JP2007129370A - Motion vector detecting apparatus and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a motion vector detecting apparatus capable of detecting a motion vector with an accuracy of decimal without increasing the power consumption and upsizing the circuit scale. <P>SOLUTION: The motion vector detecting apparatus includes a decimal accuracy motion vector detection section 104 and a motion vector detection control section 105 that detect a motion vector by using only pixel values stored in a searching region image data memory 102 in a decimal position image block comprising decimal position pixels located around integer position pixels included in an image block particularized by a motion vector with an accuracy of integer to select only the image block at decimal positions comprising decimal position pixels whose pixel values can be calculated and an image block at integer positions particularized by the motion vector with an accuracy of integer for searching objects. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus for detecting a motion vector of an encoding target block included in an encoding target picture in an image compression encoding apparatus that compresses and records an image on a storage medium such as an optical disk, a magnetic disk, or a flash memory. Yes, especially H. The present invention relates to an apparatus for detecting a motion vector in an H.264 encoding system encoding apparatus.

  Due to advances in semiconductor technology and the like, coding technology for compressing moving images has been introduced into home video recording equipment. In recent years, in particular, as broadband communication infrastructure and storage media have increased in capacity, storage, distribution, and editing of HD (High Definition) images have been required. Codes for compressing moving images Technology is becoming increasingly important.

  One of the encoding techniques for compressing moving images is motion compensation using motion vectors between pictures. MPEG1 and MPEG2 introduce motion compensation with 1/2 pixel accuracy, and the pixel value at the decimal position is obtained from the pixel values at the surrounding integer positions using a bilinear filter (see, for example, Non-Patent Document 1). H. In H.264 / AVC, motion compensation with 1/4 pixel accuracy is introduced. A pixel value at the 1/2 position is obtained using a 6TAP filter, and a pixel value at the 1/4 position is obtained using a bilinear filter (for example, non-linear). Patent Document 2).

  In general, motion vector detection stores image data of coded pictures and reference pictures in an external memory, and transfers motion vector detection block image data and search area image data to the internal memory to detect motion vectors. To do. When the pixel value at the decimal position located near the boundary of the search area is obtained, the pixel value outside the search area is used. A specific example in which pixel values outside the search area are used when obtaining pixel values at decimal positions will be described with reference to FIG.

  FIG. 11 is a diagram illustrating a positional relationship between a pixel value at a decimal position that can be calculated and a pixel value at a decimal position that cannot be calculated based on the pixel value in the search area. The positions indicated by solid circles in FIG. 11 are integer positions in the search area, the positions indicated by “Δ” and “x” are decimal positions in the search area, and the positions indicated by dotted lines “◯” are An integer position outside the search area. The position indicated by “Δ” among the decimal positions indicates a position where the pixel value can be calculated from the pixel value at the integer position in the search area using the 6TAP filter. Of the decimal positions, the position indicated by “x” cannot be calculated from the pixel value at the integer position in the search area using the 6TAP filter, but is searched to calculate the pixel value using the 6TAP filter. The position where the pixel value outside the area needs to be acquired is shown.

For example, when calculating the pixel value of a (1,4), a (-4,4), a (-2,4), a (0,4), a (2,4), a (4, 4) The pixel value of a (6, 4) is required. Of these six points, a (−4,4) and a (−2,4) are pixel values outside the search area, and therefore the pixel values stored in the external memory are used.
“ITU-T Recommendation H.262” "Draft of Version4 of H.264 / AVC"

  However, when the pixel value outside the search area is used to obtain the pixel value at the decimal position in the search area, if the pixel value outside the area necessary for calculating the pixel value at the decimal position is acquired from the external memory, the external value Since data is transferred from the memory to the internal memory, there is a problem that the memory bandwidth increases and the power consumption increases.

  Further, in order to obtain the pixel value at the decimal position in the search area by using the pixel value outside the search area, the internal memory is increased to hold the pixel value necessary for calculating the pixel value at the decimal position. Since the storage capacity of the internal memory needs to be increased, there is a problem that the circuit scale increases.

  In particular, these problems are discussed in H.C. It becomes serious in motion compensation in H.264 / MPEG4 AVC. In other words, since the bilinear filter is used in motion compensation in MPEG1 and MPEG2, the pixel values outside the search area necessary for calculating the pixel value at the decimal position are only one pixel on the left and right and one on the upper and lower lines. . However, H.C. In the motion compensation in H.264 / MPEG4 AVC, since a 6TAP filter is used, the pixel values outside the search area necessary for calculating the pixel value at the decimal position are the left and right 3 pixels and the upper and lower 3 lines.

  Furthermore, these problems become serious in the image encoding processing of HD images. This is because the search area in the HD image encoding process needs to be wider than the search area in the SD (Standard Definition) image.

  The present invention has been made to solve such a problem, and provides a motion vector detection device capable of detecting a motion vector with decimal precision without increasing power consumption or increasing the circuit scale. With the goal.

  In order to achieve the above object, a motion vector detection device according to the present invention is a motion vector detection device that detects a motion vector of an encoding target block included in an image to be encoded, and is included in a search region. Image storage means for storing pixel values of pixels to be encoded, first detection means for detecting an integer precision motion vector of the encoding target block, and an image block specified by the integer precision motion vector From the decimal position pixel that can calculate the pixel value using only the pixel value stored in the image storage means, out of the decimal position image blocks composed of the decimal position pixels around the integer position pixel. Only the fractional position image block and the integer position image block specified by the integer precision motion vector are used as search candidates. And a second detecting means for detecting.

  As a result, when detecting a motion vector with decimal precision, only the pixel value stored in the internal memory is used, and therefore it is possible to detect the motion vector without increasing the power consumption or the circuit scale. Become.

  Preferably, the second detection unit further includes a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of a pixel outside the image area stored in the image storage means is necessary for calculating the pixel value of the pixel, the image block including the pixel at the decimal position is excluded from the search candidates.

  Further, the second detection means is a pixel of a decimal position pixel located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection means with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage unit is necessary for the calculation of the value, all the image blocks at the decimal positions are further extracted from the search candidates for the encoding target block. exclude.

  Further, the second detection means is a pixel of a decimal position pixel located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection means with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage means is necessary for the calculation of the value, all the image blocks at the decimal positions and the integer positions of the encoding target block are further calculated. The image block is excluded from the search candidates.

  As described above, the pixel value of the pixel outside the area stored in the storage device is included in the search candidate for detecting the motion vector. I can't. As a result, when detecting a motion vector with decimal precision, it is possible to detect a motion vector without increasing the power consumption or the circuit scale because only the pixel values stored in the storage device are used. Become.

  Preferably, the second detection unit further includes a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of a pixel outside the image area stored in the image storage unit is necessary for calculating the pixel value of the pixel, the pixel value of the pixel at the decimal position is stored in the image storage unit. Calculate from pixel values only.

  As described above, the pixel value of the pixel outside the image area stored in the storage device is interpolated with the pixel value of the pixel stored in the storage device as the pixel value of the pixel in the decimal position necessary for the interpolation calculation. Is calculated by As a result, using only the pixel values stored in the storage device, an image block including pixels at all decimal positions in the search area can be set as a search candidate. Therefore, since the search area for motion vector detection can be expanded without increasing the search area image data memory, it is possible to improve the encoding efficiency without increasing the power consumption or the circuit scale. It becomes.

