JP2006254347A - Image encoding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image coding device that suppresses the calculation processing amount of motion vector search and hardware resources used for the calculation, and realizes highly accurate motion vector search.
An image encoding apparatus includes a first motion vector search unit 15 that performs a motion vector search for a large region, a second motion vector search unit 16 that performs a motion vector search for a small region, and a second motion vector search unit 16. A motion vector interpolation unit 19 that obtains the center point of the search range in the first delay circuit 17, the second delay circuit 18, and the second motion vector search unit that holds the search result in the first motion vector search unit until the end of the motion vector search. , A selector 20 for selecting a search result in the first motion vector search unit 15 or a search result in the second motion vector search unit 16, and holds a motion vector of the previous image necessary for performing motion vector interpolation. A motion vector memory 21 is provided.
[Selection] Figure 1

Description

  The present invention relates to an image coding apparatus that divides an input image into blocks each having a predetermined number of horizontal pixels and a predetermined number of vertical lines, and performs motion compensation prediction by detecting a motion vector of the image for each block. .

  In MPEG2, which is a moving picture coding system using motion compensation prediction, a typical conventional system for performing motion vector search is basically a method called a block matching method, and is ISO / IEC JTC1 / SC29 / WG11. / N0400 “TEST MODEL5”. This method divides the current image into blocks (for example, 16 pixels × 16 pixel lines, 16 pixels × 8 lines, etc.), and uses an error evaluation standard such as a sum of absolute differences to estimate the prediction error for the block. In this method, a block having the smallest evaluation criterion is obtained from a reference image and used as a motion vector of the block. At that time, there is a method called “full search” in which the entire search range is exhaustively searched, and the minimum point is selected from the search range.

  In addition, there are several methods for reducing the processing amount of the full search method. For example, there are a sub-sampling method that thins out search points, a multi-stage matching method that searches again finely around only candidate points obtained by the sub-sampling method (see, for example, Patent Document 1).

JP 2000-134628 A

  Since the conventional motion vector search method has been performed as described above, the total search method has a problem that the amount of hardware required for the same operation as the amount of calculation processing becomes very large. Further, if the search points are simply thinned out in order to suppress them, there is a problem that the accuracy of the motion vector is lowered and the image quality is deteriorated.

  The present invention has been made to solve the above-described problems, and is an image coding that suppresses the calculation processing amount of motion vector search and hardware resources used for the calculation, and realizes high-precision motion vector search. An object is to provide an apparatus.

  An image encoding apparatus according to the present invention is an image encoding apparatus that encodes an image by dividing the image into blocks each having a predetermined number of horizontal pixels and a predetermined number of vertical lines, and is set to the first block. A first motion vector search unit that performs a full search on the search range of the first block and searches for a motion vector of the first block; a plurality of motion vectors of the first block; A motion vector interpolation unit that performs interpolation using a motion vector of a block of an image and obtains a center point of a second search range set in the second block, and performs a full search on the second search range And a second motion vector search unit for searching for a motion vector of the second block.

  According to this invention, the first block that performs the motion vector search in the first search range (large region) is reduced, and the motion vector search is performed in the second search range (small region). The amount and the hardware scale can be reduced. Also, when setting the second search range, a more accurate motion vector search is possible by interpolating using the motion vector of the first block and the motion vector of the past image in time series. There is an effect.

Embodiment 1 FIG.
Embodiment 1 of the present invention will be described below. FIG. 1 is a configuration diagram of an image coding apparatus according to Embodiment 1. In FIG. 1, an image encoding apparatus includes an input image signal 1, a reordering unit 2 that reorders input images, a subtractor 3 that obtains a difference between input image data and predicted image data, and image data as frequency components. An orthogonal transform unit 4 for decomposing, a quantization unit 5 for performing quantization, a variable length encoding unit 6 for performing variable length encoding, a buffer 7 for storing encoded data 8, and inverse quantization for obtaining predicted image data. Inverse quantization unit 9 for performing, inverse orthogonal transform unit 10 for performing inverse orthogonal transform to obtain predicted image data, sum of differences between data obtained by inverse orthogonal transform to generate predicted image data, input image data, and predicted image data Are provided, a first image memory 12 and a second image memory 13 for realizing bidirectional prediction, and a motion compensation unit 14 for performing motion compensation. The above configuration is the same as that of a conventional image encoding device.

  In addition to the above configuration, the image coding apparatus according to Embodiment 1 includes a first motion vector search unit 15 that performs a motion vector search for a large region, a second motion vector search unit 16 that performs a motion vector search for a small region, The first delay circuit 17 and the second delay circuit 18 that hold the search result in the first motion vector search unit until the end of the motion vector search in the second motion vector search unit 16, and the center of the search range in the second motion vector search unit A selector 20 for selecting a search result in the motion vector interpolation unit 19 and the first motion vector search unit 15 for obtaining a point or a search result in the second motion vector search unit 16, necessary for performing motion vector interpolation A motion vector memory 21 that holds motion vectors of past images is provided.

