JP2009272765A - Motion vector detection apparatus and motion vector detection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a motion vector detection apparatus and a motion vector detection method, for detecting a motion vector at high speed. <P>SOLUTION: The motion vector detection apparatus is provided with: a plurality of hierarchy search parts for searching for a motion vector in a target hierarchy; and an operation part for operating a sum of motion vectors in respective hierarchies searched for by the plurality of hierarchy search parts as a motion vector to obtain. In one or more hierarchy search part having all the plurality of hierarchy search parts or hierarchies other than the lowest hierarchy as a target, a frame image is binarized with a threshold which is set higher the greater a block size is with respect to the degree of pattern complication in the search block as a reference by having a search block larger than a lower hierarchy as a unit and a motion vector in the target hierarchy is searched for with the binarized search block as a unit. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a motion vector detection device and a motion vector detection method, and is suitable for the image field.

  Conventionally, the present applicant has proposed a motion vector detection apparatus that searches for a motion vector in a target image with different resolutions for each of a plurality of hierarchies, and uses the sum of the searched motion vectors in each hierarchy as a motion vector to be obtained. (For example, refer to Patent Document 1).

  Specifically, in a higher-level processing system, for each search block of a predetermined size, each pixel value of the search block is compared with a value that is 1/2 of the sum of the maximum value and the minimum value of the brightness in the search block. The code is converted into a binary code of “1” or “0”, and then block matching is performed.

Further, in the processing system of the middle hierarchy, for each search block smaller than the search block of the upper hierarchy, each pixel value of the search block is converted into a binary code in the same manner as the upper hierarchy, and then block matching is performed. Is done.
Japanese Patent Laid-Open No. 08-088855

  By the way, in such a motion vector detection device, since the search block is converted into a binary pattern in units of pixels, the amount of data in the search block is small, and the amount of calculation of block matching can be suppressed accordingly. As a result, the motion vector to be obtained (the sum of motion vectors in each layer) can be obtained at high speed.

  However, in order to further increase the calculation speed, increase the number of hierarchies to cope with a large motion vector, or expand the search range, the amount of calculation tends to increase, and further simplification It is desired.

  The present invention has been made in consideration of the above points, and intends to propose a motion vector detection device and a motion vector detection method capable of detecting a motion vector at high speed.

  In order to solve such a problem, the present invention is a motion vector detection device, and includes a plurality of hierarchical search units that search for a motion vector of a target hierarchy, and a motion vector of each hierarchy searched by the plurality of hierarchical search units An arithmetic unit that calculates a sum of the motion vectors to be obtained, and one or more hierarchical search units that target all layers other than the lowest layer or a hierarchy other than the lowest layer, Using a search block larger than the lower layer as a unit, binarization based on a threshold set higher as the block size is larger with respect to the complexity of the pattern in the search block, and using a binarized search block as a unit, Search for a motion vector of the hierarchy to be targeted.

  The present invention is also a motion vector detection method, wherein an input step of inputting a frame image to one or two or more hierarchical search units targeted for all of a plurality of hierarchical search units or a layer other than the lowest layer In one or more hierarchical search units targeting all or a plurality of hierarchical search units other than the lowest layer, a frame image is a picture in the search block in units of search blocks larger than the lower layer. A search step for searching for a motion vector of a hierarchy to be processed in units of a binarized search block, with a threshold set higher as the block size is larger with respect to the complexity of The sum of the motion vectors of each hierarchy searched by the hierarchy search unit is calculated as a motion vector to be obtained.

  As described above, according to the present invention, a frame image is binarized for each search block with a low resolution (block size) covering a relatively wide range according to the complexity of the design of the search block, By searching for the motion vector of the target hierarchy for each search block, the amount of data required for the search block can be reduced as compared with the case where each pixel in the search block is binarized. Therefore, compared with the case where each pixel in the search block is binarized, the amount of calculation required for searching for a motion vector of the target hierarchy can be suppressed, and thus a motion vector that can detect a motion vector at high speed. A detection apparatus and a motion vector detection method can be realized.

  Hereinafter, an embodiment to which the present invention is applied will be described in detail with reference to the drawings.

(1) Overall Configuration of Motion Vector Detection Device FIG. 1 shows the overall configuration of a motion vector detection device 1 according to this embodiment. Data indicating the frame image is input to the motion vector detection device 1 in order of time series.