  More preferably, the second detection unit further includes a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage means is necessary for the calculation of the pixel value of the pixel, the number of pixels used for calculating the pixel value of the pixel at the decimal position is The pixel value at the decimal position is calculated by reducing the number of pixels that can be acquired from only the pixel value stored in the storage means.

  In addition, the second detection unit further includes a pixel at a decimal position positioned around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of a pixel outside the image area stored in the image storage unit is necessary for the calculation of the pixel value of the image, the image storage unit stores the pixel value necessary for calculating the pixel value at the decimal position. The pixel value of the decimal position is calculated by substituting the pixel value of the pixel located outside the image area with the pixel value within the image area located closest to the pixel located outside the image area. .

  In addition, the second detection unit further includes a pixel at a decimal position positioned around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of a pixel outside the image area stored in the image storage unit is necessary for the calculation of the pixel value of the image, the image storage unit stores the pixel value necessary for calculating the pixel value at the decimal position. Instead of the pixel value of a pixel located outside the image area being displayed, the pixel value shown in a line-symmetric position across the boundary between the pixel located outside the area and the image area closest to the pixel is substituted. The pixel value at the decimal position is calculated.

  By these methods, it is possible to perform interpolation calculation of pixel values at decimal positions that originally require pixel values outside the search area, so that the search area for motion vector detection can be expanded without increasing the circuit scale or power consumption. Therefore, it is possible to improve the encoding efficiency.

  The present invention can be implemented not only as such a motion vector detection device, but also as a motion vector detection method using steps characteristic of the motion vector detection device. It can also be realized as a program for causing a computer to execute steps. Needless to say, such a program can be distributed via a recording medium such as a CD-ROM or a transmission medium such as the Internet.

  According to the present invention, it is possible to improve the encoding efficiency without increasing the circuit scale of the image encoding device or increasing the power consumption.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a moving picture encoding apparatus including a motion vector detection unit according to an embodiment of the present invention.

  The moving image encoding apparatus 100 is an apparatus that encodes an input moving image, and includes a picture image data memory 101, a motion vector detection unit 107, and an encoding unit 106 as shown in FIG.

  The picture image data memory 101 stores image data of a coding target picture and a reference picture. The image data of the encoding target picture is image data of a screen to be encoded. The image data of the reference picture is image data of a screen that is referred to when detecting a motion vector, and is image data of a locally decoded screen. For example, the encoding target picture and the reference picture are both screens composed of 1920 pixels × 1088 pixels. The image data is a pixel value of a pixel included in the screen.

  The encoding unit 106 encodes a moving image in units of blocks and outputs an encoded stream. The encoding unit 106 receives the input image, the image data of the local decoding screen, and the motion vector detected by the motion vector detection unit 107. Receiving these, the encoding unit 106 generates a prediction image in which motion compensation is performed in units of blocks from the image data of the local decoding screen and the motion vector detected by the motion vector detection unit 107, and the prediction image, the input image, The difference image data and motion vector are encoded, and an encoded stream is output.

  The motion vector detection unit 107 is a part that detects a motion vector of an encoding target block. The motion vector detection unit 107 includes a search area image data memory 102, an integer accuracy motion vector detection unit 103, a decimal accuracy motion vector detection unit 104, and a motion vector detection control unit 105.

  The search area image data memory 102 is an image storage unit that stores pixel values of pixels included in the search area, and the search area image data memory 102 according to the present embodiment is included in the encoding target block. The pixel value of the pixel is also stored. In the present embodiment, both the encoding target block and the image data of the search area are transferred from the picture image data memory 101.

  The encoding target block is a unit of an image to be subjected to motion compensation, and is a part of an image included in the encoding target picture. The encoding target block is composed of, for example, 16 pixels × 16 pixels.

  The search area is an area where search is performed in motion compensation, and is a part of an image included in the reference picture. The image data stored in the search area image data memory 102 includes image data of a search area that is a search target for detecting the motion vector of the encoding target block.

  Here, FIG. 3 is a diagram illustrating a first example of the relationship between the image stored in the search area image data memory 102 and the reference picture 301.

  A hatched portion in FIG. 3 indicates a search area 302 and is a part of the reference picture 301. A black square in FIG. 3 indicates the encoding target block 303 and is a part of the encoding target picture 304. The search area image data memory 102 according to the present embodiment stores pixel values of pixels included in the search area 302 and the encoding target block 303.

  The integer precision motion vector detection unit 103 searches the position of the image block that most closely approximates the encoding target block 303 from the plurality of image blocks included in the search region 302 according to the control of the motion vector detection control unit 105 with one pixel accuracy. Then, the motion vector of the encoding target block 303 is detected with one pixel accuracy. That is, the integer accuracy motion vector detection unit 103 is a first detection unit that detects an integer accuracy motion vector of the encoding target block 303.

  Here, among the plurality of image blocks included in the search region 302, the image block that is closest to the encoding target block 303 is, for example, the pixel value of the pixel included in the encoding block 303 and the image in the search region 302. This is an image block in which the sum of absolute differences from the pixel values of the pixels included in the block is minimized.

  The decimal precision motion vector detection unit 104 is an image block that most closely approximates the encoding target block from among the search candidate image blocks according to the result of the integer precision motion vector detection unit 103 and the control of the motion vector detection control unit 105. Is searched with decimal pixel accuracy, and a motion vector is detected with decimal pixel accuracy.

  That is, the decimal precision motion vector detection unit 104 is a second detection unit that detects a motion vector from search candidates.

  When the search candidate includes an image block at the decimal position composed of pixels at the decimal position, it is necessary to calculate the pixel value of the pixel at the decimal position. The pixel value of the pixel at the decimal position is calculated by the decimal precision motion vector detection unit 104 interpolating the pixel value of the pixel at the integer position.

  For example, the motion vector with decimal precision is a motion vector detected with half pixel precision, and the pixel value of the pixel at the decimal position necessary for detection of this motion vector is calculated using a 6TAP filter.

  The motion vector detection control unit 105 controls the integer accuracy motion vector detection unit 103 and the decimal accuracy motion vector detection unit 104.

  For example, the motion vector detection control unit 105 controls the integer precision motion vector detection unit 103 to search candidates when searching the search region 302 with integer accuracy.

  Further, the motion vector detection control unit 105 is configured with pixels at decimal positions where the pixel value can be calculated using only the pixel values stored in the search area image data memory 102 with respect to the decimal precision motion vector detection unit 104. Search candidates are controlled from only the image block at the decimal position and the image block at the integer position specified by the motion vector with integer precision.

  A first example of control performed by the motion vector detection control unit 105 on the decimal precision motion vector detection unit 104 is based on an integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block 303. The pixel value of the pixel outside the image area stored in the search area image data memory 102 is necessary for calculating the pixel value of the pixel at the decimal position located around the pixel included in the image block at the specified integer position. In this case, the image block including the pixel at the decimal position is excluded from the search candidates.

  A second example of control performed by the motion vector detection control unit 105 on the decimal precision motion vector detection unit 104 is an integer precision motion detected by the integer precision motion vector detection unit 103 for the encoding target block 303. The pixel value of the pixel outside the image area stored in the search area image data memory 102 is used to calculate the pixel value of the pixel at the decimal position located around the pixel included in the image block at the integer position specified by the vector. When necessary, the pixel value of the pixel at the decimal position is calculated from the pixel value stored in the search area image data memory 102.

  Note that the function of the motion vector detection control unit 105 may be provided by the integer accuracy motion vector detection unit 103 and the decimal accuracy motion vector detection unit 104, respectively.