  Next, the operation will be described. FIG. 2 is a diagram for explaining the operation of the image coding apparatus in FIG. FIG. 3 is a diagram for explaining a motion vector search range by the first motion vector search unit 15 in FIG. In FIG. 2, a macroblock a is a macroblock for performing a large area motion vector search by the first motion vector search unit 15. The macro block b is a macro block for performing a motion vector search for a small area by the second motion vector search unit 16. In the macroblock a, the first motion vector search unit 15 performs a full search with a search range of 150 pixels × 100 lines centering on the target macroblock position (macroblock a) as shown in FIG. Do.

  FIG. 4 is a diagram for explaining a motion vector search range by the second motion vector search unit 16 in FIG. The second motion vector search unit 16 sets the motion vector search target as the macro block b. First, as shown in FIG. 2, the left and right macroblocks a closest to the macroblock b are set to V [m−2] [n] and V [m + 2] [n], respectively. Note that the motion vector of the macroblock a is detected by a full search by the first motion vector search unit 15. In addition, in the image (past image) t−1 immediately before the image t to which the macroblock b belongs, the motion vector of the macroblock c at the same position as the macroblock b is V [m ′] [n ′] and its vicinity. Let V [m′−1] [n ′] and V [m ′ + 1] [n ′] be the motion vectors of the macroblock d located on the left and right, respectively. In the first embodiment, the motion vector interpolation unit 19 interpolates these five motion vectors to obtain an interpolation vector C [m] [n] indicating the center point of the motion vector search range in the macroblock b. The second motion vector search unit 16 performs a full search using a small area centered on the interpolation vector C [m] [n], for example, 50 pixels × 50 lines as shown in FIG.

As an interpolation method by the motion vector interpolation unit 19, for example, the above five motion vectors are expressed as V [m−2] [n] = (− 20, 8).
V [m + 2] [n] = (− 5, 16)
V [m′−1] [n ′] = (− 25, −1)
V [m ′] [n ′] = (7, −3)
V [m ′ + 1] [n ′] = (8, 5)
If the interpolation vector C [m] [n] is
C [m] [n]
= {V [m−2] [n] + V [m + 2] [n] + V [m′−1] [n ′]
+ V [m ′] [n ′] + V [m ′ + 1] [n ′]} / 5
= {(− 20,8) + (− 5,16) + (− 25, −1) + (7, −3) + (8,5)}
/ 5
= (-7, 5)
As described above, the motion vector is averaged.

  As described above, according to the first embodiment, by reducing the macro block a that performs the motion vector search using the large region as the search range and performing the motion vector search of the macro block b using the small region as the search range, It is possible to reduce the calculation processing amount and the hardware resources required for the processing. In the small area search, the motion vector of the left and right macroblock a closest to the motion vector search target macroblock b, the motion vector of the macroblock c at the same position as the macroblock b in the image t-1, and the left and right sides of the motion vector search target macroblock b. Interpolation is performed using the motion vector of the macroblock d located at, and the search is performed by obtaining the center point of the search range of the macroblock b from the obtained interpolation vector, thereby enabling highly accurate motion vector search.

Embodiment 2. FIG.
The second embodiment of the present invention will be described below. FIG. 5 is a configuration diagram of an image encoding device according to the second embodiment. In addition to the configuration of the image coding apparatus according to the first embodiment (FIG. 1), the image coding apparatus according to the second embodiment searches a large area for any macroblock included in the coded data. A motion vector search control unit 22 is provided for controlling whether to perform a search for a small area.

  Next, the operation will be described. FIG. 6 is a diagram for explaining the operation of the image coding apparatus in FIG. In FIG. 6, each of the encoded image data arranged in the direction of the time axis t (in order of the past image t-3 to the current image t in time series) uses a macroblock 6a for performing a large area search and an interpolation vector. A macro block 6b for searching for a small area is provided. The motion vector search control unit 22 changes the position of each macroblock 6a in the image t-3 to the image t instead of the same position. Here, assuming that motion vectors have already been detected from image t-3 to image t-1, motion vector search for image t will be described. The motion vector search method for the macroblocks 6a and 6b is the same as in the first embodiment.

  For the macroblock 6a, the first motion vector search unit 15 performs a full search with a large area as a search range. For the macro block 6b, the motion vector interpolation unit 19 moves the motion vector of the left and right macro blocks 6a closest to the macro block 6b, that is, the motion of the macro block at the same position as the macro block 6b of the image t in the image t-1. An interpolation vector C [m] [n] is obtained by interpolating the vector and the motion vector of the macro block located in the vicinity of the vector. The second motion vector search unit 16 performs a full search in a small area centered on the interpolation vector C [m] [n]. The motion vector search is similarly performed for the images t-3 to t-1.

