically and horizontally, generates a reduced image C31, and outputs the reduced image C31 to the reduced image cache memory 4. The reduced image cache memory 4 stores a reduced image C31 output from the image reducer 3 as a reference image for primary search performed by the primary searcher 6. On the other hand, the image cache memory 5 stores an image C21 output from the search range extractor 2 as a reference image for secondary search performed by the neighborhood searcher 8.
The primary searcher 6 reads the reduced image C41 from the reduced image cache memory 4, and performs a primary search for searching a region similar to the reduced image C122 output from the image reducer 12 on the read reduced image C41 in units of macroblocks. To generate a motion vector C61 and output the generated motion vector C61 to the neighborhood searcher 8. Hereinafter, the motion vector generated by the primary searcher 6 is also referred to as a “primary motion vector”. The primary motion vector is a motion vector having an accuracy of 2 pixels indicating one of the center coordinates of the search range of the secondary search.

The search center indicator 7 outputs the center coordinates (search center coordinates) C71 of the search range of the secondary search performed by the neighborhood searcher 8. The specific contents of the search center coordinates output by the search center indicator 7 will be described later.
Based on the pixel value of the macroblock C121 to be encoded, the image feature quantity calculator 9 determines whether or not the image of the macroblock C121 is flat, that is, whether or not the variation in the pixel value of the macroblock C121 is small. The signal C91 indicating the determination result is output to the neighborhood searcher 8 and the motion vector selector 10.
The neighborhood searcher 8 reads the image C51 from the image cache memory 5, and performs a secondary search for searching for a region similar to the macroblock image C121 to be encoded on the read image C51. Specifically, the neighborhood searcher 8 uses, for example, a motion search with integer pixel accuracy centered on the primary motion vector C61 output from the primary searcher 6 and the search center coordinate C71 output from the search center indicator 7. The motion vector and the motion vector evaluation value C81 for each of the primary motion vector C61 and the search center coordinate C71 are sent to the motion vector selector 10 by performing a search within a predetermined range, for example, in a range of ± 1 pixel. Output.

Here, the “motion vector evaluation value” is used to describe the motion vector and the coding mode in the absolute value sum of the difference between each pixel in the encoding target block and the corresponding pixel on the reference image. This is a value obtained by adding a value corresponding to the required bit amount (encoding cost). The motion vector evaluation value is an index indicating a code amount or an image error when motion prediction is performed using the motion vector, and a motion vector with a smaller motion vector evaluation value moves with respect to a macroblock to be encoded. Suitable for use in making predictions.
As will be described later, the search center indicator 7 outputs a plurality of search center coordinates C71, and the neighborhood searcher 8 outputs a motion vector for each of the plurality of search center coordinates C71. When performing the secondary search, if the signal C91 indicating the determination result that the pixel value variation of the macroblock to be encoded is small is output from the image feature amount calculator 9, the neighborhood searcher 8 X Generate a motion vector for a vertical 16 pixel block (that is, a macroblock), and if a signal C91 indicating that the variation is not small is output, generate a motion vector for each horizontal 8 x vertical 8 pixel block To do.

The motion vector selector 10 selects a motion vector having the smallest motion vector evaluation value for each block from the plurality of motion vectors C81 output from the neighborhood searcher 8, and selects a motion vector C101 for each selected block. Output to the block integrator 11.
Based on the motion vector C101 for each block output from the motion vector selector 10, the block integrator 11 determines a block size to be applied to the macroblock for processing, and detects the motion vector C201 for each block as a motion vector detection. This is output to the outside of the apparatus 100 (for example, a predicted image generating unit of a moving image encoding apparatus including the motion vector detecting apparatus 100).

