JP2010074496A - Motion vector detection processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a motion vector detection processing apparatus capable of reducing an operation amount for coding without degrading the efficiency of coding. <P>SOLUTION: A reduced image is generated by a reduced image generator, and a motion vector is detected from the reduced image by a reduced motion vector detector. Next, an error value between a block centering around a pixel point around the tip of the motion vector of the reduced image and a block to be processed is calculated by a peripheral error value calculator. A search area set for a pixel point with the smallest error value, and shifted from the tip of the motion vector of the reduced image in the direction of the pixel point is set by a search area setting unit, and matching is performed by original resolution within the search area to generate the last motion vector. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

In recent years, it has become common to handle moving image information as digital signals. As an encoding format of moving image information, MPEG (Moving Picture Experts Group), which is an international standard, or H.264 is used. 26x or the like is used. In Japan, moving image information includes digital terrestrial broadcasting, BS digital broadcasting, and DVD (Digital
Versatile Disk) is MPEG-2, mobile phone videophone is MPEG-4, and one-segment broadcasting is H.264. It is compression-encoded with H.264 / AVC. In next-generation DVDs such as Blu-ray discs and HD DVDs, MPEG-2, H.264 and H.264 are used as compression encoding methods for moving images. H.264 / AVC and VC-1. Furthermore, there is AVS as a moving picture compression encoding system standardized in China.

  These video compression coding systems employ motion compensated interframe prediction. It is known that moving image signals have high signal temporal redundancy. Motion compensated interframe prediction is a technique for removing this redundancy. In the motion compensation interframe prediction technique, an image frame to be encoded is divided into a plurality of blocks, and processing is performed in units of blocks. That is, for each block, from the motion detection range located in the vicinity of the target block in an image frame (hereinafter referred to as a reference image frame) that has been previously coded in time, the same size as the target block is set. A plurality of blocks (hereinafter referred to as reference image blocks) are cut out, and among these reference image blocks, a reference image block closest to the target block is searched for.

  Next, information on the position coordinates of the closest reference block and difference information between the target block and the searched reference image block are encoded, thereby generating a code string for the target block. The motion detection is a process of searching for the most similar reference image block, and the motion vector is a relative position coordinate of each target block and the searched reference image block viewed from the block. Such motion detection is described in H.264. 264, MPEG-2, MPEG-4, VC-1, AVS, etc., are performed by normal image signal encoding processing.

Hereinafter, a conventional motion vector search method for searching for such a motion vector will be described.
FIG. 10 is a diagram illustrating the configuration of the encoding target image.

  The encoding target image 10 includes a plurality of macroblocks 11 composed of 16 horizontal pixels and 16 vertical pixels. For example, in the case of a high definition TV (HDTV) resolution moving image consisting of 1088 pixels vertically and 1920 pixels horizontally, the macroblocks are arranged vertically 68 (= 1088/16) and 120 horizontally (= 1920/16). .

FIG. 11 is a diagram for explaining the configuration of the reference image 20 for detecting a motion vector by the conventional motion vector detection method.
The reference image 20 is an image that is encoded before the encoding target image, and has the same shape and the same number of pixels as the encoding target image 10.

  In the reference image 20, a reference area 21 having a rectangular shape larger than that of the encoding block 12 is set at a position including the encoding block 12. In the reference area 21, a reference block 22 having the same size and shape as the encoding block 11 is set.