  In this case, for example, the integer precision motion vector detection unit 103 searches the image block closest to the encoding target block from the image blocks at the integer positions composed of the pixels at the integer positions included in the search region, An integer-precision motion vector of the encoding target block is detected.

  Further, for example, the decimal precision motion vector detection unit 104 includes a decimal position image block composed of pixels at decimal positions located around the integer position pixels included in the image block specified by the integer precision motion vector. , Only the decimal position image block composed of pixels at the decimal position where the pixel value can be calculated using only the pixel value stored in the search area image data memory 102 and the integer position specified by the integer precision motion vector A motion vector is detected using the image block as a search candidate.

  By detecting the motion vector in this way, only the pixel value stored in the search area image data memory 102 can be used. Therefore, it is possible to detect a motion vector without increasing the power consumption or the circuit scale.

  FIG. 2 is a block diagram illustrating a configuration of the encoding unit 106 included in the moving image encoding apparatus 100. The encoding unit 106 outputs a stream of encoded image data using the motion vector detected by the input image, the local decoded image, and the motion vector detection unit 107.

  The encoding unit 106 includes an in-plane prediction unit 201, a motion compensation unit 202, a difference calculation unit 203, an orthogonal transformation unit 204, a quantization unit 205, an inverse quantization unit 206, an inverse orthogonal transformation unit 207, an addition unit 208, and a switch 209. , And an entropy encoding unit 210.

  Since the picture image data memory 101 and the motion vector detection unit 107 are parts common to FIGS. 1 and 2, the same reference numerals are given, and the description thereof is omitted here.

  The input image is an image before compression. The input image is input to the in-plane prediction unit 201, the picture image data memory 101, and the difference calculation unit 203.

  The motion compensation unit 202 generates a predicted image using the motion vector and the local decoded image. The motion compensation unit 202 acquires the motion vector detected by the motion vector detection unit 107, extracts an optimal image region for generating a predicted image from the local decode screen stored in the picture image data memory 101, and extracts the extracted image. A motion vector is applied to the region to generate a predicted image.

  The in-plane prediction unit 201 generates a predicted image using pixel values within the same screen. The in-plane prediction unit 201 generates a predicted image by performing in-plane prediction using the pixel values of the input image and the local decoded image.

  The switch 209 switches the data input source. The switch 209 switches the data input to the difference unit 203 and the adder 208 to either intra prediction or inter prediction.

  The difference calculation unit 203 calculates the difference between the two input images. The difference calculation unit 203 calculates a difference between the pixel value of the pixel included in the input image and the pixel value of the pixel included in the predicted image, and outputs the difference to the orthogonal transform unit 204.

  The orthogonal transform unit 204 transforms image data into frequency coefficients. The orthogonal transform unit 204 transforms the difference value input from the difference calculation unit 203 into a frequency coefficient and outputs the frequency coefficient to the quantization unit 205.

  The quantization unit 205 quantizes the frequency coefficient. The quantization unit 205 quantizes the frequency coefficient input from the direct transform unit 204 and outputs the quantized value to the entropy encoding unit 210.

  The inverse quantization unit 206 performs inverse quantization. The inverse quantization unit 206 restores the frequency coefficient by inverse quantization of the quantized value input from the quantization unit 205, and outputs the frequency coefficient to the inverse orthogonal transform unit 207.

  The inverse orthogonal transform unit 207 converts the frequency coefficient into a pixel value. The inverse orthogonal transform unit 207 converts the frequency coefficient input from the inverse quantization unit 206 into a pixel difference value by performing inverse frequency conversion, and outputs the pixel difference value to the addition unit 208.

  The adder 208 generates image data by adding the two input image data. The adder 208 adds the predicted image output from either the in-plane prediction unit 201 or the motion compensation unit 202 and the pixel difference value according to the switching of the switch 209, and generates a local decoded image.

  The entropy encoding unit 210 performs entropy encoding processing. The entropy encoding unit 210 performs entropy encoding on the quantized value, the motion vector, and the like, and outputs a stream obtained thereby.

  Next, processing executed in the moving image encoding apparatus 100 including the motion vector detection unit 107 configured as described above will be described with reference to FIG.

  FIG. 4 is a flowchart of processing executed in the video encoding apparatus 100 including the motion vector detection unit 107 according to the present embodiment.

  First, the picture image data memory 101 receives image data of an encoding target picture that is an input image, and stores the received image data (S401).

  Next, the picture image data memory 101 receives the image data of the reference picture locally decoded by the encoding unit 106 from the encoding unit 106 and stores it (S402).

  Next, the search area image data memory 102 stores the encoding target block 303 transferred from the picture image data memory 101 (S403).

  Next, the search area image data memory 102 stores the image data of the search area 302 set as a part of the reference picture 301 (S404).

  Next, the integer accuracy motion vector detection unit 103 and the decimal accuracy motion vector detection unit 104 detect a motion vector under the control of the motion vector detection control unit 105 (S405). Details of the motion vector detection process (S405) will be described below with reference to FIGS.

  First, in the processing shown in FIGS. 5 to 7, the motion vector detection control unit 105 is composed of pixels at decimal positions inside and around the image block at the integer position searched by the integer precision motion vector detection unit 103. If there is an image block that includes a pixel that requires a pixel value of a pixel outside the region stored in the search region image data memory 102 for calculation of the pixel value in a plurality of decimal position image blocks. Exclude blocks from search candidates. Hereinafter, processing performed by the motion vector detection unit 107 will be described with reference to FIGS.

  FIG. 5 is a flowchart showing a first example of the details of the motion vector detection processing S405 shown in FIG. In the processing shown in FIG. 5, the control performed by the motion vector detection control unit 105 is performed on the image block at the integer position specified by the integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block. When the pixel value of the pixel outside the image area stored in the search area image data memory 102 is necessary for calculation of the pixel value of the pixel at the decimal position located around the included pixel, the pixel at the decimal position is included. This is control for excluding image blocks from search candidates.

  The integer precision motion vector detection unit 103 determines the position of the image block that most closely approximates the encoding target block from the search region stored in the search region image data memory 102 according to the control of the motion vector detection control unit 105 by one pixel. A search is made with accuracy, and an integer-precision motion vector is detected from the position (S501).

  Next, the motion vector detection control unit 105 can search for an image block composed of the image block searched by the integer precision motion vector detection unit 103 and pixels at decimal positions located around the image block as search candidates. As a block, the processing from S503 to S504 is performed for each (S502).

  Next, the motion vector detection control unit 105 determines whether or not a pixel value that is not included in the search region is necessary for calculating the pixel value at the decimal position (S503). With this process, it is determined whether or not to exclude the image block including the pixel at the decimal position around the pixel at the integer position obtained from the integer precision motion vector detection unit 103 from the motion vector candidates.

  Next, when the motion vector control unit 105 determines that a pixel value that is not included in the search region is not necessary for calculation of the pixel value at the decimal position (No in S503), the decimal accuracy motion vector detection unit 104 receives this determination. Then, the image block at the decimal position including the pixel value at the decimal position is set as a candidate for motion vector search, and an evaluation value of the motion vector is calculated. (S504)

  If the motion vector control unit 105 determines that a pixel value that is not included in the search region is necessary for calculation of the pixel value at the decimal position (Yes in S503), the decimal accuracy motion vector detection unit 104 receives the determination. Then, the image block at the decimal position including the pixel value at the decimal position is excluded from the motion vector search candidates. Therefore, the motion vector evaluation value is not calculated for the image block at the decimal position.