  As shown in FIG. 2, the dividing unit 2 divides the frame image into, for example, vertical and horizontal 8 pixels × 8 pixel blocks (hereinafter also referred to as reference blocks), and the data of these reference blocks is converted into the complexity detection unit 3. And the third hierarchy processing unit 30.

  The complexity calculation unit 3 calculates the difference between the minimum value and the maximum value (hereinafter also referred to as a dynamic range) for each reference block using the data given from the dividing unit 2, and indicates the calculation result. Data is given to the first hierarchical processing unit 10 and the second hierarchical processing unit 20, respectively. As the dynamic range is larger, the reference block tends to include complex patterns such as edges and details.

  As shown in FIG. 3, the first hierarchical processing unit 10 has a motion vector (hereinafter, referred to as a first hierarchical block) in units of a block of 32 pixels × 32 pixels (hereinafter, also referred to as a first hierarchical block) that is larger than the reference block. This is a processing unit for calculating (also referred to as a first layer motion vector). Specifically, the first layer processing unit 10 includes a first layer complexity calculation unit 11, a binarization unit 12, a frame delay unit 13, and a block matching calculation unit 14.

  The first hierarchy complexity calculation unit 11 uses the data (the maximum and minimum luminance values in each reference block or the dynamic range that is the difference) provided from the complexity calculation unit 3 for each first hierarchy block. The dynamic range is calculated, and data indicating the calculation result is given to the binarization unit 12.

  The first hierarchy complexity calculator 11 uses the maximum and minimum luminance values in the reference block (8 pixels × 8 pixels) already obtained by the complexity calculator 3 or the dynamic range that is the difference between them. As compared with the case of calculating from the beginning without using this, the calculation can be performed with a simple calculation.

  Incidentally, as this calculation, for example, the maximum value among the 16 maximum values respectively calculated by the complexity calculation unit 3 for the 16 reference blocks corresponding to the first hierarchical block, and the 16 minimum values There is a method of obtaining a minimum value in the above by a comparison operation and obtaining a new dynamic range from the difference between them. Further, for example, there is a method of obtaining the maximum value of the 16 dynamic ranges calculated by the complexity calculation unit 3 for the 16 reference blocks corresponding to the first hierarchical block.

  The binarizing unit 12 uses the data respectively given from the dividing unit 2 and the first layer complexity calculating unit 11 to generate a frame image in units of first layer blocks based on the threshold value TH1 set by the threshold setting unit. The binarization is performed, and the binarized data of each first layer block is supplied to the frame delay unit 13 and the block matching calculation unit 14, respectively.

  Therefore, as shown in FIG. 4, the binarizing unit 12 can roughly represent a complicated pattern portion and a simple pattern portion in the frame image. Specifically, it can be represented by 1-bit data for each first hierarchical block (vertical and horizontal 32 pixels × 32 pixels).

  The frame delay unit 13 temporarily stores the data supplied from the binarization unit 12 and delays the data by a time corresponding to one frame, and then supplies the data to the block matching calculation unit 14. In this frame delay unit 13, the data provided from the binarization unit 12 is represented by 1 bit, so that the memory capacity can be greatly reduced as compared to the case where binarization is not performed.

  The block matching calculation unit 14 uses the data given from the binarization unit 12 and the data given from the frame delay unit 13 to calculate each first layer block in the current and past frame images as shown in FIG. , Each size is reduced to a size corresponding to the reference block.

  Then, the block matching calculation unit 14 calculates the first layer motion vector (motion amount and direction) of the dynamic target in the current frame image with respect to the past frame image in units of the reduced first layer block, and the calculation result Is provided to the second hierarchical processing unit 20 and the vector addition unit 4.

  Here, a specific calculation method of the first layer motion vector in this embodiment will be described. That is, the block matching calculation unit 14 includes all or part of the first layer blocks (white blocks in FIG. 5) having a dynamic range (complex picture) equal to or higher than the threshold TH1 among the first layer blocks in the current frame image. Are determined as candidate blocks in a predetermined order.

  When the block matching calculation unit 14 determines the candidate block, as shown in FIG. 6, the candidate block is searched from the search range AR1 based on the portion corresponding to the candidate block NB in the past frame image. A portion (indicated by a one-dot difference line in FIG. 6) where the absolute value sum of luminance differences from NB is minimum is searched.

  Then, the block matching calculation unit 14 calculates the motion amount and the direction between the searched portion and the portion corresponding to the candidate block NB as a first layer motion vector (indicated by an arrow in FIG. 6).