When the coding block 12 shown in FIG. 10 is selected as a block for detecting a motion vector, the reference block 22 is arranged in the coding block 11 while being moved one pixel at a time in the reference area 21. The difference absolute value sum S expressed by the equation (1) is calculated between the pixel value of the pixel and the pixel value of the pixel arranged in the reference block 22.
S = Σ | C (i, j) -R (x + i, y + j) | ... Formula (1)
Here, Σ represents summation at i = 0, 1, 2,..., N, j = 0, 1, 2,..., M, and (n + 1) is the horizontal direction of the coding block. The number of pixels, (m + 1) is the number of pixels in the vertical direction of the coding block, C (i, j) is the pixel value at the position of row j and column i of the coding block to be processed, and x is the displacement in the horizontal direction of the reference position The quantity y represents the amount of displacement in the vertical direction of the reference position, and R (x + i, y + j) represents the pixel value in the j row and i column of the reference block at the point of displacement (x, y). Then, the position (x, y) of the reference block 22 that minimizes the absolute difference value sum S is detected as the motion vector 23. When a macro block is employed as the encoding block, n = 15 and m = 15. In moving image encoding, the same processing is performed on all macroblocks included in the encoding target image to detect a motion vector.

  The method of calculating the sum of absolute differences at all points in the reference region 21 and examining the motion vector omnipresently is called a full search method. In such a full search method, since all the pixels in the reference area 21 are examined brute-force, the calculation amount becomes enormous.

  Therefore, a hierarchical search method is known as a search method for reducing the amount of calculation. FIG. 12 is a schematic diagram for explaining a reduced coded image 30 in a motion vector detection device using a conventional hierarchical search method, and FIG. 13 is a motion vector detection device using a conventional hierarchical search method. FIG. 14 is a schematic diagram for explaining a method of detecting a motion vector in a motion vector detection apparatus using a conventional hierarchical search method.

  In the hierarchical search method, first, the encoded image 10 is reduced to 1 / N in the horizontal direction, and the reduced encoded image 30 is reduced to 1 / M in the vertical direction. The reduced encoded image 30 has the same number of macroblocks as the encoded image 10. However, the number of pixels included in one macroblock 31 is 1 / N times in the horizontal direction and 1 / M times in the vertical direction. Similarly, a reduced reference image 40 is generated by reducing the reference image 20 by 1 / N in the horizontal direction and by 1 / M in the vertical direction. In the reduced reference image 40, the reference area 21 becomes a reduced reference area 41 that is 1 / N times in the horizontal direction and 1 / M times in the vertical direction. In the primary search, the reduced motion vector 43 is detected in the reduced coding block and the reduced search range 41.

  In the secondary search, motion vector detection is performed between the encoding block 12 and the reference image 20. The search range is N pixels in the horizontal direction and M pixels in the vertical direction indicated by 51 in FIG. 14. The center point of the search range is the position of the vector 52 obtained by multiplying the reduced motion vector 43 N times in the horizontal direction and M times in the vertical direction. It becomes. The position of the reference block that gives the minimum error value in the secondary search is obtained as the motion vector 53 of the coding block 12.

  For example, when the search range is an area composed of 15 × 15 pixels of horizontal ± 7 and vertical ± 7, the amount of calculation necessary for detecting a motion vector for one encoded block is the above-described all search method. In this case, 15 × 15 = 225 matching of 16 × 16 pixels, that is, 57600 pixel matching is required. On the other hand, in the horizontal 1/4 and vertical 1/4 hierarchical search methods, (15/4) × (15/4) = about 25 times 4 × 4 pixel matching and secondary search in the primary search. The motion vector 53 can be detected by 4 × 4 = 16 16 × 16 pixel matchings, that is, 4496 pixel matchings. Therefore, the calculation amount is about 1/12 of the calculation amount in the above-described full search method.

Thus, according to the hierarchical search method, the calculation amount can be significantly reduced as compared with the above-described full search method.
Patent Document 1 describes a motion vector detection device that can reduce the amount of calculation for encoding without degrading the encoding efficiency.