  As described above, according to the second embodiment, the position of the macro block 6a having the large area as the search range is changed by the encoded image data arranged in the time axis t direction. Therefore, it is possible to suppress the bias of the interpolation vector caused by searching the large area for the macroblock at the same position. As a result, the accuracy of the interpolation vector C [m] [n] for obtaining the center point of the small region search range can be increased.

Embodiment 3 FIG.
The third embodiment of the present invention will be described below. FIG. 7 is a configuration diagram of an image encoding device according to Embodiment 3. In addition to the configuration of the image coding apparatus according to the first embodiment (FIG. 1), the image coding apparatus according to the third embodiment includes a motion vector search target macroblock and a macroblock selected to obtain an interpolation vector. A block correlation evaluation unit 23 for evaluating the correlation is provided.

  Next, the operation will be described. FIG. 8 is a diagram for explaining the operation of the image coding apparatus in FIG. The motion vector search method of the third embodiment is different from that of the first embodiment in how to obtain an interpolation vector. Specifically, in FIG. 8, in order to obtain an interpolation vector indicating the center point of the search range in the motion vector search target macroblock 8 b, as in the first embodiment, positions near the left and right of the macroblock 8 b of the image t. Motion vectors are selected for the macro block 8a to be executed, the macro block 8c at the same position as the macro block 8b of the image t-1, and the macro block 8d located on the left and right in the vicinity of the macro block 8c. In the block correlation evaluation unit 23, error evaluation such as comparison of the sum of absolute differences is performed on each of the macroblocks 8a, 8c, and 8d. As a result, the motion vector of the macro block having a high correlation with the macro block 8b is weighted more than the motion vector of the other macro block to obtain the interpolation vector C [m] [n] of the macro block 8b.

For example, in FIG. 8, when the macroblock 8d having the motion vector V [m ′ + 1] [n ′] has a high correlation with the macroblock 8b, the interpolation vector C [m] [n] Asking.
C [m] [n]
= {2 * V [m-2] [n] + 2 * V [m + 2] [n] + 2 * V [m'-1] [n ']
+ 2 × V [m ′] [n ′] + 5 × V [m ′ + 1] [n ′]} / 13

  As described above, according to the third embodiment, the accuracy of the interpolation vector C [m] [n] is weighted by weighting the motion vector of the macroblock 8d having high correlation with the motion vector search target macroblock 8b. It is possible to increase. Alternatively, the motion vector of the highly correlated macroblock 8d may be used as the center of the search range of the macroblock 8b as it is.

Embodiment 4 FIG.
The fourth embodiment of the present invention will be described below. FIG. 9 is a configuration diagram of an image coding apparatus according to Embodiment 4. In addition to the configuration of the image coding apparatus according to the first embodiment (FIG. 1), the image coding apparatus according to the fourth embodiment includes a subsampling unit 24 that performs subsampling on an image for which motion vector search is performed. I have.

  Next, the operation will be described. FIG. 10 is a diagram for explaining the operation of the image coding apparatus in FIG. In FIG. 10, the sub-sampling unit 24 obtains average pixel data 10b from four pixel data 10a in the image data 10c, and performs 1/4 sub-sampling on the image data 10c using the average pixel data 10b. The image data 10d is output. By subsampling the original image data 10c to ¼ by the subsampling unit 24, the search range (large region) shown in FIG. 3 and the search range (small region) shown in FIG. The amount of data required is reduced.

  As described above, according to the fourth embodiment, the motion vector search range is reduced by performing sub-sampling on the image on which the motion vector search is performed. Hardware resources used for the processing can be suppressed.

Embodiment 5. FIG.
The fifth embodiment of the present invention will be described below. FIG. 11 is a diagram for explaining the fifth embodiment. In Embodiment 1, the motion vector of the macroblock a in the horizontal direction of the macroblock b is used when obtaining the interpolation vector for the macroblock b subject to motion vector search (see FIG. 2). In the fifth embodiment, as shown in FIG. 11, the interpolation vector is obtained using the macroblock a in the vertical direction of the macroblock b and the macroblocks c and d in the vertical direction for the image t-1. Note that, for the horizontal macroblock (Embodiment 1) and the vertical macroblock (Embodiment 5), the respective motion vectors may be interpolated together, or only the vertical macroblock can be interpolated. You may obtain | require the interpolation vector of b.