  Since the first hierarchical block (vertical and horizontal 32 pixels × 32 pixels) as a candidate target is 1 bit, the block matching calculation unit 14 performs a simple calculation (exclusive OR operation) on a bit-by-bit basis. As a result, the amount of calculation can be greatly reduced.

  In addition to this, the block matching calculation unit 14 is limited to a part of the number of candidate blocks that should be candidates in the current frame image, and limited to a part of the candidate range in the past frame image. Further, the amount of calculation can be further reduced.

  On the other hand, as shown in FIG. 7, the second hierarchy processing unit 20 is a block of 16 pixels × 16 pixels, for example, which is larger than the reference block and smaller than the first hierarchy block (hereinafter also referred to as a second hierarchy block). Is a processing unit that calculates a motion vector (hereinafter also referred to as a second layer motion vector). Specifically, the second hierarchy processing unit 20 includes a second hierarchy complexity calculation unit 21, a binarization unit 22, a frame delay unit 23, and a block matching calculation unit 24.

  The second hierarchy complexity calculation unit 21 uses the data (the maximum and minimum luminance values in each reference block, or the dynamic range that is the difference between them) given from the complexity calculation unit 3 for each second hierarchy block. The dynamic range is calculated, and data indicating the calculation result is given to the binarization unit 22.

  In the same way as the first hierarchy complexity calculator 11, the second hierarchy complexity calculator 21 does not use the dynamic range of the reference block (vertical and horizontal 8 pixels × 8 pixels) already obtained by the complexity calculator 3. Compared to the above, the dynamic range of the second hierarchical block (vertical and horizontal 16 pixels × 16 pixels) larger than the reference block can be calculated by a simple calculation. Incidentally, this calculation may be the same as or different from the first hierarchy complexity calculation unit 11.

  The binarizing unit 22 uses the data respectively given from the dividing unit 2 and the second hierarchical complexity calculating unit 21 to generate a frame image in units of second hierarchical blocks based on the threshold TH2 set by the threshold setting unit. The data is binarized, and the binarized data of each second layer block is supplied to the frame delay unit 23 and the block matching calculation unit 24, respectively.

  The threshold TH2 is searched for a motion vector in units of a second hierarchical block (vertical and horizontal 16 pixels × 16 pixels) finer than the first hierarchical block (vertical and horizontal 32 pixels × 32 pixels). It is set as a value lower than the set threshold value TH1.

  Therefore, as shown in FIG. 8, the binarizing unit 22 can roughly express a complicated pattern portion and a simple pattern portion in the frame image although they are finer than the first layer. Specifically, the second hierarchical block (vertical 16 pixels × 16 pixels) is doubled vertically and doubled twice in total with respect to the first hierarchical block (vertical 32 pixels × 32 pixels). Each hierarchical block can be represented by 4-bit data.

  Similar to the frame delay unit 13, the frame delay unit 23 temporarily stores the data supplied from the binarization unit 22, delays the data by a time corresponding to one frame, and then performs a block matching operation unit. 24. In this frame delay unit 23, the data provided from the binarization unit 22 is 4 bits, so that the memory capacity can be greatly reduced as compared with the case where the binarization is not performed.

  The block matching calculation unit 24 uses the data given from the binarization unit 22 and the data given from the frame delay unit 23 to calculate each second hierarchical block in the current and past frame images as shown in FIG. , Each size is reduced to a size corresponding to the reference block.

  Then, the block matching calculation unit 24 calculates the second layer motion vector (motion amount and direction) of the dynamic target in the current frame image with respect to the past frame image in units of the reduced second layer block, and the calculation result Is provided to the third hierarchical processing unit 30 and the vector addition unit 4.

  Here, a specific calculation method of the second layer motion vector in this embodiment will be described. In other words, the block matching calculation unit 24 adds all or a part of the second layer block (white block in FIG. 9) having a dynamic range (complex picture) equal to or higher than the threshold TH2 among the second layer blocks in the current frame image. Are determined as candidate blocks in a predetermined order.

  In addition, when the candidate block is determined, the block matching calculation unit 24 uses the data provided from the block matching calculation unit 14 of the first hierarchical processing unit 10 to start from the portion corresponding to the candidate block in the past frame image. A position shifted by the first layer motion vector is recognized.