Patent Document 2 describes a hierarchical motion vector detection method that can avoid the possibility of erroneous detection of motion vectors.
Patent Document 3 discloses a motion vector detection device capable of realizing IC miniaturization, low power consumption, and low cost.
JP 2004-193906 A Japanese Unexamined Patent Publication No. 7-154801 Japanese Patent Laid-Open No. 2000-102015

  In recent years, the resolution of moving images has been increased. In particular, there is an increasing demand for encoding a high-resolution moving image even in a moving image capturing function of a portable moving image capturing device such as a mobile phone. When moving picture encoding is performed with a portable device driven by a built-in battery, there is a demand to reduce power consumption as much as possible due to the limitation of battery capacity, and it is necessary to realize moving picture encoding processing with low power consumption. In order to reduce the power consumption of the moving image encoding process, it is effective to reduce the amount of calculation. However, as a result of reducing the amount of calculation in order to realize low power consumption, the motion vector detection accuracy is lowered and the coding efficiency is not allowed to deteriorate.

  The present invention has been made to solve such a problem, and it is an object of the present invention to provide a motion vector detection processing apparatus capable of reducing the amount of calculation for encoding without degrading encoding efficiency. It is.

  A motion vector detection processing apparatus according to the present invention is a motion vector detection processing apparatus that detects a motion vector necessary for motion compensation interframe prediction in a moving image encoding process, and detects a motion vector with coarse accuracy. Motion vector detection means; peripheral error value calculation means for calculating error values at points around the first precision motion vector detected by the first precision motion vector detection means; the first precision motion vector and the peripheral error A search range determining means for determining a search range for second precision motion detection specific to each surrounding point from error values at surrounding points calculated by the value calculating means, and a search range determined by the search range determining means Second precision motion vector detection means for performing motion vector detection with high accuracy.

  ADVANTAGE OF THE INVENTION According to this invention, the motion vector detection processing apparatus which can reduce the computational complexity for encoding, without degrading encoding efficiency can be provided.

The motion vector detection device according to the embodiment of the present invention is configured to use the encoded image based on the reduced motion vector detected based on the reduced encoded image obtained by reducing the encoded image and the reduced reference image obtained by reducing the reference image. A motion vector detecting device for detecting a motion vector of a coded block included in the reduced reference image, wherein the reduced motion vector of the reduced coded block included in the reduced encoded image is obtained based on pixel data included in the reduced reference image. A reduced motion vector detector to detect, a peripheral error value calculator to calculate error values at points around the reduced motion vector, and the most similar among the points at which the peripheral error value calculator has calculated similarity values A gradient detector that detects the direction of a point having a high degree when viewed from the point indicated by the reduced motion vector, and a detection by the reduced motion vector detector. The reduced motion vector, a search range setter for setting a search range for detecting the motion vector in the reference image from the direction detected by the gradient detector, and the search range setter set the search range setter A motion vector detector for detecting the motion vector in a search range.

  In the motion vector detection device according to the present embodiment, a search range for detecting a motion vector in a reference image based on the reduced motion vector detected by the reduced motion vector detector and the direction detected by the gradient detector. Is set, and a motion vector is detected in the set search range. For this reason, since motion vector detection is performed in the search range corresponding to the direction detected by the gradient detector, a range in which no motion vector exists can be excluded from the search range, and motion vector detection is performed in a narrow search range. be able to. As a result, it is possible to reduce the amount of calculation for detecting a motion vector without reducing the encoding efficiency.

FIG. 1 is a block diagram of a motion vector detection processing apparatus according to an embodiment of the present invention.
The reduced image generator 600 receives the encoded image 610 as an input, and outputs a reduced encoded image 611. The reduced image generator 600 receives the reference image 612 as an input, and outputs a reduced reference image 613. The reduced motion vector detector 601 receives the reduced encoded image 611 and the reduced reference image 613 as inputs, and outputs a reduced motion vector 614 for the processing target macroblock. The peripheral error value calculator 602 receives the reduced encoded image 611, the reduced reference image 613, and the reduced motion vector 614 as inputs, and positions 615 of points around the reduced motion vector in the reference image 612 and the surrounding points as the center. The error value 616 of each macroblock is output. The search range setter 603 receives the reduced motion vector 614, the position 615 of the surrounding points of the reduced motion vector, and the error value 616 of each macroblock centered on the surrounding points as inputs, and outputs the search range 617. The motion vector detector 604 receives the encoded image 610, the reference image 612, the reduced motion vector 614, and the search range 617 as inputs, and outputs a motion vector 618.