  Here, the evaluation value is a numerical value representing how close the encoding target block and the image block in the search area are. For example, it is the sum of absolute values of differences between pixel values of pixels included in the encoding target block and pixel values of pixels included in the image block in the search area.

  Next, the decimal precision motion vector detection unit 104 outputs the motion vector having the minimum evaluation value as the detected motion vector. (S505)

  Hereinafter, a specific example of the control performed by the motion vector detection control unit 105 in the process shown in FIG. 5 will be described with reference to FIG. The interpolation method here is assumed to be a 6TAP filter.

  First, for example, in the integer precision motion vector detection process (S501), when a motion vector indicating an image block at an integer position including a (4,4) in the upper left corner is detected by the integer precision motion vector detection unit 103, a decimal number is calculated. Image blocks that can be candidates in the precision motion vector detection unit 104 are a (3,3), a (4,3), a (5,3), a (3,4), a (4,4), a These are nine image blocks each including nine points (5,4), a (3,5), a (4,5), and a (5,5) in the upper left corner. The motion vector detection control unit 105 repeats the processing of S503 and S504 for these image blocks that can be search candidates (S502).

  The motion vector detection control unit 105 determines whether or not a pixel value not included in the search region is necessary for calculation of the pixel value at the decimal position for all the pixels included in each image block (S503).

  Among these nine points, an image block that includes three points a (5,4), a (4,5), and a (5,5) in the upper left corner is a 6TAP filter based only on the pixel values in the search area. Can be used to calculate the pixel value at the decimal position (No in S503). Of the nine points above, five points a (3,3), a (4,3), a (5,3), a (3,4), a (3,5) The pixel value at the decimal position cannot be calculated from the pixel value alone using the 6TAP filter (Yes in S503). Therefore, the decimal precision motion vector detection unit 104 adds a (4,4) to a (5,4), a (4,5), a (5,5), and sets each of the four points to the upper left corner. Evaluation values are calculated for the four image blocks included (S504). The decimal precision motion vector detection unit 104 detects a motion vector from the position indicated by the block having the smallest evaluation value and the position of the encoding target block (S505).

  Next, for example, in the integer precision motion vector detection process (S501), when the motion vector indicating the image block at the integer position including a (2,2) in the upper left corner is detected by the integer precision motion vector detection unit 103, a The processing of S503 and S504 is repeated for the image block at the decimal position including the eight decimal positions around (2, 2) at the upper left corner (S502).

  The pixel values at the integer positions outside the search area are required to calculate the pixel values at the eight decimal positions around a (2,2) (Yes in S503). Therefore, as a result of the iterative process in S502, search candidates are only image blocks including a (2,2) in the upper left corner, and the decimal precision motion vector detection unit 104 detects a motion vector from the position of the block (S505). .

  FIG. 6 is a flowchart showing a second example of details of the motion vector detection process S405 shown in FIG. In the processing shown in FIG. 6, the control performed by the motion vector detection control unit 105 is a pixel included in an image block specified by an integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block. If the pixel value of the pixel outside the image area stored in the search area image data memory 102 is necessary for calculating the pixel value of the pixel at the decimal position located around the image block, the image block including the pixel at the decimal position Is excluded from the search candidates, and for the encoding target block, all the image blocks at the decimal position are excluded from the search candidates.

  That is, in the process shown in FIG. 6, in the above case, the motion vector detection control unit 105 controls the result of the integer precision motion vector detection unit 103 as the motion vector detected by the motion vector detection unit 107 according to the present invention. I do.

  The integer precision motion vector detection unit 103 detects the motion vector with integer precision (S601). This process is the same as S501 shown in FIG.

  Next, the motion vector detection control unit 105 performs the processing for all the pixels included in the eight decimal position image blocks configured from the decimal position pixels located inside and around the image block detected with integer precision. It is determined whether or not a pixel value that is not included in the search area is necessary for calculating the pixel value (S602).

  Next, when the motion vector control unit 105 determines that pixel values of integer positions that are not included in the search region are not necessary for calculation of pixel values for all of the pixels included in the above eight image blocks ( In step S602, the decimal precision motion vector detection unit 104 receives the determination, calculates an evaluation value of each image block, and detects a motion vector from the position of the image block having the minimum evaluation value (S603).

  Further, when the motion vector control unit 105 determines that a pixel value at an integer position that is not included in the search region is necessary for calculation of the pixel value even at one of the pixels included in the eight image blocks. In response to the determination, the decimal precision motion vector detection unit 104 outputs the motion vector detected by the integer precision motion vector detection unit 103 as the detected motion vector (S604).

  A specific example of control performed by the motion vector detection control unit 105 in the process shown in FIG. 6 will be described with reference to FIG. The interpolation method here is assumed to be a 6TAP filter.

  First, for example, in integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S601), a motion vector indicating an image block at an integer position including a (6,6) in the upper left corner is detected as an integer precision motion vector. A process when detected by the unit 103 will be described. The eight decimal pixels located around a (6,6) are a (5,5), a (6,5), a (7,5), a (5,6), a ( 7,6), a (5,7), a (6,7), a (7,7). The motion vector detection control unit 105 calculates pixel values of integer positions that are not included in the search area for calculating the pixel values of all the pixels included in the eight decimal positions of the image block including the eight points in the upper left corner. Is determined whether it is not necessary (S602).

  The pixel values of all the pixels included in the eight image blocks including these eight points in the upper left corner can be calculated from the pixel values only in the search region using the 6TAP filter. For this reason, the motion vector detection control unit 105 performs the processing for all the pixels included in the eight image blocks including the pixels at the decimal positions located around the image block at the integer position detected by the integer precision motion vector detection unit 103. Then, it is determined that pixel values at integer positions not included in the search area are not necessary for the calculation of pixel values (No in S602).

  The decimal precision motion vector detection unit 104 having received such a determination adds nine image blocks obtained by adding the image block at the integer position including a (6,6) in the upper left corner to the above-described eight decimal position image blocks. An evaluation value is calculated for, and a motion vector is detected from the position of the block having the smallest evaluation value and the position of the encoding target block (S603).

  Next, for example, in the integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S601), a motion vector indicating an image block at an integer position including a (4,4) in the upper left corner is an integer precision motion vector. Processing when detected by the detection unit 103 will be described. The eight decimal pixels located around a (4,4) are a (3,3), a (4,3), a (5,3), a (3,4), a ( 5,4), a (3,5), a (4,5), a (5,5). Of these 8 points, 5 points a (3,3), a (4,3), a (5,3), a (3,4), a (3,5) are within the search area. The pixel value at the decimal position cannot be calculated from the pixel value alone using the 6TAP filter.

  Therefore, the motion vector detection control unit 105 needs a pixel value at an integer position that is not included in the search region to calculate a pixel value at a decimal position located around the pixel at the integer position detected by the integer precision motion vector detection unit 103. (Yes in S602).

  The decimal precision motion vector detection unit 104 that has received this determination moves from the position indicated by the image block including the a (4,4) detected by the integer precision motion vector detection unit 103 in the upper left corner and the position of the encoding target block. A vector is detected (S604).