FIG. 2 is a diagram illustrating an example in which the primary searcher 6 performs a primary search. The primary searcher 6 performs a primary search using a direct search method.
A reduced image C122 in the figure is a reduced image of a macroblock to be encoded, which is output from the image reducer 12. The non-reference images P22 and P23 and the reference image P24 are reduced images C41 read from the reduced image cache memory 4. Further, the reduced image C122 (including the image to be encoded) and the reference image P24 are separated by three images in the time direction with the non-reference images P22 and P23 interposed therebetween.
As shown in the figure, in the direct search method, the primary searcher 6 sets a search range corresponding to the coordinates of the encoding target macroblock directly in the reference image P24, and the encoding target macro is within the set search range. The primary motion vector 51 is determined by matching the block with the reference image.
Specifically, the primary searcher 6 includes pixels of the reduced image C122 of the encoding target macroblock and the region of the same size on the reference image within the search range set on the reduced reference image C41. The motion vector evaluation value is calculated based on the sum of the absolute difference values of the pixels, and the motion vector indicating the region where the motion vector evaluation value is the smallest within the search range is calculated as the primary motion. Let it be a vector C61.

Note that the method for the primary search by the primary searcher 6 is not limited to the direct search method.
FIG. 3 is a diagram illustrating an example in which the primary searcher 6 performs a primary search using a telescopic search method. The reduced image C122, the non-reference images P22 and P23, and the reference image P24 in the figure are the same as those described in FIG.
First, the primary searcher 6 applies a search range (in the non-reference image P32 in the non-reference image P32) to the non-reference image P32 that is adjacent to the reduced image C122 (including the image to be encoded) in the time direction. A search range centered on the coordinates of the encoding target block C122 is set, and a motion search is performed. Next, the primary searcher 6 sets a search range corresponding to the coordinates indicated by the search result R32 in the non-reference image P32 in the non-reference image P33 adjacent to the non-reference image P32 in the time direction, and performs a motion search. . Furthermore, the primary searcher 6 sets a search range corresponding to the coordinates indicated by the search result R33 in the non-reference image P33 in the reference image P34 adjacent to the non-reference image P33 in the time direction, and performs a motion search. Then, the primary searcher 6 uses the search result R34 obtained by this motion search as the final motion search result (primary motion vector C61).
In the direct search method, in order to follow high-speed movement of an object in an image, the search range must be sufficiently wide, which increases the amount of calculation. On the other hand, in the telescopic search method, the search is sequentially performed from the temporally closest image to the reference image, and the motion vector of the search result is repeatedly set as the search center point of the next image. Even if a single search range is set to be relatively narrow, it is possible to follow the movement of a wide range of objects on the reference image.

FIG. 4 is a diagram illustrating an example of a motion vector obtained as a result of the primary search.
As a result of the primary search, as shown in FIG. 4, the primary searcher 6 detects and outputs a motion vector in a macroblock unit of 16 × 16 pixels as a primary search result C61.

FIG. 5 is a diagram illustrating an example of the search range of the secondary search performed by the neighborhood searcher 8. In the figure, the neighborhood searcher 8 performs an integer pixel precision motion search in the range of ± 1 pixel around the coordinates indicated by the primary motion vector C61 on the non-reduced image C51. In this case, the neighborhood searcher 8 performs a motion search for all nine motion vectors as shown in the figure, and selects a motion vector having the smallest motion vector evaluation value, thereby obtaining a motion vector with integer pixel accuracy. To detect.
In the neighborhood searcher 8, as in the case of the primary searcher 6, the pixel difference absolute value between the macroblock to be encoded and the reference image is calculated within the range of 16 × 16 macroblocks. Unlike the case of the primary searcher 6, the sum of absolute differences is calculated for each horizontal 8 × vertical 8 pixel block, and an optimal motion vector is detected for each horizontal 8 × vertical 8 pixel block. Alternatively, the neighborhood searcher 8 may perform a motion search independently for each horizontal 8 × vertical 8 pixel block and detect an optimal motion vector for each horizontal 8 × vertical 8 pixel block.