  Here, the search range 617 has a relatively center position that is biased in the direction in which a point most similar to the processing target macroblock exists among points around the position corresponding to the reduced motion vector in the reference image 612. A small area.

FIG. 2 is a flowchart of processing for detecting a motion vector of one macroblock.
In step 1, a reduced encoded image is generated by a reduced image generator. The reduced encoded image is generated by sampling the pixels of the encoded image every specified number. For example, when generating a reduced encoded image reduced to 1 / N in the vertical direction and 1 / M in the horizontal direction, the vertical direction is set to sampling points at every N pixels, and the horizontal direction is set to every M pixels. Create an image from an image consisting of The pixel value at the position in the horizontal direction X and the vertical direction Y of the encoded image is represented by P (X, Y), and the pixel value at the position in the horizontal direction x and the vertical direction y of the reduced encoded image is represented by Q (x, y). Then, Q (x, y) is given by equation (2).
Q (x, y) = P (M × x + m, N × y + n) (2)
However, m represents an arbitrary integer of 0 or more and less than M, and n represents an arbitrary integer of 0 or more and less than N.

The reduced reference image in step 2 is also generated in the same manner. That is, the pixel value at the position of the reference image in the horizontal direction X and the vertical direction Y is P ′ (X, Y), and the pixel value at the position of the reduced reference image in the horizontal direction x and vertical position y is Q ′ (x, y). Q ′ (x, y) is given by equation (3).
Q ′ (x, y) = P ′ (M × x + m, N × y + n) (3)
However, m represents an arbitrary integer of 0 or more and less than M, and n represents an arbitrary integer of 0 or more and less than N.

FIG. 3 shows a case where a reduced encoded image reduced to 1/2 in the vertical direction and 1/4 in the horizontal direction is generated. The points indicated by black squares are sampling points.
In step 3, a reduced motion vector is detected for each macroblock using the reduced encoded image and the reduced reference image generated in step 1. At each point in the reduced search range, the error value S is calculated using the equation (1), and the point (x, y) that gives the minimum error value S is represented by the reduced motion vector 91 from the macroblock to be processed. The tip. When the reduction ratio of the reduced encoded image and the reduced reference image is 1/2 vertical and 1/4 horizontal, the number of pixels in the vertical direction of the macroblock is 8 (= 16/2) and the number of pixels in the horizontal direction is 4 ( = 16/4), Σ in equation (1) is summed for i = 0, 1, 2, 3, j = 0, 1, 2,.

  In step 4, the reference image at the position where the eight pixel points (points A, B, C, D, E, F, G, H) around the reduced motion vector 91 obtained in step 3 are the leading edge of the reduced motion vector. And error values SAD_A, SAD_B, SAD_C, SAD_D, SAD_E, SAD_F, SAD_G, and SAD_H between these and the macroblock in the encoded image to be processed are calculated using the equation (1), and the coordinates of the above eight pixel points are calculated. And the error value are transmitted to the search range determiner. FIG. 4 shows the positional relationship between the reduced motion vector and the surrounding eight points. Assuming that the horizontal component of the reduced motion vector 91 is MVS_X and the vertical component is MVS_Y, the coordinates of the surrounding eight points are point A (MVS_X-1, MVS_Y), point B (MVS_X + 1, MVS_Y), point C (MVS_X, MVS_Y-1), point D (MVS_X, MVS_Y + 1), point E (MVS_X-1, MVS_Y-1), point F (MVS_X + 1, MVS_Y-1), point G (MVS_X-1, MVS_Y + 1), point H (MVS_X + 1, MVS_Y + 1). The error value SAD_A is an error value in the macro block centered on the point A. Similarly, SAD_B, SAD_C, SAD_D, SAD_E, SAD_F, SAD_G, and SAD_H are error values at points B, C, D, E, F, G, and H, respectively.