  FIG. 7 is a flowchart illustrating a third example of details of the motion vector detection process S405 illustrated in FIG. In the processing shown in FIG. 7, the control performed by the motion vector detection control unit 105 is a pixel included in an image block specified by an integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block. If the pixel value of the pixel outside the image area stored in the search area image data memory 102 is necessary for calculating the pixel value of the pixel at the decimal position located around the image block, the image block including the pixel at the decimal position Are excluded from the search candidates, and all the image blocks at the decimal positions are excluded from the search candidates for the encoding target block, and further, the image blocks at the integer positions are excluded from the search candidates for the encoding target block. Control.

  The integer precision motion vector detection unit 103 detects the motion vector with integer precision (S701). This process is the same as S501 shown in FIG.

  Next, the motion vector detection control unit 105 performs the processing for all the pixels included in the eight decimal position image blocks configured from the decimal position pixels located inside and around the image block detected with integer precision. It is determined whether or not a pixel value that is not included in the search area is necessary for calculating the pixel value (S702).

  Next, when the motion vector control unit 105 determines that pixel values of integer positions that are not included in the search region are not necessary for calculation of pixel values for all of the pixels included in the above eight image blocks ( In response to this determination, the decimal precision motion vector detection unit 104 calculates an evaluation value of each image block and detects a motion vector from the position of the image block having the minimum evaluation value (S703).

  Further, when the motion vector control unit 105 determines that a pixel value at an integer position that is not included in the search region is necessary for calculation of the pixel value even at one of the pixels included in the eight image blocks. In response to the determination, the decimal precision motion vector detection unit 104 does not detect a motion vector (S704).

  A specific example of control performed by the motion vector detection control unit 105 in the process shown in FIG. 7 will be described with reference to FIG. The interpolation method here is assumed to be a 6TAP filter.

  First, for example, in integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S701), a motion vector indicating an image block at an integer position including a (6,6) in the upper left corner is detected as an integer precision motion vector. The processing in the case of being detected by the unit 103 (after No in S702, S703), in the example in which the motion vector detection control unit 105 performs the control shown in FIG. 6, an integer position including a (6,6) in the upper left corner This is the same as the processing when the motion vector indicating the image block is detected by the integer precision motion vector detection unit 103 (No in S602, then S603).

  Next, for example, in integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S701), a motion vector indicating an image block at an integer position including a (4,4) in the upper left corner is an integer precision motion vector. Processing when detected by the detection unit 103 will be described. The eight decimal pixels located around a (4,4) are a (3,3), a (4,3), a (5,3), a (3,4), a ( 5,4), a (3,5), a (4,5), a (5,5). Of these 8 points, 5 points a (3,3), a (4,3), a (5,3), a (3,4), a (3,5) are within the search area. The pixel value at the decimal position cannot be calculated from the pixel value alone using the 6TAP filter.

  Therefore, the motion vector detection control unit 105 performs pixel values for all the decimal position pixels included in the eight decimal position image blocks located around the integer position pixels detected by the integer precision motion vector detection unit 103. It is determined that a pixel value at an integer position that is not included in the search region is necessary for the calculation of (Yes in S702).

  The motion vector detection control unit 105 that has made such a determination excludes all image blocks from the search candidates for the encoding target block. Therefore, the detection of the motion vector by the decimal precision motion vector detection unit 104 is prohibited (S704).

  As described above, the motion vector detection unit 107 executes the processing shown in FIGS. 5 to 7 so that the decimal precision motion vector detection unit 104 can search for a decimal precision motion vector unless the search candidates are limited. . In addition, the motion vector is detected by limiting the search candidates by using only the pixel values included in the image area stored in the search area image data memory 102. Therefore, it is possible to improve the image quality and coding efficiency of the compressed image while suppressing an increase in power consumption and an increase in circuit scale.

  Next, in the processing shown in FIGS. 8 to 10, the motion vector detection control unit 105 is composed of pixels at the decimal positions inside and around the image block at the integer position detected by the integer precision motion vector detection unit 103. If there is an image block that includes pixels that require pixel values of pixels outside the area stored in the search area image data memory 102 for calculation of pixel values, among the plurality of decimal position image blocks Control for adjusting the number of integer pixels used for the control, and control for substituting the pixel value stored in the search area image data memory 102 for the insufficient pixel value.

  Under such control of the motion vector detection control unit 105, the decimal precision motion vector detection unit 104 calculates the pixel values of all the decimal positions from the pixel values stored in the search area image data memory 102. Hereinafter, the processing performed by the motion vector detection unit 107 will be described with reference to FIGS.

  FIG. 8 is a flowchart showing a fourth example of details of the motion vector detection processing S405 shown in FIG. In the processing shown in FIG. 8, the control performed by the motion vector detection control unit 105 is a pixel included in an image block specified by an integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block. If the pixel value of the pixel outside the image area stored in the search area image data memory 102 is necessary for calculating the pixel value of the pixel at the decimal position located around the pixel area, the pixel value of the pixel at the decimal position is calculated. This is control for calculating the pixel value at the decimal position by reducing the number of pixels used for the calculation to the number of pixels that can be acquired from the pixel values stored in the search area image data memory 102.

  The integer precision motion vector detection unit 103 detects the motion vector with integer precision (S801). This process is the same as S501 shown in FIG.

  Next, the motion vector detection control unit 105 calculates a pixel for each pixel included in the eight decimal positions of the image block including the pixels within and around the image block detected with integer precision. It is determined whether or not a pixel value that is not included in the search area is necessary for calculating the value (S802).

  Next, when the motion vector control unit 105 determines that the pixel value of the pixel at the integer position not included in the search region is not necessary for the calculation of the pixel value for the pixel included in the above-described eight decimal position image blocks. (No in S802), the decimal precision motion vector detection unit 104 receives the determination and calculates the pixel value of the pixel using a predetermined number of TAPs (S803).

  If the motion vector control unit 105 determines that the pixel value of the pixel at the integer position not included in the search region is necessary for the calculation of the pixel value for the pixel included in the eight image blocks (S802). Yes) For the pixel, the pixel value is calculated by making the number of TAPs smaller than the predetermined number of TAPs.

  Further, the decimal precision motion vector detection unit 104 calculates an evaluation value of each motion vector using the calculated pixel value of the decimal position, and outputs a motion vector having the smallest evaluation value as a detected motion vector ( S804).

  A specific example of control performed by the motion vector detection control unit 105 in the motion vector detection process S5 illustrated in FIG. 8 will be described with reference to FIG. In this interpolation method, a 6TAP filter is used. When the 6TAP filter cannot be used, a 4TAP filter or a 2TAP filter is used.

  First, in integer precision motion vector detection processing executed by, for example, the integer precision motion vector detection unit 103 (S801), a motion vector indicating an image block at an integer position including a (6,6) in the upper left corner is detected as an integer precision motion vector. A process when detected by the unit 103 will be described. The eight decimal pixels located around a (6,6) are a (5,5), a (6,5), a (7,5), a (5,6), a ( 7,6), a (5,7), a (6,7), a (7,7). The motion vector detection control unit 105 uses the eight decimal positions including these eight points in the upper left corner as search candidates, and for each pixel included in each search candidate, the pixel value is included in the search area for calculation. It is determined whether or not a pixel value at a non-integer integer position is necessary (S802).

