JP4895853B2 - Motion vector detection apparatus and motion vector detection system - Google Patents

Description

  The present invention relates to a technology of an apparatus for detecting a motion vector from an image signal, and more particularly to a motion vector detection apparatus that improves the detection accuracy of a motion vector.

As a conventional motion vector detection device for an image signal, there is a configuration based on a block matching method. That is, a target block in which the target image is divided into a plurality of blocks and motion is detected, and a plurality of candidate regions (hereinafter referred to as candidate blocks) within a predetermined search range in a frame before or after the target image. The degree of correlation is evaluated. A candidate block having the highest correlation in the search range is determined, and a displacement between the candidate block and the target block is set as a motion vector. Further, as one of the plurality of candidate blocks described above, there is a configuration in which detection performance is improved by using an average result of detected motion vectors in a block in a line direction including the target block.
Japanese Patent Laying-Open No. 2005-287047

  In a motion vector detection device for image signals, when there are a plurality of different motions such as an object and background moving in different directions in the block of interest, the motion vector detection device is originally detected by the motions that exist. There was a problem that a correct motion vector should not be detected effectively.

  The present invention solves the above-described conventional problems, and an object thereof is to effectively detect a correct motion vector even when a plurality of different motions are mixed.

  In order to achieve the above-described object, the motion vector detection device of the present invention is a motion vector detection apparatus including N reference blocks in a predetermined range (N is an integer of 2 or more) in an image signal divided into a plurality of blocks. A motion vector detection device that generates a motion vector candidate signal of a reference block based on a vector signal, and uses the motion vector candidate signal for detection of a motion vector signal of a reference block, and quantizes N motion vector signals, Quantization means that generates N quantized signals, and a quantized signal counter that counts the number of quantized signals for each quantization level based on the N quantized signals and generates a quantized signal Vector signal integration for generating a vector signal by vector-integrating motion vector signals for each quantization level based on means and N quantized signals And a representative quantization level generating means for generating a representative quantization level representing the quantization level having the maximum number and a representative count quantized signal representing the number at the representative quantization level based on the stage and the count quantization signal A representative integrated vector signal generating means for generating a representative integrated vector signal representing the integrated vector signal at the representative quantization level, and a motion vector for generating a motion vector candidate signal by dividing the representative integrated vector signal by the representative count quantized signal Candidate signal generating means.

  In the motion vector detection system of the present invention, in the image signal divided into a plurality of blocks, a predetermined range is set to include N blocks (N is an integer of 2 or more) including a reference block. It has a range setting means, a predetermined range motion vector detection means for generating motion vector signals of N blocks in the predetermined range, and the motion vector detection device described above.

  According to the present invention, the most representative motion vector can be extracted as a candidate vector in a plurality of blocks in the vertical direction, the horizontal direction, or the adjacent block including the target block (reference block). The correlation degree can be evaluated and the motion vector in the target block can be detected more effectively and correctly.

  Hereinafter, examples relating to embodiments of the present invention will be described with reference to the drawings. In addition, all the numbers described in the following are illustrated in order to specifically explain the present invention, and the present invention is not limited to the illustrated numbers. The following embodiments are configured using hardware and / or software, but the configuration using hardware can be configured using software, and the configuration using software uses hardware. Can be configured.

  The motion vector detection apparatus according to the present invention is based on motion vector signals of N blocks (N is an integer of 2 or more) including reference blocks in a predetermined range in an image signal divided into a plurality of blocks. A motion vector detection device that generates a motion vector candidate signal and uses the motion vector candidate signal for detection of a motion vector signal of a reference block, and quantizes N motion vector signals and generates N quantized signals respectively. A quantizing means for counting, the number of quantized signals for each quantization level based on N quantized signals, and a quantized signal counting means for generating a quantized quantized signal; and N quantized signals Based on the above, vector signal integration means for vectorizing motion vector signals for each quantization level and generating an integrated vector signal, and based on the count quantized signal Representative quantization level generating means for generating a representative quantization level representing the maximum quantization level and a representative count quantized signal representing the number at the representative quantization level, and an integrated vector signal at the representative quantization level. Representative integrated vector signal generating means for generating a representative integrated vector signal, and a motion vector candidate signal generating means for generating a motion vector candidate signal by dividing the representative integrated vector signal by the representative count quantized signal. It is a feature.