  In step 5, the reduced motion vector transmitted from the reduced motion vector detector in step 3 and the coordinates and 8 pixel points of the 8 pixel points transmitted from the peripheral error value calculator in step 4 are used as the tips of the reduced motion vectors. The search range is determined from the error value at the position and transmitted to the motion vector detector.

  FIG. 5 shows a flowchart of the process in step 5 of FIG. In step 11, an enlarged motion vector (MVS_X ′, MVS_Y ′) is obtained by multiplying the reduced motion vector (MVS_X, MVS_Y) by the vertical reduction rate N and the horizontal reduction rate M. Here, MVS_X ′ = MVS_X × M, MVS_Y ′ = MVS_Y × N. In step 12, it is determined which of SAD_A, SAD_B, SAD_C, SAD_D, SAD_E, SAD_F, SAD_G, and SAD_H has the minimum value. In step 13, the search range is determined from the position of the pixel point having the minimum error value obtained in step 12. When SAD_A is the minimum, the search range is determined as shown in FIG. In FIG. 6A, a hatched square point represents a point indicated by the expanded motion vector (MVS_X ′, MVS_Y ′), and a range 110 surrounded by a thick line represents a search range. FIG. 6A shows that the search range is from MVS_X′−4 to MVS_X ′ in the horizontal direction, and the search range is from MVS_Y′−2 to MVS_Y ′ + 2 in the vertical direction. Similarly, when SAD_B, SAD_C, SAD_D, SAD_E, SAD_F, SAD_G, and SAD_H are minimum, the search ranges are shown in FIGS. 6 (B), (C), (D), (E), (F), (G), respectively. ) And (H). The determined search range is transmitted to the motion vector detector. The search range is set in FIG. 6A when the minimum error value is SAD_A, set in FIG. 6B when the minimum error value is SAD_B, and so on, depending on the position of the pixel point that gives the minimum error value. The search range is determined in advance, and when the pixel point that gives the minimum error value is determined, the set search range may be set. For example, when the minimum error value is SAD-A, the neighborhood value on the left side is searched, and FIG. 6A is the search range.

  In step 6, the search range transmitted from the search range setter in step 5 and the reduced motion vector transmitted from the reduced motion vector detector are received, and at each pixel point of the designated search range, the expression ( The error value S is calculated using 1), and the point (x, y) that gives the minimum error value S is set as a motion vector. Since the number of pixels in the vertical direction of the macroblock is 16 and the number of pixels in the horizontal direction is 16, Σ in equation (1) is i = 0, 1, 2,..., 15, j = 0, 1, 2, ,..., 15 is summed.

The search range determination method described above calculates an error value of pixels around the tip of the reduced motion vector, and determines a search range biased in a direction with a small error value. Conventionally, for all motion vector candidates in the search area around the tip of the reduced motion vector, the error value between the reference image and the encoded image in the original resolution image is calculated, and the final motion vector is calculated. However, in this case, since the entire region at the tip of the reduced motion vector is searched, the amount of calculation required for the search increases. On the other hand, in the above embodiment, a direction with a smaller error value is detected at the tip of the reduced motion vector, and a smaller area biased in that direction is set as a search range. That is, matching is performed on an image with the original resolution only in a relatively small area in the direction in which a macroblock more similar to the block to be processed of the encoded image exists at the tip of the reduced motion vector. Therefore, the motion vector detection accuracy can be maintained, and the amount of calculation required for the search can be reduced. The size of this relatively small area should be determined by those skilled in the art implementing this embodiment. If the area is large, the motion vector detection accuracy increases, but the amount of computation increases.If the region is small, the motion vector detection accuracy decreases, but there is a trade-off relationship that the amount of computation increases. Set the size of the area according to the type of device to be used. For example, when a large battery such as a video camera can be mounted, the amount of calculation may be increased. Therefore, the size of the area is increased, and in the case of a camera mounted on a mobile phone, the battery To reduce consumption, set the area size to a smaller size.