FIG. 6 is a diagram illustrating an example in which the neighborhood searcher 8 detects an optimal motion vector for each horizontal 8 × vertical 8 pixel block.
As described with reference to FIG. 5, the neighborhood searcher 8 calculates the absolute difference value of 16 pixels in the horizontal direction and 16 pixels in the vertical direction = 256 pixels for the nine points near the search center such as the coordinates indicated by the primary motion vector C61. Then, the neighborhood searcher 8 calculates the motion vector evaluation value by summing up the sum of absolute differences for each horizontal 8 × vertical 8 pixel block, that is, horizontal 8 × vertical 8 = 64 pixels. The neighborhood searcher 8 detects a motion vector that minimizes the calculated motion vector evaluation value for every 8 horizontal × 8 vertical blocks.
Here, as described with reference to FIG. 5, even if the neighborhood searcher 8 calculates a motion vector having the minimum motion vector evaluation value for each horizontal 8 × vertical 8 pixel block, the primary motion vector C 61 is at most. Only the motion vector C81 is detected from motion vectors separated by ± 1 pixel. By increasing the number of motion search targets, it can be expected that the coding efficiency can be improved by detecting a region whose image is more similar to the block to be encoded. Therefore, the neighborhood searcher 8 performs a motion search for the neighborhood centered on the search center coordinate C71 output from the search center indicator 7 in addition to the primary motion vector C61, and in units of horizontal 8 × vertical 8 pixel blocks, A motion vector having the smallest motion vector evaluation value is obtained.

The search center indicator 7 outputs a number of search center coordinates C71 corresponding to the neighborhood search calculation amount that the neighborhood searcher 8 can process. In particular,
(A) Motion vector prediction value and zero vector (b) Primary motion vector of left adjacent block or primary motion vector of left adjacent macroblock, and primary motion vector of upper adjacent block or upper adjacent macroblock (c ) For the primary motion vector of the lower adjacent macroblock or the primary motion vector of the right adjacent macroblock, (a) and (b) and (c) or (a ) And (b), only (a), or any other combination is set as the search center coordinate C71. Here, the “left adjacent block” is a block adjacent to the left of the block on which motion search is performed. Similarly, “left adjacent macroblock”, “upper adjacent block”, “upper adjacent macroblock”, and the like are blocks or macroblocks adjacent to each other in the direction of the motion search block. The search center coordinates C71 output by the search center indicator 7 are not limited to the above (a), (b), (c), and combinations thereof. For example, only the motion vector prediction value may be used as the search center coordinate C71. Hereinafter, in this embodiment, the case where the search center indicator 7 outputs (a), (b), and (c) as the search center coordinates C71 will be described.

FIG. 7 is a diagram showing a motion vector prediction value and a zero vector, which are the search center coordinates of (a).
As described above, the motion vector prediction value is a vector indicating an average value of the motion amount of the image of the adjacent block whose motion vector has already been detected. Therefore, when the image of the macroblock to be encoded is moving similarly to the adjacent block, a motion vector having a small motion vector evaluation value can be detected by performing a motion search in the vicinity of the motion vector prediction value. There is a high possibility that a region of an image having a high degree of similarity with the block to be encoded can be detected. When the image of the macroblock to be encoded moves in the same manner as the adjacent block, the image of the macroblock to be encoded and the image of the adjacent block are images of the same moving object, A case where the background image is almost stationary or moves (pans) uniformly can be considered.
The “zero vector” here is a vector indicating the origin. The zero vector indicates the position of the block to be encoded on the reference image in relative coordinates from the position of the block to be encoded. Therefore, when the macroblock image to be encoded is an image of a stationary object, background, or the like, a motion vector with a small motion vector evaluation value can be detected by performing a motion search in the vicinity of the motion vector prediction value. Is expensive.

FIG. 8 is a diagram showing the motion vector of the left adjacent block or the primary motion vector of the left adjacent macroblock, and the motion vector of the upper adjacent block or the primary motion vector of the upper adjacent macroblock, which are the search center coordinates in (b) above. It is.
When the image of the macro block B81 to be coded is similar to the image of the left adjacent block B82, such as when the image of the macroblock B81 to be encoded is continuous with the image of the left adjacent block B82, motion search is performed in the vicinity of the motion vector of the left adjacent block B82. Thus, there is a high possibility that a motion vector having a small motion vector evaluation value can be detected. The primary motion vector of the left adjacent macroblock is the motion of the left adjacent block B82 when the motion vector of the left adjacent block B82 cannot be obtained for some reason, for example, when motion search of a plurality of macroblocks is performed in parallel. It is conceivable to substitute for a vector.
Note that the motion search may be performed for all four horizontal 8 × vertical 8 pixel blocks B 811 to B 814 using the motion vector of the left adjacent block B 82 or the primary motion vector of the left adjacent macro block. In order to reduce the amount of calculation for motion search, motion search is performed only for the horizontal 8 × vertical 8 pixel blocks B811 and B813 adjacent to the left adjacent macroblock.