  Then, the block matching calculation unit 24 uses the recognized position as a reference and selects a portion in which the absolute value sum of the luminance difference from the candidate block is minimum from the search range smaller than the search range used by the block matching calculation unit 14. A second layer motion vector is calculated by searching.

  Since the second hierarchical block (vertical and horizontal 16 pixels × 16 pixels) as a candidate target is 4 bits, the block matching calculation unit 24 performs a simple calculation (exclusive OR operation) on a bit-by-bit basis. As a result, the amount of calculation can be greatly reduced.

  In addition to this, the block matching calculation unit 24 limits the number of candidate blocks to be candidates in the current frame image to a part, and sets the candidate range in the past frame image to be narrower than the first hierarchy. The amount of computation can be further reduced in terms of the point.

  On the other hand, as shown in FIG. 2, the third hierarchy processing unit 30 calculates a motion vector (hereinafter also referred to as a third hierarchy motion vector) in units of reference blocks (vertical and horizontal 8 pixels × 8 pixels). It is a processing unit. Specifically, the third layer processing unit 30 includes a compensation unit 31, a frame delay unit 32, and a block matching calculation unit 33.

  The compensation unit 31 temporarily stores the data given from the division unit 2 and compensates for the time delay required for the processing in the first hierarchical processing unit 10 or the second hierarchical processing unit 20, and then the frame delay unit 32 and Each is given to the block matching calculation unit 33.

  Similar to the frame delay units 13 and 23, the frame delay unit 32 temporarily stores the data supplied from the compensation unit 31 and delays the data by a time corresponding to one frame, and then the block matching calculation unit. 33.

  The block matching calculation unit 33 calculates the third-layer motion vector (motion amount and direction) of the dynamic target in the current frame image with respect to the past frame image, using the reference block as a unit, and data indicating the calculation result as a vector This is given to the adder 4.

  Here, a specific calculation method of the third layer motion vector in this embodiment will be described. That is, the block matching calculation unit 33 determines all or part of the third hierarchical blocks in the current frame image as candidate blocks in a predetermined order.

  In addition, when the block matching calculation unit 33 determines a candidate block, the data provided from the block matching calculation unit 14 of the first hierarchy processing unit 10 and the data provided from the block matching calculation unit 24 of the second hierarchy processing unit 20 Is used to recognize a position shifted from the portion corresponding to the candidate block in the past frame image by the first layer motion vector and the second layer motion vector.

  Then, the block matching calculation unit 24 uses the recognized position as a reference, and selects a portion in which the absolute value sum of the luminance difference from the candidate block is minimum from the search range smaller than the search range used by the block matching calculation unit 24. A third layer motion vector is calculated by searching.

  The block matching calculation unit 33 can reduce the calculation amount in that the candidate range in the past frame image is limited to a range narrower than the second hierarchy.

  The vector addition unit 4 includes data provided from the block matching calculation unit 14 of the first hierarchy processing unit 10, data provided from the block matching calculation unit 24 of the second hierarchy processing unit 20, and block matching calculation of the third hierarchy processing unit 30. Using the data provided from the unit 33, as shown in FIG. 10, the motion vectors V1 to V3 calculated in the respective layers are added, and data indicating the addition result is output.

  Since this data indicates the sum of low-resolution motion vectors (first-layer motion vector to third-layer motion vector) calculated in each layer, the motion vector when viewed in units of pixel levels in the frame image Is the true motion vector to be obtained.

  In this way, the motion vector detection device 1 detects motion vectors at the search levels of the highest hierarchy (first hierarchy), middle hierarchy (second hierarchy), and lowest hierarchy (third hierarchy) that are larger than the pixels. By searching, the search for the motion vector at the pixel level is speeded up.

(2) Motion Vector Detection Processing Procedure Next, a motion vector detection processing procedure in the motion vector detection device 1 will be described. As shown in FIG. 11, the motion vector detection device 1 waits for the first frame image in step SP1, and when receiving the frame image, proceeds to the next step SP2 and proceeds to the first to third layers. Each processing unit 10, 20, 30 is input.

  Then, the motion vector detection apparatus 1 proceeds to step SP3 and causes the processing units 10, 20, and 30 to search for the motion vector of the target layer. That is, the first hierarchical processing unit 10 searches for a motion vector at a binary level (FIG. 4) in units of blocks (32 pixels × 32 pixels) corresponding to the lowest resolution (FIG. 3).