  Next, the operation will be described. In FIG. 11, let V [m] [n−2] and V [m] [n + 2] be the motion vectors of the macroblock a that has been subjected to full search up and down closest to the motion vector search target macroblock b. In addition, in the image t−1 input immediately before the image t to which the macroblock b belongs, the motion vector of the macroblock c at the same position as the macroblock b is V [m ′] [n ′], and the vicinity thereof is up and down. Let V [m ′] [n′−1] and V [m ′] [n ′ + 1] be the motion vectors of the macroblock d located. The motion vector interpolation unit 16 in FIG. 1 obtains an interpolation vector C [m] [n] by interpolating these five motion vectors. Since the subsequent processing is the same as that of the first embodiment, the description thereof is omitted.

  As described above, according to the fifth embodiment, the interpolation vector C [m] is obtained by using the motion vector of the vertical macroblock in addition to the motion vector of the horizontal macroblock of the first embodiment. The accuracy of [n] can be increased.

Embodiment 6 FIG.
The sixth embodiment of the present invention will be described below. FIG. 12 is a diagram for explaining the sixth embodiment. In the sixth embodiment, when the interpolation vector of the macroblock b of the image t in the first embodiment is obtained, a part of the motion vector of the immediately preceding image t-1 is saved, thereby storing the amount of memory used for the arithmetic processing. Reduce. For example, as shown in FIG. 12, a motion vector at the same position as the macroblock b subject to motion vector search is stored among the motion vectors of the immediately preceding image t-1 stored in the motion vector memory 21 (see FIG. 1). If not, the interpolation vector of the macroblock b is obtained using the motion vector of the macroblock d located near the same position (upper right, lower right, upper left, lower left).

  As described above, according to the sixth embodiment, by storing a part of the motion vector of the immediately preceding image t-1 in the motion vector memory 21, the amount of memory used for hardware arithmetic processing is reduced. be able to. In addition to Embodiment 1 or Embodiment 5, by selecting a macroblock in an oblique direction from the same position as the macroblock b in the image t-1, the accuracy of the interpolation vector C [m] [n] is further improved. Can be increased.

It is a block diagram of the image coding apparatus which concerns on Embodiment 1 of this invention. It is a figure explaining operation | movement of the image coding apparatus of FIG. It is a figure explaining the motion vector search range by the 1st motion vector search part in FIG. It is a figure explaining the motion vector search range by the 2nd motion vector search part in FIG. It is a block diagram of the image coding apparatus which concerns on Embodiment 2 of this invention. It is a figure explaining operation | movement of the image coding apparatus of FIG. It is a block diagram of the image coding apparatus which concerns on Embodiment 3 of this invention. It is a figure explaining operation | movement of the image coding apparatus of FIG. It is a block diagram of the image coding apparatus which concerns on Embodiment 4 of this invention. It is a figure explaining operation | movement of the image coding apparatus of FIG. It is a figure explaining Embodiment 5 of this invention. It is a figure explaining Embodiment 6 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Input image signal, 2 Rearrangement part, 3 Subtractor, 4 Orthogonal transformation part, 5 Quantization part, 6 Variable length coding part, 7 Buffer, 8 Encoded data, 9 Inverse quantization part, 10 Inverse orthogonal transformation part 11 adder, 12 first image memory, 13 second image memory, 14 motion compensation unit, 15 first motion vector search unit, 16 second motion vector search unit, 17 first delay circuit, 18 second delay circuit, 19 motion vector interpolation unit, 20 selector, 21 motion vector memory, 22 motion vector search control unit, 23 block correlation evaluation unit, 24 subsampling unit.

Claims (6)

In an image encoding device for encoding an image by dividing the image into blocks each having a predetermined number of horizontal pixels and a predetermined number of vertical lines,
A first motion vector search unit that performs a full search for the first search range set in the first block and searches for a motion vector of the first block;
Interpolation is performed using a plurality of motion vectors of the first block and a motion vector of a block of an image past the image in time series, and a center point of a second search range set in the second block is determined. A desired motion vector interpolation unit;
An image coding apparatus comprising: a second motion vector search unit that performs a full search on the second search range and searches for a motion vector of the second block. The motion vector interpolation unit
A motion vector of the first block located in the vicinity of the second block in the image;
A motion vector of a block at the same position as the second block in the image in the past image;
The interpolation vector indicating the center point of the second search range is obtained by interpolating a motion vector of a block located in the vicinity of the block at the same position in the past image. Image coding apparatus.   The image code according to claim 2, further comprising: a motion vector search control unit configured to set the position of the first block in the image to be different from the position of the first block in the past image. Device. The first block located near the second block in the image, the block located at the same position as the second block in the past image, and the block located near the block located at the same position A block correlation evaluation unit for evaluating the correlation between each block and the second block;
The image coding apparatus according to claim 2, wherein the motion vector interpolation unit weights a motion vector of the highly correlated block.   The image coding apparatus according to claim 2, further comprising a subsampling unit that generates a subsampled image in which the number of pixels is reduced by subsampling the image data. 6. The image encoding apparatus according to claim 2, further comprising a motion vector memory that stores a motion vector of the past image.

Images