  Similarly, when the same motion as the image of the upper adjacent block B83, such as when the image of the macro block B81 to be encoded is continuous with the image of the upper adjacent block B83, the motion vector of the upper adjacent block B83 or the upper adjacent block It is highly possible that a motion vector having a small motion vector evaluation value can be detected by performing motion search in the vicinity of the primary motion vector of the macro block. Note that the motion search may be performed for all four horizontal 8 × vertical 8 pixel blocks B 811 to B 814 using the motion vector of the upper adjacent block B 83 or the primary motion vector of the upper adjacent macro block. In order to reduce the amount of calculation for motion search, motion search is performed only for the horizontal 8 × vertical 8 pixel blocks B811 and B812 adjacent to the upper adjacent macroblock.

FIG. 9 is a diagram showing the primary motion vector of the lower adjacent macroblock and the primary motion vector of the right adjacent macroblock, which are the search center coordinates of (c).
In order to use these primary motion vectors as search center coordinates, it is necessary that the primary motion vector of the lower adjacent macroblock and the primary motion vector of the right adjacent macroblock have already been generated when performing the secondary search. . Therefore, as shown in the lower part of FIG. 9, the primary search by the primary searcher 6 is a collection of macroblocks arranged in a row in the horizontal direction from the left end to the right end of one macroblock row (image (frame)) than the neighbor searcher 8. ) It must be done ahead of time.

  When the same motion as the image of the lower adjacent macroblock B93 is performed, such as when the image of the macroblock B91 to be encoded is continuous with the image of the lower adjacent macroblock B93, the vicinity of the primary motion vector of the lower adjacent macroblock B93 It is highly possible that a motion vector having a small motion vector evaluation value can be detected by performing a motion search. Note that the motion search may be performed on all four horizontal 8 × vertical 8 pixel blocks B911 to B914 using the primary motion vector of the lower adjacent macroblock B93. However, in this embodiment, the amount of motion search is reduced. Therefore, a motion search is performed only for the horizontal 8 × vertical 8 pixel blocks B913 and B914 adjacent to the lower adjacent macroblock B93.

  Similarly, when the same motion as the image of the right adjacent macroblock B92 is performed, such as when the image of the macroblock B91 to be encoded is continuous with the image of the right adjacent macroblock B92, the primary motion of the right adjacent macroblock B92 It is highly possible that a motion vector having a small motion vector evaluation value can be detected by performing motion search in the vicinity of the vector. Note that the motion search may be performed on all four horizontal 8 × vertical 8 pixel blocks B911 to B914 using the primary motion vector of the right adjacent macroblock B83. However, in this embodiment, the amount of motion search is reduced. Therefore, a motion search is performed only for the horizontal 8 × vertical 8 pixel blocks B911 and B912 adjacent to the right adjacent macroblock B92.

FIG. 10 shows an example of motion vectors and motion vector evaluation values output by the neighborhood searcher 8 to the motion vector selector 10, and an example in which the motion vector selector 10 selects motion vectors and motion vector evaluation values. FIG.
The neighborhood searcher 8 performs the motion search described with reference to FIG. 5 with respect to each of the four horizontal 8 × vertical 8 pixel blocks B101 to B104 as the secondary search target, centering on each of the primary motion vector C61 and the plurality of search center coordinates C71. A motion search is performed on nine target points, and the motion vector having the smallest motion vector evaluation value among the nine points and the motion vector evaluation value are output to the motion vector selector 10. The motion vector selector 10 selects a motion vector having the smallest motion vector evaluation value from among the motion vectors output from the neighborhood searcher 8 for each of the blocks B101 to B104. For example, in the example of FIG. 10, the motion vector selector 10 selects the motion vector MV31 having the minimum motion vector evaluation value of 48 for the block B101. Similarly, the motion vector MV12 is selected for the block B102, the motion vector MV13 is selected for the block B103, and the motion vector MV34 is selected for the block B104. The motion vector selector 10 outputs the motion vector C101 for each selected block to the block integrator 11.