Alternatively, as another embodiment, the following can be considered.
The method in which the reduced image generator generates a reduced encoded image in step 1 and the method in which the reduced image generator generates a reduced reference image in step 2 are not limited to sampling methods. As a method other than the sampling method, a method using a low-pass filter can be considered. If the pixel value at the position in the horizontal direction x and the vertical direction y of the reduced encoded image or the reduced reference image is represented by Q (x, y), Q (x, y) is given by Expression (4).
Q (x, y) = Σ (Ri × (ΣCj × P (N × x + i, M × y + j))) Equation (4)
In Equation (4), the outer Σ represents the sum for i, the inner Σ represents the sum for j, N represents the horizontal reduction ratio, and M represents the vertical reduction ratio. For example, when generating a reduced image of a horizontal image 1/4 and a vertical image 1/4 with a horizontal filter tap number 7 and a vertical filter tap number 7, N = 4 and M = 4, and the outside Σ is summed at i = −3, −2, −1, 0, 1, 2, 3 and Σ inside is summed at j = −3, −2, −1, 0, 1, 2, 3 I take the. Ri and Cj represent filter coefficients.

The reduced motion vector detection method in step 3 and the motion vector detection method in step 6 are not limited to the full search method. For example, when a reduced search range or a search range is specified by a range surrounded by a bold line 120 in FIG. 7, instead of evaluating error values at all points in the range, it is indicated by a hatched square. A method is conceivable in which an error value is evaluated only at the checker pattern-like point 121 and a reduced motion vector or a motion vector is determined from a point that gives the smallest error value. Alternatively, when a reduced search range or a search range is specified by a range surrounded by a bold line 140 in FIG. 8, instead of evaluating error values at all points in the range, it is indicated by a hatched square. A method is conceivable in which an error value is evaluated only at the grid-like point 141 and a reduced motion vector or a motion vector is determined from the point that gives the smallest error value.

The error value used in the reduced motion vector detection or the motion vector detection is not limited to the sum of absolute differences expressed by Equation (1). For example, instead of calculating the sum of absolute differences for all the pixel values included in the encoding target block, the sum of absolute differences may be calculated only at checker pattern-like points. In this case, the error value S is calculated using equation (1). When i = 0, 1, 2, 3, 4,..., 15, i is an even number, j = 0, 2, 4 , 6,..., 14, i is an odd number, j = 1, 3, 5, 7,. Alternatively, the sum of squared differences may be used instead of the sum of absolute differences. In this case, the error value is calculated by equation (5). S = Σ ((C (i, j) −R (x + i, y + j)) × (C (i, j) −R (x + i, y + j)))) Equation (5)
Alternatively, it is conceivable to use an error value such as Equation (6) that takes into account the magnitude of the motion vector component.
S = Σ | C (i, j) −R (x + i, y + j) | + λ × (Log2 (x) + Log2 (y)) (6)
The unit of the reduced motion vector detection in step 3 or the motion vector detection in step 6 is not limited to a macroblock, but is a block unit in which a plurality of macroblocks are combined or a block composed of a smaller number of pixels than the number of pixels included in one macroblock. This can be done in units.

  The method of obtaining pixel points around the reduced motion vector in step 4 is not limited to the eight points shown in FIG. For example, only four points A, B, C, and D in FIG. 4 may be used, and as shown in FIG. 9, points a, b, c, and d shifted by two pixels from the reduced motion vector in the reduced reference image. , E, f, g, and h.