  On the other hand, the second hierarchical processing unit 20 searches for a motion vector at a binary level (FIG. 8) in units of blocks (vertical and horizontal 16 pixels × 16 pixels) corresponding to intermediate resolution (FIG. 7). On the other hand, the third layer processing unit 30 is caused to search for a motion vector at a multi-value level in units of blocks (vertical and horizontal 8 pixels × 8 pixels) corresponding to the highest resolution (FIG. 3).

  In addition, when the motion vectors of the target layer are searched for by the processing units 10, 20, and 30, the motion vector detection device 1 proceeds to the next step SP4, and obtains the sum of these motion vectors for the frame image. Calculated as a power vector.

  Thereafter, in step SP5, the motion vector detection device 1 returns to step SP2 if an unsearched frame image remains, and ends the motion vector detection processing procedure if it does not remain.

  In this way, the motion vector detection device 1 is adapted to detect the motion vector of the frame image given in chronological order.

(3) Operation and Effect In the above configuration, the motion vector detection device 1 in the first layer processing unit 10 is a block (vertical and horizontal 32 pixels × 32 pixels) larger than the lower layer block (vertical and horizontal 16 pixels × 16 pixels). The frame image is binarized on the basis of the threshold TH1 (FIG. 4), and the motion vector of the highest layer (low resolution) for the frame image is searched.

  On the other hand, in the second layer processing unit 20, the motion vector detection device 1 uses the upper layer threshold TH <b> 1 in units of blocks (vertical and horizontal 16 pixels × 16 pixels) larger than the lower layer blocks (vertical and horizontal 8 pixels × 8 pixels). The frame image is binarized based on a threshold TH2 set lower than that (FIG. 8), and the frame image is smaller than the first layer from the position corresponding to the motion vector searched in the first layer. A motion vector in the second layer (medium resolution) is searched in the search range.

  On the other hand, in the third layer processing unit 30, the motion vector detection device 1 responds to the motion vector searched in the first layer and the second layer with the minimum block (vertical and horizontal 8 pixels × 8 pixels) as a unit (FIG. 2). The motion vector of the third layer (high resolution) is searched from the determined position within the search range smaller than the second layer.

  Then, the motion vector detection device 1 calculates the sum of motion vectors searched in each layer as a motion vector to be obtained (FIG. 10).

  Therefore, this motion vector detection device 1 is a unit of 1 bit in the first layer and a unit of 4 bits in the second layer, and the motion vector is obtained by a simple operation (exclusive OR operation) (by block matching). It is possible to search and to greatly reduce the amount of calculation.

  Here, FIG. 12 shows a difference in data amount between the hierarchical motion vector by this technique and the hierarchical motion vector by the technique described in the cited document 1. As is clear from FIG. 12, in this method, the first hierarchical block and the second hierarchical block are binarized as a unit, so that binary values are obtained in units of pixels in the first hierarchical block and the second hierarchical block. It can be seen that the amount of data is greatly reduced compared to the conventional method converted into a code, and the motion vector can be searched by a simple operation (exclusive OR operation).

  In the conventional method, the block is code-converted in units of pixels on the basis of 1/2 of the sum of the maximum value and the minimum value in the block. In other words, since the pixels are binarized based on the variable value determined according to the complexity of the picture in the block, the block can be patterned with high accuracy according to the picture.

  However, as in the present method, even when the binary assignment criterion (threshold values TH1 and TH2) for the block is fixed and the block is binarized, it is possible to maintain approximately the same accuracy as the conventional method. Confirmed by the applicant. The thresholds TH1 and TH2 in this confirmation are thresholds that are linearly related to the block size as shown in FIG.

  The amount of change in the pixel value (picture) in the entire block tends to converge regardless of the picture in the block as the block size is increased (the motion vector is searched at a lower resolution). Therefore, in the conventional method, reducing the binarization unit in the block and making the binarization standard variable according to the complexity of the picture are not reflected in the accuracy, but rather are a factor of useless calculation. Become. On the other hand, this method is useful in that it can maintain approximately the same accuracy as the conventional method while avoiding this factor.

  According to the above configuration, since a motion vector is searched at a binary level with a block corresponding to a low resolution covering a relatively wide range as a unit, a binary allocation criterion (threshold TH1, TH1, Even when TH2) is fixed, a motion vector can be searched by a simple calculation while maintaining a predetermined accuracy, and thus a motion vector can be detected at high speed.