When the motion vector values output from the motion vector selector 10 are the same for adjacent blocks, the block integrator 11 integrates both blocks into a larger size block.
For example, in FIG. 10, when the values of the motion vector MV31 of the block B101 and the motion vector MV12 of the block B102 are equal, and the values of the motion vector MV13 of the block B103 and the motion vector MV34 of the block B104 are equal. The block integrator 11 integrates the block B101 and the block B102, and also integrates the block B103 and the block B104 to generate two horizontal 16 × 8 vertical pixel blocks, and is a unit block for performing motion prediction The size is determined to be 16 × 8 pixels (blocks B1521 and B1522 in FIG. 15).

Alternatively, if the value of the motion vector MV31 of the block B101 is equal to the value of the motion vector MV13 of the block B103, and the value of the motion vector MV12 of the block B102 is equal to the value of the motion vector MV34 of the block B104, the block integrator 11 The block B101 and the block B103 are integrated, and the block B102 and the block B104 are integrated to generate two horizontal 8 × vertical 16 pixel blocks, and the block size of a unit for performing motion prediction is set to 8 × horizontal. The length is determined to be 16 pixels (blocks B1531 and B1532 in FIG. 15).
Alternatively, when the motion vector values of the blocks B101 to B104 are all equal, the block integrator 11 integrates the four blocks B101 to B104 to generate a horizontal 16 × vertical 16 pixel block, and performs motion prediction. Is determined to be 16 × 16 pixels (block B1511 in FIG. 15).
On the other hand, when the motion vector values of adjacent blocks are different from each other, the block integrator 11 does not perform block integration, and sets the block size of a unit for performing motion prediction to 8 × 8 pixels (block B1551 in FIG. 15). To B1554).
As described above, the block integrator 11 determines the block size (inter-coding mode) and the motion vector C201, and outputs them to the outside of the motion vector detection device 100.

  Here, if the macroblock image to be encoded is flat and the reference image has a similar flat image area, there is not much difference in the motion vector evaluation value in any part of the flat image area. It is thought that it does not occur. In this case, the motion vector evaluation value slightly changes due to the influence of disturbance such as noise, and the motion vector selected by the neighborhood searcher 8 or the motion vector selector 10 is different from the case where there is no influence of the disturbance. sell. Therefore, looking at each 8 × 8 pixel block of the macroblock, if the macroblock image is flat, motion vectors in different directions are easily selected for each block due to disturbance such as noise. In this case, the block integrator 11 does not perform block integration, and the encoding efficiency decreases.

  In addition, in a video in which the entire screen is moved (panned) in the same direction, it is desirable that a motion vector near the motion vector prediction value reflecting the movement of the entire screen is selected. On the other hand, when a motion vector near the motion vector prediction value is not selected in some horizontal 8 × vertical 8 pixel blocks due to the influence of noise or the like, for example, when a motion vector near the zero vector is selected, When processing adjacent macroblocks, the search center indicator 7 calculates a motion vector prediction value based on a motion vector near the zero vector. As a result, the motion vector prediction value calculated by the search center indicator 7 does not correctly reflect the movement of the entire screen. For this reason, there is a high possibility that a motion vector near the motion vector prediction value is not selected even in a macro block adjacent to the block. Similarly, there is a high possibility that a motion vector near the motion vector prediction value is not selected even in a macro block adjacent to this macro block. As described above, mixing of noise or the like into one block may affect a plurality of macroblocks, resulting in a decrease in encoding efficiency, and an image error is included in a decoded image at the time of encoding / decoding. There is a risk that subjective quality will be reduced.

Therefore, the image feature quantity calculator 9 calculates an index indicating pixel value variation such as dispersion of pixel values as the feature quantity of the encoding target macroblock image C121, and determines whether or not the macroblock image is flat. Then, it is determined whether or not the pixel value variation is small, and the determination result C91 is output to the neighborhood searcher 8 and the motion vector selector 10. When the determination result that the variation of the pixel value is small is output, the neighborhood searcher 8 does not calculate the absolute difference value and calculate the motion vector in the horizontal 8 × vertical 8 pixel block unit, and the horizontal 16 × vertical 16 A motion search is performed in units of macroblocks of pixels, and a motion vector of the macroblock and a motion vector evaluation value are output. Then, the motion vector selector 10 selects a motion vector having the smallest motion vector evaluation value from among the motion vectors of the macroblock output from the neighborhood searcher 8. Thereby, it is possible to prevent the blocks from being finely divided in a flat image.
For example, the image feature amount calculator 9 determines that the variation in the pixel value of the macroblock is small when Expression (1) is satisfied.