  In the motion vector detection system of the present invention, in the image signal divided into a plurality of blocks, a predetermined range is set to include N blocks (N is an integer of 2 or more) including a reference block. It has a range setting means, a predetermined range motion vector detection means for generating motion vector signals of N blocks in the predetermined range, and the motion vector detection device described above.

(Embodiment 1)
FIG. 1 is a block diagram illustrating a configuration of the motion vector detection device according to the first embodiment. The motion vector detection apparatus according to the first embodiment includes a predetermined range setting unit 9, a predetermined range motion vector detection unit 10, a region determination unit 11, a region distinct number counter unit 12, a region-specific vector value integration unit 13, and a maximum number region determination unit. 14, an integrated value selection unit 15, a division unit 16, and a target block motion vector detection unit 17. A device including the region determination unit 11, the region distinct number counter unit 12, the region-specific vector value integration unit 13, the maximum number region determination unit 14, the integration value selection unit 15, and the division unit 16 is also referred to as a motion vector detection device. The range setting unit 9, the predetermined range motion vector detection unit 10, and a device including the motion vector detection device are also referred to as a motion vector detection system.

  First, a motion vector detection configuration for an image signal included in the predetermined range motion vector detection unit 10 and the target block motion vector detection unit 17 will be described. In the motion vector detection configuration, an image signal is divided into a plurality of motion vector detection blocks (hereinafter simply referred to as blocks), and an arbitrary one block of interest for detecting motion and a plurality of blocks within a predetermined search range. The degree of correlation is obtained between candidate areas (hereinafter referred to as candidate blocks). The block of interest is also called a reference block. The search range is set to a desired range over one or more frames of the previous frame and / or the subsequent frame with respect to the frame including the block of interest.

  The motion vector detection configuration generates a displacement vector of the target block for the candidate block having the highest degree of correlation. The motion vector of the target block is detected as a projection vector of the displacement vector onto the frame including the target block in the motion vector detection configuration.

  In FIG. 1, a predetermined range setting unit 9 sets a predetermined range S9 indicating a range of a plurality of blocks including a target block in a frame including the target block. The predetermined range motion vector detection unit 10 generates a motion vector signal for each block in a plurality of blocks including the target block in the predetermined range S9 by the above-described motion vector detection configuration, and outputs it as a motion vector signal S10.

  The region determination unit 11, the region distinct number counter unit 12, the region-specific vector value integration unit 13, the maximum number region determination unit 14, the integration value selection unit 15, and the division unit 16 are obtained in a plurality of blocks within the predetermined range S9. The accuracy of each motion vector signal S10 is increased by the configuration described below, and one motion vector candidate signal S16 is generated. The target block motion vector detection unit 17 detects again the motion vector of the target block based on the motion vector candidate signal S16. In some cases, the detection result in the target block motion vector detection unit 17 is the motion vector signal S10 in the predetermined range motion vector detection unit 10 together with the detection result of the similar configuration in the blocks other than the target block in the predetermined range S9. The accuracy of the motion vector is increased.

  The motion vector signal S10 has horizontal and vertical components of the image signal. For each direction, the motion vector value is a finite value and there is an effective range. Therefore, the effective range can be divided into a plurality of motion vector regions (hereinafter simply referred to as regions). In the following description, the motion vector is limited to one of the horizontal and vertical components, but the other component can be similarly described.

For example, when the effective range of the motion vector value is from −128 to +127, as an example of dividing into four regions of region 1, region 2, region 3, and region 4, vector values corresponding to each region The range of
Region 1 is from -128 to -65,
Region 2 is from -64 to -1,
Region 3 is from 0 to +63
Region 4 is from +64 to +127,
As described above, it is possible to make the range of each region uniform. Here, values such as −128 and −65 are called boundary values. Each region represents a range between boundary values.

Also,
Region 1 is from -128 to -60,
Region 2 is from -59 to -5,
Region 3 is from +4 to +58,
Region 4 is from +59 to +127,
As described above, it is possible to set the size of each area to be unequal.

further,
Region 1 is from -128 to -65,
Region 2 is from -64 to -5,
Region 3 is from +4 to +63
Region 4 is from +64 to +127,
As described above, it is possible to set a range that does not belong to any region in the effective range.