(4) Other Embodiments In the complexity calculation unit 3 in the above-described embodiment, the difference (dynamic range) between the maximum value and the minimum value of the luminance in the target block is calculated. However, this calculation target is not limited to this embodiment.

  For example, the degree of separation between the maximum value and the minimum value in the target block, such as the ratio between the maximum value and the minimum value of luminance, can be set as the calculation target. Also, the absolute value of the luminance difference between adjacent pixels in the target block can be calculated. In short, any object that indicates the degree of change (complexity of a pattern) in the target block can be applied as a calculation target.

  Moreover, although the threshold value setting unit in the above-described embodiment has been described above that the threshold values TH1 and TH2 to be set in the binarization units 12 and 22 may be fixed, it may be variable.

  As a specific example, there is an aspect in which the thresholds TH1 and TH2 are varied according to the frequency distribution of the dynamic range in each reference block of the frame image and the average thereof. As these parameters are related to the complexity of the pattern, as shown in FIG. 14, the frequency distribution width of the dynamic range becomes wider as the pattern included in the entire frame image has more edges and details. The average of the frequency distribution exhibits a high value.

  Therefore, if the threshold values TH1 and TH2 are set higher as the frequency distribution width of the dynamic range in each reference block and / or the average thereof are larger, the complex picture portion and the simple picture portion are roughly divided. However, it becomes possible to express more accurately. In this way, it is possible to separate a block that truly includes edges and details from a block that does not include edges with high accuracy.

  As another specific example, an input unit for inputting the block size of the first hierarchical block, the second hierarchical block, and the third hierarchical block (reference block) that is a target in each layer is provided, and the block is input from the input unit There is an aspect in which the thresholds TH1 and H2 are varied according to the size. In this way, according to the frame image to be applied, the user can arbitrarily specify the detection accuracy of the motion vector. Incidentally, in this case, as shown in FIG. 13, the thresholds TH1 and H2 may be determined according to the table in which the block size and the threshold are in a linear relationship and the block size input from the input unit. Is possible.

  In addition, although the binarization units 12 and 22 in the above-described embodiment use one binarization method, it is also possible to use one binarization method selected from the two binarization methods.

  For example, the binarization units 12 and 22 acquire the frequency distribution width and average of the dynamic range of the reference block unit in the frame image to be processed, and the thresholds provided for the frequency distribution width and the average. .

  Here, when both the frequency distribution width and the average thereof are less than the corresponding threshold values, the binarizing units 12 and 22 assume that the complexity of the pattern of the frame image to be processed is not so high, The binarization method in the form is used. That is, the target block (first hierarchical block, second hierarchical block) is binarized according to the magnitude of the dynamic range or the like in the target block with respect to the reference values (threshold values TH1 and TH2).

  On the other hand, when one or both of the frequency distribution and the average thereof are equal to or higher than the corresponding threshold values, the binarization units 12 and 22 assume that the complexity of the pattern of the frame image to be processed is high, and the ADRC A binarization method by conversion (the method of Cited Document 1) is used. That is, each pixel in the target block (first hierarchical block, second hierarchical block) is represented by a code of “0” or “1” or n depending on the dynamic range in the target block with respect to the reference values (threshold values TH1, TH2). Encodes a bit code.

  In this way, according to the complexity of the pattern in the entire frame image, the binarization method in this embodiment and the binarization method by ADRC conversion that is more accurate than the binarization method but has a large load are used. By switching, it is possible to grasp the movement of the pattern portion more accurately while suppressing processing addition as much as possible.

  In addition, the block matching calculation units 14 and 24 in the above-described embodiment convert the frame image (FIGS. 4 and 8) binarized in units of blocks of each layer into the reference block before obtaining the motion vector of the target layer. (FIGS. 5 and 9), the reduction process may be omitted.

  In the above embodiment, simple motion vectors of three layers are obtained. However, the number of layers is not limited to three, and may be any number as long as it is two or more layers.

  Further, in the above-described embodiment, the motion vector of the lowest layer is searched at the multi-value level in units of the minimum block (vertical and horizontal 8 pixels × 8 pixels). However, this lowest layer motion vector may also be searched at the binary level.

  The present invention can be used in the field of image processing.