  Here, the “macroblock variance value” refers to the sum of absolute values of the difference between each pixel in the macroblock and the average value of all the pixels. Α is a threshold value, for example, a predetermined constant.

FIG. 11 is a flowchart showing a processing procedure in which the motion vector detection apparatus 100 detects a motion vector for the encoding target macroblock C121.
In step S11, the image reducer 12 generates an image C122 obtained by reducing the encoding target macroblock C121 to half the number of pixels both vertically and horizontally, and outputs the image C122 to the primary searcher 6. The primary searcher 6 Using the reduced image C122, a motion search on the reduced reference image C41 is performed in units of macroblocks, and a primary motion vector C61 is generated and output to the neighborhood searcher 8.
In step S <b> 12, the search center indicator 7 outputs search center coordinates C <b> 71 indicating the center coordinates of the region in which the neighborhood searcher 8 performs the secondary search.
In step S <b> 13, the image feature amount calculator 9 calculates an image feature amount indicating whether or not the image of the encoding target macroblock C <b> 121 is flat based on the pixel value of the encoding target macroblock C <b> 121.

In step S14, the image feature amount calculator 9 determines whether or not the variation in the pixel value of the macroblock C121 to be encoded is small based on the calculated image feature amount. If it is determined that the variation is small (step S14: YES), the process proceeds to step S21. If it is determined that the variation is not small (step S14: NO), the process proceeds to step S15.
In step S15, the neighborhood searcher 8 performs a motion search on the reference image C51 using the encoding target macroblock C121, and the primary motion vector C61 and the search center indicator 7 output by the primary searcher 6 For each neighborhood of search center coordinates C71 to be output, a motion vector having a minimum motion vector evaluation value is selected for each horizontal 8 × vertical 8 pixel block, and the selected motion vector and its motion vector evaluation value are Output to the motion vector selector 10.

In step S16, the motion vector selector 10 selects a motion vector having a minimum motion vector evaluation value for each horizontal 8 × vertical 8 pixel block from the motion vectors output from the neighborhood searcher 8, and selects the selected block. Each motion vector is output to the block integrator 11.
In step S17, the block integrator 11 integrates blocks that are adjacent to each other and have the same motion vector value, thereby determining the block size of the macroblock to be encoded and the motion vector of each block.
In step S <b> 18, the block integrator 11 outputs the determined block size and motion vector to the outside of the motion vector detection device 100. Thereafter, the process of detecting a motion vector for the encoding target macroblock C121 ends.

In step S21, the neighborhood searcher 8 performs a motion search on the reference image C51 using the encoding target macroblock C121, and the primary motion vector C61 and the search center indicator 7 output by the primary searcher 6 For each neighborhood of the search center coordinates C71 to be output, a motion vector having a minimum motion vector evaluation value is selected in units of horizontal 16 × vertical 16 pixel blocks, and the selected motion vector and its motion vector evaluation value are Output to the motion vector selector 10.
In step S22, the motion vector selector 10 selects a motion vector having a minimum motion vector evaluation value in units of 16 × 16 pixels from the motion vector output from the neighborhood searcher 8, and selects the selected motion. The vector is output to the block integrator 11. Thereafter, the process proceeds to step S18.