  The method of determining the search range in step 5 is not limited to the method described above. For example, instead of determining the search range from the point giving the minimum error value, a method of determining the search range from the point giving the maximum error value can be considered. In this case, when the maximum value among SAD_A, SAD_B, SAD_C, SAD_E, SAD_F, SAD_G, and SAD_H is SAD_A, the search range is determined as FIG. 6B. Similarly, when the maximum value is SAD_B, SAD_C, SAD_D, SAD_E, SAD_F, SAD_G, SAD_H, the search ranges are shown in FIGS. 6 (A), (D), (C), (H), (G), (F), respectively. ), (E). Or, in step 13 of FIG. 5, when the minimum value among SAD_A, SAD_B, SAD_C, SAD_D, SAD_E, SAD_F, SAD_G, and SAD_H is larger than the threshold value SAD_TH, the search range is determined as FIG. If it is smaller than the threshold value SAD_TH, a method of determining from among FIGS. 6 (A), (B), (C), (D), (E), (F), (G), (H) can be considered. .

  The present embodiment may be provided in the form of a software program executed on the CPU, or may be provided in the form of hardware configured with a dedicated circuit.

In addition to the above embodiments, the following supplementary notes are disclosed.
(Appendix 1)
A motion vector detection processing device for detecting a motion vector necessary for motion compensation inter-frame prediction in moving image encoding processing,
First motion vector detection means for detecting a motion vector with a first accuracy;
A peripheral error value calculating means for calculating an error value at a point around the first motion vector detected by the first motion vector detecting means;
Search range determining means for determining a motion detection search range specific to each surrounding point from the error values at the surrounding points calculated by the first motion vector and the surrounding error value calculating means;
In the search range determined by the search range determination means, second motion vector detection means for performing motion vector detection with a second accuracy higher than the first accuracy;
A motion vector detection processing apparatus comprising:
(Appendix 2)
The first motion vector detection means performs first motion vector detection between a reduced encoded image obtained by reducing an image to be encoded and a reduced reference image obtained by reducing a reference image. Motion vector detection processing device.
(Appendix 3)
The motion vector detection processing apparatus according to appendix 1, wherein the first motion vector detection means performs error value calculation at a sampled search position and performs first motion vector detection.
(Appendix 4)
The motion vector detection processing apparatus according to appendix 1, wherein a center point of the search range determined by the search range determination unit is shifted from a point indicated by the first motion vector.
(Appendix 5)
The motion vector detection processing apparatus according to appendix 4, wherein the search range has a center point shifted in a direction in which the surrounding points are shifted from a point indicated by the first motion vector.
(Appendix 6)
A motion vector detection processing method for detecting a motion vector necessary for motion compensation interframe prediction in a video encoding process,
Detect motion vectors with first accuracy,
Calculating error values at points around the first motion vector detected by the first motion vector detection step;
From the first motion vector and the error value at the surrounding points calculated in the calculating step, a search range for motion detection specific to each surrounding point is determined,
In the search range determined in the search range determination step, motion vector detection is performed with a second accuracy higher than the first accuracy.
A motion vector detection processing method characterized by the above.
(Appendix 7)
A program for causing a computer to implement a motion vector detection processing method for detecting a motion vector necessary for motion compensation interframe prediction in a moving image encoding process,
Detect motion vectors with first accuracy,
Calculating error values at points around the first motion vector detected by the first motion vector detection step;
From the first motion vector and the error value at the surrounding points calculated in the calculating step, a search range for motion detection specific to each surrounding point is determined,
In the search range determined in the search range determination step, motion vector detection is performed with a second accuracy higher than the first accuracy.
A program characterized by causing processing to be performed.
(Appendix 8)
Hierarchical motion that performs motion vector detection with coarse accuracy and motion vector detection with fine accuracy around the motion vector detected with the coarse accuracy. A vector detection device,
A coarse-grain motion vector detector for detecting motion vectors with coarse accuracy;
A peripheral error value calculator for calculating an error value at a point around the coarse grain motion vector detected by the coarse grain motion vector detector;
A search range determiner for determining a search range for fine-grained motion detection from error values at surrounding points calculated by the coarse-grained motion vector and the peripheral error value calculator;
A motion vector detection device comprising a motion vector detector that performs motion vector detection with a fine accuracy within a search range determined by the search range determiner.