  The above is an example of setting the range of each area when the effective range is divided into four areas, but various range settings can be made in addition to the above-described example. Similarly, when the effective range is divided into a plurality of regions other than four, various implementations are possible. In the above description, it is assumed that the motion vector signal S10 is divided into each region sandwiched between boundary values. In another description, the motion vector signal S10 may be quantized. In this case, the above-described region corresponds to the quantization level. In the above example, since there are four regions, the motion vector signal S10 is quantized to four quantization levels.

  In FIG. 1, the region determination unit 11 determines which region the motion vector signal S <b> 10 corresponds to among regions obtained by dividing the effective range of motion vector values in advance. Which region the input motion vector signal S10 corresponds to can be determined by size comparison with boundary values between the regions. The region determination unit 11 is also called a quantization unit, and the determination result S11 of the region determination unit 11 is also called a quantized signal.

  The area distinction number counter unit 12 is provided for each divided area, and sequentially counts the determination result S11 in the corresponding area. The area distinction number counter unit 12 is also called a quantized signal counting part, and the count result S12 of the area distinction number counter part 12 is also called a count quantization signal. Moreover, the vector value accumulation | aggregation part 13 for every area | region is provided for every divided | segmented area | region, and when the determination result S11 is an area | region applicable in an applicable area | region, the motion vector signal S10 is vector-integrated sequentially. That is, for example, the horizontal direction component and the vertical direction component are sequentially integrated. The region-specific vector value integration unit 13 is also called a vector signal integration unit, and the integration result S13 of the region-specific vector value integration unit 13 is also called an integration vector signal.

  A series of operations by the region determination unit 11, the region distinction number counter unit 12, and the region-specific vector value integration unit 13 as described above are performed on the motion vector signal S10, and the result is input. This shows the detection state of the motion vector value. Therefore, by limiting the motion vector signal S10 to a plurality of blocks within the predetermined range S9, a motion vector detection state in the limited plurality of blocks can be obtained.

  For example, when the predetermined range S9 includes a plurality of blocks in the vertical direction of the image signal including the block of interest, the count result S12 of the region-specific number counter unit 12 and the integration result S13 of the region-specific vector value integration unit 13 are: This indicates a detection state of motion vectors in a plurality of blocks in the vertical direction of the image signal including the target block.

  Further, when the predetermined range S9 includes a plurality of blocks in the horizontal direction of the image signal including the block of interest, the count result S12 of the area distinction number counter unit 12 and the integration result S13 of the area-specific vector value integration unit 13 are: This indicates the detection state of motion vectors in a plurality of blocks in the horizontal direction of the image signal including the block of interest.

  Further, when the predetermined range S9 is composed of a plurality of blocks adjacent to the target block, the count result S12 of the region-specific number counter unit 12 and the integration result S13 of the region-specific vector value integration unit 13 include the target block. This indicates a detection state of motion vectors in a plurality of blocks adjacent to the target block of the signal.

  Further, when the predetermined range S9 is composed of all the blocks of one frame, the counting result S12 of the area distinction number counter unit 12 and the integration result S13 of the area-specific vector value integration unit 13 include all of the one frame including the target block. It shows the detection state of the motion vector in the block.

  As described above, the count result S12 of each area of the area distinction number counter unit 12 indicating the motion vector detection state is a representative vector value with respect to the motion vector values of a plurality of blocks in the predetermined range S9. This indicates the degree of whether or not this is an existing area. As an example, the region where the counting result S12 is maximum is a region where the most representative vector value exists for the motion vector values of a plurality of blocks in the predetermined range S9. This area is called a representative area.

  The maximum number area determination unit 14 generates a representative area signal S14A indicating an area where the counting result S12 is maximum and a representative number signal S14B indicating the counting result S12 in the representative area, and each has the most representative vector value. To obtain the representative area and the number of blocks corresponding to the representative area. The maximum number area determination unit 14 is also called a representative quantization level generation unit.

  Further, the integrated value selection unit 15 selects the integration result in the representative region represented by the representative region signal S14A from the integration results S13 of the vector value integration unit 13 for each region, and the representative region where the most representative vector value exists. The representative integrated value signal S15 representing the integrated result of the vector values corresponding to is generated. The integrated value selection unit 15 is also called a representative integrated vector signal generation unit.