As described above, when the primary searcher 6 performs a wide range primary search using the reduced image, the search is performed in units of macroblocks of 16 × 16 pixels, not 8 × 8 pixels. There are a large number of pixels for which motion vector evaluation values are calculated. Thereby, it is hard to receive the influence of disturbances, such as noise, and can track a moving body image more correctly.
Further, the neighborhood searcher 8 performs motion search not only in the neighborhood of the primary motion vector C61 but also in the neighborhood of the zero vector, the neighborhood of the motion vector prediction value, or the neighborhood of the motion vector of the block or macroblock adjacent vertically and horizontally. Therefore, it is possible to select a more appropriate motion vector corresponding to a still image or an image that moves like an adjacent block.
In addition, the neighborhood searcher 8 generates a motion vector for a limited block size, for example, every horizontal 8 × vertical 8 pixels, so there is no need to perform a motion search in various block sizes, and the amount of motion search calculation is reduced. it can. In addition, since the block integrator 11 integrates blocks according to the value of the motion vector, it is possible to select an appropriate block size and increase the encoding efficiency.
Furthermore, since the image feature amount calculator 9 detects a macroblock with a flat image and suppresses division into fine blocks, the image feature amount calculator 9 is hardly affected by disturbances such as noise even when the image is flat.

Note that the reduction ratio of the reference image used when the primary searcher 6 performs the primary search is not limited to 1/2 of the above-described vertical and horizontal directions, and the reference image used when the neighborhood searcher 8 performs the secondary search is as follows: It is not limited to non-reduced ones. For example, in order to expand the search range while suppressing the amount of computation, the primary searcher 6 performs a primary search using a reference image with a reduction ratio of 1/4 in both the vertical and horizontal directions, and the neighborhood searcher 8 selects an unreduced reference image. A secondary search may be performed by using this. Alternatively, the primary searcher 6 performs a primary search using a reference image with a reduction ratio of 1/4 both vertically and horizontally, and the neighborhood searcher 8 performs a secondary search using a reference image with a reduction ratio of 1/2 both vertically and horizontally. May be performed. In this case, the primary searcher 6 performs a search using the reduced image C122 of the encoding target macroblock reduced to 4 × 4 pixels.
Further, the reference image used when the primary searcher 6 performs the primary search may not have the same vertical / horizontal reduction ratio. For example, the primary searcher 6 may perform a primary search using a reduced image C122 of horizontal 4 × vertical 8 pixels that has been reduced to 1/2 vertical and 1/4 horizontal.

Further, the target of the secondary search is not limited to the nine nearby points. For example, the neighborhood searcher 8 performs a secondary search for 25 points within a range of ± 2 pixels around a primary vector or the like, or performs a secondary search with a decimal accuracy such as 0.25 pixel or 0.5 pixel. For example, a wider range may be searched. Thereby, a search for a more appropriate motion vector can be expected.
Further, the primary searcher 6 and the neighborhood searcher 8 may use a plurality of reference image candidates. For example, the primary searcher 6 performs a primary search with respect to a plurality of reference image candidates, selects a reference image having the smallest motion vector evaluation value and a primary motion vector, and outputs them to the neighborhood searcher 8. Also good. The neighborhood searcher 8 performs a neighborhood search centered on the position indicated by the primary motion vector on the reference image (non-reduced reference image) corresponding to the reference image (reduced reference image) selected by the primary searcher 6. The information indicating the reference image and the motion vector resulting from the secondary search are output for each block.
Furthermore, the condition that the block integrator 11 integrates blocks into a larger block is not limited to the case where the motion vector values of adjacent blocks are the same. For example, the blocks may be integrated when Expression (2) is established.

  Here, β is a threshold value, for example, a predetermined constant. For example, when β = 0.25, the block integrator 11 integrates blocks when the absolute value of the motion vector difference between adjacent blocks is 0.25 pixels or less. In this way, by making it easier to integrate blocks, the number of motion vectors to be encoded is reduced, and an improvement in encoding efficiency can be expected. When integrating blocks, the block integrator 11 determines, for example, the motion vector of any block before integration as the motion vector of the block after integration.

Note that a program for realizing all or part of the functions of the motion vector detection apparatus 100 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into a computer system and executed. You may process each part by. Here, the “computer system” includes an OS and hardware such as peripheral devices.
Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design changes and the like within a scope not departing from the gist of the present invention.

  The present invention provides motion prediction such as a video encoder that encodes a moving picture digital signal in accordance with an MPEG (Moving Picture Experts Group) system or an ITU-T (The International Telecommunication Union Telecommunication Standardization Sector) H series system. In the video encoder that performs the above, it is suitable for use in a motion vector detection device, a motion vector detection method, and a program that output a motion vector.