It is a block diagram which shows the whole structure of the motion detection apparatus by this Embodiment. It is a photograph which shows the example of a division | segmentation of a reference | standard block. It is a photograph which shows the example of a block used as the unit at the time of calculating the motion vector of a 1st hierarchy. It is the schematic which shows the binarization example of a 1st hierarchy block. It is the schematic which shows the example of reduction of a 1st hierarchy block. It is the schematic where it uses for description of block matching. It is a photograph which shows the example of a block used as the unit at the time of calculating the motion vector of a 2nd hierarchy. It is the schematic which shows the binarization example of a 2nd hierarchy block. It is the schematic which shows the example of reduction of a 2nd hierarchy block. It is the schematic where it uses for description of the addition of the motion vector obtained in each hierarchy. It is a flowchart which shows a motion vector detection process procedure. It is the schematic where it uses for description of the difference in the data amount by the conventional method and this method which are required for a hierarchical motion vector. It is the schematic which shows the relationship between a block size and the binarization reference value (threshold value) of a block. It is the schematic which shows the relationship between the frequency distribution and the average of the dynamic range of the reference | standard block unit in a frame image, and the pattern of a frame image.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Motion vector detection apparatus, 2 ... Division | segmentation part, 3 ... Complexity calculation part, 4 ... Vector addition part, 10 ... 1st hierarchy process part, 20 ... 2nd hierarchy process part, 30 ... Third layer processing unit, 11, 21... First layer complexity calculation unit, 12, 22... Binary processing unit, 13, 23, 32... Frame delay unit, 14, 24, 33. Part, 31 ... compensation part.

Claims (8)

A plurality of hierarchical search units for searching for a motion vector of a target hierarchy;
An arithmetic unit that calculates a sum of motion vectors of each layer searched by the plurality of hierarchical search units as a motion vector to be obtained;
One or two or more hierarchy search units targeting all or a hierarchy other than the lowest layer of the plurality of hierarchy search units,
The frame image is binarized based on a threshold value set higher as the block size is larger with respect to the complexity of the pattern in the search block in units of search blocks larger than the lower layer, and binarized search block A motion vector detection device that searches for a motion vector of a hierarchy to be processed in units of. A first calculation unit that calculates the complexity of the pattern in the frame image for each search block assigned as the minimum size to the lowest layer;
A pattern in a search block that is a unit in one or more search units targeting a layer other than the lowest layer, using the complexity of the pattern in each search block calculated by the first calculation unit. The motion vector detection device according to claim 1, further comprising: a second calculation unit that calculates the complexity of.   The motion according to claim 2, wherein the first calculation unit calculates the degree of separation between the maximum value and the minimum value in the search block assigned as the minimum size to the lowest layer as the complexity of the pattern. Vector detection device. A search in which a threshold having a linear relationship with the block size is used as a unit for one or two or more hierarchy search parts that target all or a hierarchy other than the lowest layer among the plurality of hierarchy search parts. The motion vector detection device according to claim 3, further comprising a threshold value setting unit that is set according to a block size of the block. The input part which inputs the block size of the search block which should be made into a unit with respect to one or two or more hierarchy search parts for all the said multiple hierarchy search parts or hierarchies other than the lowest layer. 5. The motion vector detection device according to 4. The distribution of the degree of difference between the maximum value and the minimum value in each of the search blocks with respect to one or more hierarchy search parts that target all or a hierarchy other than the lowest layer of the plurality of hierarchy search parts The motion vector detection device according to claim 3, further comprising: a threshold setting unit configured to set a higher threshold as the average of the distribution is larger. One or two or more hierarchy search units targeting all or a hierarchy other than the lowest layer of the plurality of hierarchy search units,
When both or one of the distribution spread and the average of the distribution of the degree of difference between the maximum value and the minimum value in each of the search blocks exceeds a predetermined value, each block in the block is set in units of blocks larger than the lower layer. The motion vector detection device according to claim 6, wherein the pixel is binarized based on the threshold value. An input step for inputting a frame image to one or more hierarchical search units targeting all or a hierarchy other than the lowest layer of the multiple hierarchical search units;
In one or two or more hierarchical search units targeting all or a hierarchy other than the lowest layer of the plurality of hierarchical search units, the frame image is a search block larger than the lower layer as a unit, and the pattern in the search block A search step for binarizing a threshold value set higher as the block size is larger with respect to the complexity, and searching for a motion vector of a hierarchy to be processed in units of binarized search blocks;
A motion vector detection method comprising: an operation step of calculating a sum of motion vectors of each hierarchy searched by the plurality of hierarchy search units as a motion vector to be obtained.