  With respect to each pixel included in the eight decimal position image blocks including the eight points in the upper left corner, the pixel value of the decimal position can be calculated from only the pixel value in the search region using the 6TAP filter. Therefore, the motion vector detection control unit 105 determines that a pixel value at an integer position that is not included in the search region is not necessary for the calculation of the pixel value for each of these pixels (No in S802). The decimal precision motion vector detection unit 104 that has received such determination calculates the pixel value of each pixel using a 6TAP filter, and adds a (6,6) to the image block at the decimal position including the above eight points in the upper left corner. The evaluation values of the nine image blocks including the image block at the integer position included in the upper left corner are calculated, and a motion vector is detected from the position of the image block having the smallest evaluation value and the position of the encoding target block (S804). ).

  Next, for example, in integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S801), a motion vector indicating an image block at an integer position including a (4,4) in the upper left corner is an integer precision motion vector. Processing when detected by the detection unit 103 will be described. The eight decimal pixels located around a (4,4) are a (3,3), a (4,3), a (5,3), a (3,4), a ( 5,4), a (3,5), a (4,5), a (5,5). Of these 8 points, 5 points a (3,3), a (4,3), a (5,3), a (3,4), a (3,5) are within the search area. The pixel value at the decimal position cannot be calculated from the pixel value alone using the 6TAP filter. Therefore, the motion vector detection control unit 105 determines that a pixel value at an integer position that is not included in the search area is necessary for calculating the pixel value (Yes in S802). The motion vector detection control unit 105 that has made such a determination adjusts the number of TAPs of the filter (S803).

A specific method for adjusting the number of TAPs in each pixel will be described.
The 5 points a (3,3), a (4,3), a (5,3), a (3,4), a (3,5) cannot be interpolated by the 6TAP filter. For this reason, the motion vector detection control unit 105 uses a (3,3), a (4,3), a (5,3), a (3,4) as 4TAP filters that enable interpolation of these five points. , A (5,4), a (3,5), a (4,5), a (5,5), the decimal precision motion vector detection unit 104 is controlled so as to be used at all eight points.

  For example, when obtaining the pixel value of the pixel located at a (1, 2), the motion vector detection control unit 105 controls the decimal precision motion vector detection unit 104 to use a 2TAP filter. The pixel values of pixels at integer positions that can be used to calculate a (1,2) from the pixel values stored in the search area image data memory 102 are limited to a (0,2) and a (2,2). This is because neither a 6TAP filter nor a 4TAP filter can be used. As described above, the number of TAPs of the filter is determined by how many pixel values of pixels at integer positions that can be used as pixel values of pixels at the decimal positions to be calculated are stored in the search area image data memory 102. .

  FIG. 9 is a flowchart illustrating a fifth example of details of the motion vector detection process S405 illustrated in FIG. In the process shown in FIG. 9, the control performed by the motion vector detection control unit 105 is a pixel included in an image block specified by an integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block. If the pixel value of the pixel outside the image area stored in the search area image data memory 102 is required for calculating the pixel value of the pixel at the decimal position located around the pixel, the pixel value at the decimal position is calculated. For the pixel value of the pixel located outside the area stored in the search area image data memory 102 necessary for the purpose, the pixel value in the image area located closest to the pixel located outside the image area is substituted. In this way, the pixel value at the decimal position is calculated.

  The integer precision motion vector detection unit 103 detects the motion vector with integer precision (S901). This process is the same as S501 shown in FIG.

  Next, the determination process performed by the motion vector detection control unit 105 in step S902 is the same as that in step S802 illustrated in FIG.

  Next, for the pixels included in the image blocks at the eight decimal positions positioned around the image block detected with integer precision, the motion vector control unit 105 does not include the search region in the pixel value calculation. If it is determined that the pixel value of this pixel is not necessary (No in S902), the decimal precision motion vector detection unit 104 receives this determination and calculates the pixel value of that pixel (S904).

  If the motion vector control unit 105 determines that the pixel value of the pixel at the integer position not included in the search region is necessary for the calculation of the pixel value for the pixel included in the eight image blocks (S902). Yes), the decimal precision motion vector detection unit 104 adds the pixel value at the integer position stored in the search area image data memory 102 to the pixel value at the integer position outside the search area necessary for calculating the pixel value at the decimal position. The pixel value is calculated by substituting.

  Here, the pixel value to be substituted is an integer pixel value located at the boundary in the search area closest to the pixel at the integer position outside the area necessary for calculating the pixel value at the decimal position.

  Further, the decimal precision motion vector detection unit 104 calculates an evaluation value of the motion vector using the calculated pixel value at the decimal position, and outputs the motion vector having the minimum evaluation value as the detected motion vector (S904). ).

  A specific example of control performed by the motion vector detection control unit 105 in the motion vector detection process S505 illustrated in FIG. 9 will be described with reference to FIG. The interpolation method here is assumed to be a 6TAP filter.

  First, for example, in integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S901), a motion vector indicating an image block at an integer position including a (6,6) in the upper left corner is detected as an integer precision motion vector. The processing in the case of being detected by the unit 103 (after No in S902, S904), in S801 shown in FIG. 8, the motion vector indicating the image block at the integer position including a (6,6) in the upper left corner is an integer precision motion. This is the same as the processing when it is detected by the vector detection unit 103 (No in S802, then S804).

  Next, for example, in integer precision motion vector detection processing executed by the integer precision motion vector detection unit 103 (S901), a motion vector indicating an image block at an integer position including a (4,4) in the upper left corner is an integer precision motion vector. Processing when detected by the detection unit 103 will be described.

  The eight decimal pixels located around a (4,4) are a (3,3), a (4,3), a (5,3), a (3,4), a ( 5,4), a (3,5), a (4,5), a (5,5). Of these 8 points, 5 points a (3,3), a (4,3), a (5,3), a (3,4), a (3,5) are within the search area. With only the pixel value, the pixel value at the decimal position cannot be calculated using the 6TAP filter. Therefore, the motion vector detection control unit 105 determines that a pixel value at an integer position that is not included in the search region is necessary for calculation of the pixel value for each of these five pixels (Yes in S902).

  The motion vector detection control unit 105 that has made such a determination substitutes a pixel value located at the boundary of the search region closest to the pixel as a pixel value outside the required search region (S903). The pixel in the region closest to the pixel outside the required search region is located between the intersection of the vertical line drawn from the pixel outside the required search region to the boundary of the search region and the boundary line of the search region, A pixel in a region.

  For example, to calculate the pixel value of a (3,4), a (-2,4), a (0,4), a (2,4), a (4,4), a (6, 4) The pixel value of a (8,4) is required. However, a (−2, 4) is a pixel outside the search area. Therefore, the motion vector detection control unit 105 substitutes the pixel value of a (0,4), which is the pixel in the search region closest to a (-2,4), as the pixel value of a (-2,4). (S903). The same applies to a (3,3), a (4,3), a (5,3), and a (3,5).

  Further, calculation of the pixel value of a (1, 2) will be described as an example of the boundary value substitution process (S903) executed by the motion vector detection control unit 105. To calculate the pixel value of a (1,2), a (-4,2), a (-2,2), a (0,2), a (2,2), a (4,2 ), A (6,2) pixel values are required. However, a (−2, 4) and a (−2, 2) are pixels outside the search area. Therefore, the motion vector detection control unit 105 searches for the pixel values of a (-2,4) and a (-2,2) that are closest to a (-2,4) and a (-2,2), respectively. The pixel value of a (0,2), which is a pixel in the area, is substituted (S903).