  The division unit 16 receives the representative number signal S14B and the representative integrated value signal S15 obtained as described above, and divides the representative integrated value signal S15 by the representative number signal S14B, thereby obtaining the most representative motion vector. A motion vector candidate signal S16 representing an average value of motion vectors in a representative area where the value exists is generated. The average of the motion vectors in the representative area is called a representative vector. The division unit 16 is also called a motion vector candidate signal generation unit.

  The representative vector represented by the motion vector candidate signal S16 is a motion vector representing the detection state with respect to the motion vector value detected in each block in the predetermined range S9, and it can be assumed that the existence probability is high. The target block motion vector detection unit 17 divides the image signal into a plurality of blocks and detects a motion vector for the target block, and then generates a motion vector candidate signal S16 that is a vector value that can be assumed to have a high existence probability. Based on this, a motion vector is detected.

  As described above, in the first embodiment, when the motion vector of the target block is detected, the most representative motion vector is obtained based on the motion vector already obtained in each block in the predetermined range including the target block. The representative vector is used as a motion vector candidate again to detect the motion vector of the block of interest. Thereby, even when a plurality of different motions coexist, it is possible to effectively detect a motion vector with high accuracy. As a result, it is possible to provide high-quality and high-efficiency processing, such as compression encoding of image signals and generation of interpolated images, which are motion-compensated using motion vectors.

  In the first embodiment, the image signal is configured in units of frames, and the search range is set to the previous frame and / or the subsequent frame of one frame including the block of interest. As another embodiment, the image signal may be configured in units of fields, and the search range may be set in one field before and / or after the field including the block of interest.

  The above description of the embodiments is merely an example embodying the present invention. The present invention is not limited to these examples, and can be easily configured by those skilled in the art using the technology of the present invention. It can be expanded to various examples.

  The present invention can be used for a motion vector detection device and a motion vector detection system.

FIG. 2 is a block diagram showing a configuration of a motion vector detection device according to the first embodiment

Explanation of symbols

DESCRIPTION OF SYMBOLS 9 Predetermined range setting part 10 Predetermined range motion vector detection part 11 Area | region determination part 12 Area | region distinct number counter part 13 Area | region vector value integration part 14 Maximum number area | region determination part 15 Integration value selection part 16 Division part 17 Attention block motion vector detection part

Claims (6)

In the image signal divided into a plurality of blocks, a motion vector candidate signal of a reference block is generated based on motion vector signals of N blocks (N is an integer of 2 or more) including the reference block in a predetermined range, and motion A motion vector detection apparatus that uses a vector candidate signal to detect a motion vector signal of a reference block,
Quantizing means for quantizing N motion vector signals and generating N quantized signals respectively;
Quantized signal counting means for counting the number of quantized signals for each quantization level based on the N quantized signals and generating a counted quantized signal;
Vector signal integrating means for vectorizing the motion vector signals for each quantization level based on the N quantized signals and generating an integrated vector signal;
Representative quantization level generating means for generating a representative quantization level representing a quantization level with the largest number based on the count quantization signal, and a representative count quantization signal representing the number at the representative quantization level;
Representative integrated vector signal generating means for generating a representative integrated vector signal representing the integrated vector signal at the representative quantization level;
A motion vector detection apparatus comprising: motion vector candidate signal generation means for generating a motion vector candidate signal by dividing a representative integrated vector signal by a representative count quantized signal. A predetermined range setting means for setting a predetermined range to include N blocks (N is an integer of 2 or more) including a reference block in an image signal divided into a plurality of blocks;
Predetermined range motion vector detection means for generating motion vector signals of N blocks in the predetermined range;
A motion vector detection system comprising: the motion vector detection device according to claim 1.   The motion vector detection system according to claim 2, wherein the predetermined range setting unit sets the predetermined range in a vertical direction of the image signal.   The motion vector detection system according to claim 2, wherein the predetermined range setting unit sets the predetermined range in a horizontal direction of the image signal.   The motion vector detection system according to claim 2, wherein the predetermined range setting unit sets the predetermined range so as to include two or more blocks adjacent to the reference block.   The motion vector detection system according to claim 2, wherein the predetermined range setting means sets the predetermined range so that all blocks in one frame are included.

Images