It is a block block diagram of the motion vector detection processing apparatus according to the embodiment of the present invention. It is a flowchart of the process which detects the motion vector of 1 macroblock. It is a figure which shows the case where the reduction | restoration encoding image reduced to vertical 1/2 and horizontal 1/4 is produced | generated. It is a figure which shows the positional relationship of a reduction | decrease motion vector and eight surrounding points. It is a flowchart of the process of step 6 of FIG. It is a figure which shows the example of a setting of a search range. It is a figure explaining the method of determining a reduced motion vector and a motion vector from the point which evaluates an error value only at a checkered pattern-like point and gives the minimum error value in the point. It is a figure explaining the method to determine a reduction | decrease motion vector and a motion vector from the point which gives the minimum error value in the error value only in the grid | lattice-like point shown with the oblique line. It is a figure which shows another example of the method of taking the pixel point around a reduction | decrease motion vector. It is a figure explaining the structure of an encoding process target image. It is a figure for demonstrating the structure of the reference image for the conventional motion vector detection method to detect a motion vector. It is a schematic diagram for demonstrating the reduction | restoration encoding image 30 in the motion vector detection apparatus using the conventional hierarchical search method. It is a schematic diagram for demonstrating the reduction | restoration reference image 40 in the motion vector detection apparatus using the conventional hierarchical search method. It is a schematic diagram for demonstrating the method of detecting the motion vector in the motion vector detection apparatus using the conventional hierarchical search method.

Explanation of symbols

600 Reduced image generator 601 Reduced motion vector detector 602 Peripheral error value calculator 603 Search range setter 604 Motion vector detector

Claims (5)

A motion vector detection processing device for detecting a motion vector necessary for motion compensation inter-frame prediction in moving image encoding processing,
First motion vector detection means for detecting a motion vector with coarse accuracy;
Peripheral error value calculating means for calculating error values at points around the first motion vector detected by the first motion vector detecting means;
Search range determining means for determining a motion detection search range specific to each surrounding point from the error value at the surrounding point calculated by the first motion vector and the surrounding error value calculating means;
Second motion vector detection means for performing motion vector detection with a second accuracy higher than the first accuracy in the search range determined by the search range determination means;
A motion vector detection processing apparatus comprising:   The motion vector detection processing apparatus according to claim 1, wherein a center point of a search range determined by the search range determination unit is shifted from a point indicated by the first motion vector.   The motion vector detection processing device according to claim 2, wherein a center point of the search range is shifted in a direction in which the surrounding points are shifted from a point indicated by the first motion vector. . A motion vector detection processing method for detecting a motion vector necessary for motion compensation interframe prediction in a video encoding process,
Detect motion vectors with coarse accuracy,
Calculating error values at points around the first motion vector detected by the first motion vector detection step;
From the first motion vector and the error value at the surrounding points calculated in the calculating step, a search range for motion detection specific to each surrounding point is determined,
In the search range determined in the search range determination step, motion vector detection is performed with a second accuracy higher than the first accuracy.
A motion vector detection processing method characterized by the above. A program for causing a computer to implement a motion vector detection processing method for detecting a motion vector necessary for motion compensation interframe prediction in a moving image encoding process,
Detect motion vectors with coarse accuracy,
Calculating error values at points around the first motion vector detected by the first motion vector detection step;
From the first motion vector and the error value at the surrounding points calculated in the calculating step, a search range for motion detection specific to each surrounding point is determined,
In the search range determined in the search range determination step, motion vector detection is performed with a second accuracy higher than the first accuracy.
A program characterized by causing processing to be performed.

Images