  FIG. 10 is a flowchart illustrating a sixth example of the details of the motion vector detection processing S405 illustrated in FIG. In the process shown in FIG. 10, the control performed by the motion vector detection control unit 105 is a pixel included in an image block specified by an integer precision motion vector detected by the integer precision motion vector detection unit 103 for the encoding target block. If the pixel value of the pixel outside the image area stored in the search area image data memory 102 is required for calculating the pixel value of the pixel at the decimal position located around the pixel, the pixel value at the decimal position is calculated. The pixel value of the pixel located outside the image area stored in the search area image data memory 102 necessary for the line between the pixel located outside the image area and the area closest to the pixel This is a control for substituting the pixel value shown at the symmetrical position and calculating the pixel value at the decimal position.

  The integer precision motion vector detection unit 103 detects the motion vector with integer precision (S1001). This process is the same as S501 shown in FIG.

  Next, the determination process performed by the motion vector detection control unit 105 in S1002 is the same as that in S802 shown in FIG.

  Next, an integer that is not included in the search area in the calculation of the pixel value of the motion vector detection control unit 105 for the pixels included in the image blocks at the eight decimal positions positioned around the image block detected with integer precision. When it is determined that the pixel value of the pixel at the position is not necessary (No in S1002), the decimal precision motion vector detection unit 104 receives the determination and calculates the pixel value of the pixel (S1004). This is the same as the processing performed in S904 shown.

  In addition, when the motion vector control unit 105 determines that the pixel values of the pixels at the integer positions not included in the search region are necessary for the calculation of the pixel values for the pixels included in the above eight image blocks (S1002). Yes), the decimal precision motion vector detection unit 104 calculates a pixel value using a substitute value and detects a motion vector (S1004), which is the same as the processing performed in S904 shown in FIG.

  The difference between the process in S1004 in this figure and the process in S904 in FIG. 9 is the position of the pixel to be substituted.

  That is, the pixel value that the motion vector detection control unit 105 substitutes for the pixel value not stored in the search area image data memory 102 in S904 of FIG. 9 is an integer position outside the area that is necessary for calculating the pixel value at the decimal position. This pixel is located at the boundary of the area stored in the search image area data memory 102 that is located closest to the other pixel.

  On the other hand, the pixel value that the motion vector detection control unit 105 substitutes for the pixel value that is not stored in the search area image data memory 102 in S1004 in this figure is out of the area necessary for calculating the pixel value at the decimal position. The pixel value of the pixel at the integer position and the pixel at the symmetrical position across the boundary of the area stored in the search image area data memory 102.

  A specific example of control performed by the motion vector detection control unit 105 in the motion vector detection process S5 illustrated in FIG. 10 will be described with reference to FIG. The interpolation method here is assumed to be a 6TAP filter.

  Up to now, as described in the explanation of this figure, the difference between the process shown in this figure and the process shown in FIG. 9 is that the pixel value of the pixel necessary for calculating the pixel value of the pixel at the decimal position by the 6TAP filter. Is not stored in the search area image data memory 102, the pixel value at which position in the area stored in the search area image data memory 102 is to be substituted.

  That is, the processes in S1001, S1002, and S1004 are the same as the processes in S901, S902, and S904, respectively. Therefore, details of these processes are omitted.

  Hereinafter, a specific description will be given of which pixel value is to be substituted, taking as an example the case where the pixel values of a (3,4) and a (1,2) are calculated.

  First, a case where the pixel value of a (3,4) is calculated will be described. To calculate the pixel value of a (3,4), a (-2,4), a (0,4), a (2,4), a (4,4), a (6,4) , A (8,4) pixel value is required. However, a (−2, 4) is a pixel outside the search area. Therefore, the motion vector detection control unit 105 uses a (0,4) which is a pixel at a symmetrical position with respect to a (-2,4) as a pixel value of a (-2,4) across the boundary in the search area. ) Is substituted (S1003).

  Next, a case where the pixel value of a (1,2) is calculated will be described. To calculate the pixel value of a (1,2), a (-4,2), a (-2,2), a (0,2), a (2,2), a (4,2 ), A (6,2) pixel values are required. However, a (-4,2) and a (-2,2) are pixels outside the search area. Therefore, the motion vector detection control unit 105 uses a (0,2) which is a pixel at a symmetrical position with respect to a (-2,2) as a pixel value of a (-2,2) across the boundary in the search area. ) Is substituted (S1003).

  In addition, as the pixel value of a (-4,2), the pixel value of a (2,2), which is a pixel located symmetrically across the boundary in the search region with a (-4,2), is substituted. (S1003).

  As described above, the motion vector detection unit 107 executes the processing shown in FIGS. 8 to 10, so that the decimal positions in and around the image block at the integer position detected by the integer precision motion vector detection unit 103 are detected. Among the plurality of decimal position image blocks made up of pixels, there is an image block that includes pixels that require pixel values of pixels outside the region stored in the search region image data memory 102 for calculation of pixel values. Even in such a case, it is possible to detect a motion vector with decimal precision. Therefore, as compared with the method described with reference to FIGS. 5 to 7, the search area for motion vector detection can be expanded without increasing the search area image data memory 102. Therefore, it is possible to further improve the image quality and coding efficiency of the compressed image while suppressing an increase in power consumption and an increase in circuit scale.

  The motion vector detection unit 107 according to the embodiment of the present invention has been described above, but the present invention is not limited to this embodiment.

  For example, a modification of each filter can be considered for use in interpolation of pixel values at decimal positions. First, the filter used in the embodiment is the same as that of the motion vector detection unit of the present invention. When used in an H.264 encoding device, a 6TAP filter described in the standard may be used for motion vector detection, or a filter such as a 4TAP filter or a 2TAP filter may be used. Further, the decimal precision motion vector detection unit 104 may use different methods for interpolation calculation of pixels at decimal positions in the vertical direction and the horizontal direction.

  In the example illustrated in FIG. 8, when calculating the pixel value of the pixel at the decimal position that cannot be calculated by the 6TAP filter, the motion vector detection control unit 105 can use the search area image data memory 102 that can be used to calculate the pixel value of the pixel. The number of TAPs is reduced to 4 TAP and 2 TAP according to the number of pixel values stored. However, if the calculation cannot be performed using the 6TAP filter, the 2TAP filter may always be applied.

  Furthermore, when using a 4TAP filter and a 6TAP filter, the coefficients related to the pixel value at each integer position include, for example, -1/8, 5/8, 5/8, and -1/8 in the 4TAP filter. . In the 6TAP filter of the H.264 encoding standard, there are 1/32, −5/32, 20/32, 20/32, −5/32, and 1/32. However, the value of the coefficient used in each of these filters may be another value.

  Further, for example, the search area image data memory 102 may store all the lines of the reference picture including the search area as shown in FIG.

  FIG. 12 is a diagram illustrating a second example of the relationship between the region stored in the search region image data memory 102 and the reference picture. The encoding target block 303 and the search area 302 are the same as those in FIG. The search area image data memory 102 stores only the search area of the reference picture in FIG. 3, whereas FIG. 12 stores all the lines 305 including the search area in the reference picture. Different.

  In this case, the pixel value of the pixel at the decimal position may be calculated using not only the pixel in the search area 302 but also the pixel value of the pixel in the line 305 including the search area.

  Further, for example, in the present embodiment, only a generation method with 1/2 pixel accuracy is described, but a motion vector detection control method with arbitrary 1 / M pixel accuracy can be performed in the same manner, and decimal pixels The motion vector detection with accuracy may be performed with arbitrary 1 / M pixel accuracy. Here, “M” represents a power of 2.