DESCRIPTION OF SYMBOLS 1 Image memory 2 Search range cutout device 3, 12 Image reducer 4 Reduced image cache memory 5 Image cache memory 6 Primary searcher 7 Search center indicator 8 Neighbor searcher 9 Image feature-value calculator 10 Motion vector selector 11 Block integration 100 Motion vector detection device

Claims (7)

A motion vector detection device that detects a motion vector indicating a position of a region corresponding to the first block on a reference image for each first block obtained by dividing a macroblock of an input moving image,
A primary searcher that performs a primary motion search on the macroblock to detect a primary motion vector;
A search center indicator that outputs a search center coordinate indicating a center coordinate of a range in which a secondary motion search is performed;
A neighborhood searcher that detects a motion vector by performing a secondary motion search for each coordinate between the coordinates indicated by the primary motion vector and the search center coordinates and for each second block obtained by dividing the macroblock;
A motion vector selector for selecting a motion vector for each of the second blocks from among the motion vectors detected by the neighborhood searcher;
Based on the motion vector selected by the motion vector selector, the second block or a block obtained by combining the second blocks adjacent to each other is defined as the first block, and the motion of the second block A block integrator for determining a motion vector of the first block based on a vector;
A motion vector detection apparatus comprising:   The search center coordinates include a zero vector indicating a position of the macroblock or a motion vector prediction value calculated from a motion vector of a block adjacent to the second block. Motion vector detection device.   The search center coordinates are the motion vector of the left adjacent block adjacent to the left of the macroblock, the primary motion vector of the left adjacent macroblock adjacent to the left of the macroblock, or adjacent to the macroblock. The motion vector detection apparatus according to claim 1, further comprising: a motion vector of an upper adjacent block that performs an upper adjacent block, or the primary motion vector of an upper adjacent macroblock that is adjacent to the macro block.   The search center coordinates include the primary motion vector of a right adjacent macroblock adjacent to the right of the macroblock or the primary motion vector of a lower adjacent macroblock adjacent below the macroblock. The motion vector detection device according to any one of claims 1 to 3. Based on the pixel value of the macroblock, an index indicating pixel value variation is calculated, and further includes an image feature quantity calculator that determines whether or not the macroblock pixel value variation is small,
2. The secondary searcher, wherein the image feature quantity calculator determines that the macroblock is the second block when it is determined that variations in pixel values of the macroblock are small. 5. The motion vector detection device according to any one of items 1 to 4. A motion vector detection method for a motion vector detection apparatus that detects a motion vector indicating a position of an area corresponding to a first block on a reference image for each first block obtained by dividing a macroblock of an input moving image. ,
Performing a primary motion search on the macroblock to detect a primary motion vector; and
A search center instruction step for outputting a search center coordinate indicating a center coordinate of a range in which a secondary motion search is performed;
A neighborhood search step for detecting a motion vector by performing a secondary motion search for each coordinate between the coordinate indicated by the primary motion vector and the search center coordinate and for each second block obtained by dividing the macroblock;
A motion vector selection step of selecting a motion vector for each of the second blocks from the motion vectors detected in the neighborhood search step;
Based on the motion vector selected in the motion vector selection step, the second block or a block obtained by combining the second blocks adjacent to each other is defined as the first block, and the second block A block integration step for determining a motion vector of the first block based on a motion vector;
A motion vector detection method comprising: A program for causing a computer to detect a motion vector indicating a position of an area corresponding to a first block on a reference image for each first block obtained by dividing a macro block of an input moving image,
Performing a primary motion search on the macroblock to detect a primary motion vector; and
A search center instruction step for outputting a search center coordinate indicating a center coordinate of a range in which a secondary motion search is performed;
A neighborhood search step for detecting a motion vector by performing a secondary motion search for each coordinate between the coordinate indicated by the primary motion vector and the search center coordinate and for each second block obtained by dividing the macroblock;
A motion vector selection step of selecting a motion vector for each of the second blocks from the motion vectors detected in the neighborhood search step;
Based on the motion vector selected in the motion vector selection step, the second block or a block obtained by combining the second blocks adjacent to each other is defined as the first block, and the second block A block integration step for determining a motion vector of the first block based on a motion vector;
For causing the computer to execute.

Images