  Further, for example, the search procedure executed by the decimal-precision motion vector detection unit 104 when detecting a motion vector with 1 / M pixel accuracy performs a motion vector search with 1/2 pixel accuracy, and a motion vector with 1/4 pixel accuracy. A motion vector with 1 / M pixel accuracy may be detected step by step, such as detection. Alternatively, a 1 / M pixel precision motion vector may be detected by searching all positions with 1 / M pixel precision existing around the position indicated by the input one pixel precision motion vector.

  Each functional block in the block diagrams shown in FIGS. 1 and 2 is typically realized as an LSI which is an integrated circuit. This LSI may be made into one chip or a plurality of chips. As an example of a plurality of chips, it is conceivable that each functional block other than the memory is formed into one chip. Note that an LSI may be referred to as an IC, a system LSI, a super LSI, or an ultra LSI depending on the degree of integration.

  Further, the method of circuit integration of the moving picture coding apparatus 100 including the motion vector detection unit 107 according to the present invention is not limited to LSI. The integration method may be realized by a dedicated circuit or a general-purpose processor. An FPGA (Field Programmable Gate Array) that can be programmed after the manufacture of the LSI or a reconfigurable processor that can reconfigure the connection and setting of the circuit cells inside the LSI may be used.

  Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology. Biotechnology can be applied.

  Further, among the functional blocks, only the unit for storing data may be configured separately without being integrated into one chip.

  According to the motion vector detecting apparatus of the present invention, the H.264 and H.264 circuits are suppressed while suppressing the increase in power consumption and the circuit scale. H.264 video can be encoded, and can be applied to personal computers, HDD recorders, DVD recorders, video cameras, mobile phones with cameras, and the like. Also, the present invention can be applied to an encoding device provided with this motion vector detection device.

It is a block diagram which shows the structure of the moving image encoding apparatus provided with the motion vector detection part which concerns on embodiment of this invention. It is a block diagram which shows the structure of the encoding part contained in a moving image encoder. It is a figure which shows the 1st example of the relationship between the image which the search area | region image data memory has memorize | stored, and a reference picture. It is a flowchart of the process performed in a moving image encoder provided with the motion vector detection apparatus which concerns on this Embodiment. 5 is a flowchart showing a first example of details of motion vector detection processing S405 shown in FIG. 4. 6 is a flowchart showing a second example of details of motion vector detection processing S405 shown in FIG. 4. 6 is a flowchart showing a third example of details of motion vector detection processing S405 shown in FIG. 4. It is a flowchart which shows the 4th example of the detail of the motion vector detection process S405 shown in FIG. It is a flowchart which shows the 5th example of the detail of motion vector detection process S405 shown in FIG. It is a flowchart which shows the 6th example of the detail of motion vector detection process S405 shown in FIG. It is a figure which shows the positional relationship of the pixel value of the decimal position which can be calculated with the pixel value in a search area | region, and the pixel value of the decimal position which cannot be calculated. It is a figure which shows the 2nd example of the relationship between the area | region which the search area | region image data memory has memorize | stored, and a reference picture.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Moving image encoder 101 Picture image data memory 102 Search area image data memory 103 Integer precision motion vector detection 104 Decimal precision motion vector detection 105 Motion vector detection control part 106 Encoding part 107 Motion vector detection part 201 In-plane prediction part 202 Motion compensator 203 Subtractor 204 Orthogonal transformer 205 Quantizer 206 Inverse quantizer 207 Inverse orthogonal transformer 208 Adder 209 Switch 210 Entropy encoder

Claims (10)

A motion vector detection device that detects a motion vector of an encoding target block included in an image to be encoded,
Image storage means for storing pixel values of pixels included in the search area;
First detection means for detecting an integer-precision motion vector of the encoding target block;
Pixels stored in the image storage unit among the decimal position image blocks composed of pixels at the decimal positions located around the integer position pixels included in the image block specified by the integer precision motion vector A motion vector is detected by using only an image block at a decimal position composed of pixels at a decimal position at which a pixel value can be calculated using only the value and an image block at an integer position specified by the integer precision motion vector as search candidates. A motion vector detecting device. The second detection unit further includes a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When a pixel value of a pixel outside the image area stored in the image storage unit is necessary for calculation of the value, an image block including the pixel at the decimal position is excluded from the search candidates. Item 2. The motion vector detection device according to Item 1. The second detection means is a pixel value of a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection means with respect to an encoding target block. When the pixel values of the pixels outside the image area stored in the image storage means are required for the calculation, all the image blocks at the decimal positions are further excluded from the search candidates for the encoding target block. The motion vector detection device according to claim 1 or 2, wherein The second detection means is a pixel value of a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection means with respect to an encoding target block. When the pixel values of the pixels outside the image area stored in the image storage means are required for the calculation, all the image blocks at the decimal positions and the image blocks at the integer positions are further included in the encoding target block. The motion vector detection device according to claim 1, wherein the motion vector detection device is excluded from the search candidates. The second detection unit further includes a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage unit is necessary for the calculation of the value, the pixel value of the pixel at the decimal position is only the pixel value stored in the image storage unit. The motion vector detection device according to claim 1, wherein the motion vector detection device is calculated from: The second detection unit further includes a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage unit is necessary for the calculation of the value, the image storage unit stores the number of pixels used for calculating the pixel value of the pixel at the decimal position. The motion vector detection device according to claim 5, wherein the pixel value at the decimal position is calculated by reducing the obtained pixel value to a number of pixels that can be acquired. The second detection unit further includes a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage unit is necessary for the calculation of the value, the image storage unit stores the pixel value necessary for calculating the pixel value in the decimal position. The pixel value of the decimal position is calculated by substituting the pixel value in the image area located closest to the pixel located outside the image area for the pixel value of the pixel located outside the existing area. The motion vector detection device according to claim 5, wherein The second detection unit further includes a pixel at a decimal position located around a pixel included in an image block specified by an integer-precision motion vector detected by the first detection unit with respect to the encoding target block. When the pixel value of the pixel outside the image area stored in the image storage unit is necessary for the calculation of the value, the image storage unit stores the pixel value necessary for calculating the pixel value in the decimal position. The pixel value of the pixel located outside the image area is substituted with the pixel value shown at a line-symmetric position across the boundary between the pixel located outside the area and the image area closest to the pixel. The motion vector detection device according to claim 5, wherein a pixel value at a decimal position is calculated. A motion vector detection method for detecting a motion vector of an encoding target block included in an image to be encoded,
A first detection step of detecting an integer-precision motion vector of the encoding target block;
Pixels stored in the image storage unit among the decimal position image blocks composed of pixels at the decimal positions located around the integer position pixels included in the image block specified by the integer precision motion vector A motion vector is detected by using only an image block at a decimal position composed of pixels at a decimal position at which a pixel value can be calculated using only the value and an image block at an integer position specified by the integer precision motion vector as search candidates. A motion vector detection method comprising: 2 detection steps. An integrated circuit for detecting a motion vector of an encoding target block included in an image to be encoded,
Image storage means for storing pixel values of the search area;
First detection means for detecting an integer-precision motion vector of the encoding target block;
Pixels stored in the image storage unit among the decimal position image blocks composed of pixels at the decimal positions located around the integer position pixels included in the image block specified by the integer precision motion vector A motion vector is detected by using only an image block at a decimal position composed of pixels at a decimal position at which a pixel value can be calculated using only the value and an image block at an integer position specified by the integer precision motion vector as search candidates. An integrated circuit comprising: 2 detection means.

Images