JP4385875B2 - Motion vector detection device and program - Google Patents

Description

  The present invention relates to a motion vector detection device and a program, and more particularly to a motion vector detection device and a program suitable for block matching motion detection.

  In encoding processing for compressing moving image data, a difference between a reference image (reference frame) obtained by decoding an already encoded image and a current image (current frame) to be encoded is detected. Therefore, it is common to detect a motion vector based on a macroblock which is a predetermined number of pixel groups. That is, a block (reference block) on the reference image that minimizes an error from the macroblock set on the current image is searched, and a motion vector between the reference coordinates on the macroblock and the reference coordinates on the reference block is obtained. To detect.

  As a method for searching for a reference block on such a reference image, for example, a “full search procedure” is known. In this method, calculation is performed for all block patterns in the search area on the reference image. Here, for example, since sum of absolute difference (SAD) is generally used for motion vector detection between blocks, detection is performed in a full search procedure for all block patterns. While it is certain, the total calculation amount for motion vector detection becomes enormous. In addition, even if there is little change in the image, the calculation is performed for all blocks, so the processing may not be efficient.

  On the other hand, a “step search procedure” (multi-stage search) is also known in which a motion vector is simply detected by performing a plurality of steps without performing calculations for all block patterns. In particular, when searching for a fractional pixel unit, the amount of calculation can be reduced by using a step search procedure. That is, in the step search procedure, a coarse search (integer search) in units of integer pixels is performed in the first stage, and a fine search (fractional search) in units of fractional pixels is performed in the second stage around the integer position where the difference value is minimum. Do as. By introducing a coarse search in the first stage, the amount of calculation can be reduced compared to the full search procedure.

  However, in the fractional search in the second stage, the difference absolute value sum calculation is performed with the fractional pixel accuracy in the eight directions around the optimum position searched in the integer search, so that the calculation amount in the second stage increases. In order to solve such a problem, a method for reducing the amount of calculation in the second stage has been proposed (for example, Patent Document 1).

  In the method of Patent Document 1, the block having the best matching result (best block) and the second best block (next block) are identified from the blocks searched by the integer search in the first stage, and the best block is determined. The direction of the next block viewed from the viewpoint is the direction of the target fraction position of the fraction search in the vicinity of the center of the best block. That is, by using the vector_min indicating the center of the best block and the vector_second indicating the center of the next block, the fractional position to be subjected to the fractional search is determined, and matching is performed only at the relevant fractional position to obtain the fractional search (first The amount of computation is reduced in two stages.

  In the above method, vector_min and vector_second are determined based on the matching order in units of blocks, and the fractional search direction is determined based on this. For this reason, for example, when the integer search range is relatively wide, a block that is distant from the best block may be considered as the next block. In such a case, the distance from the search center to the next block may be longer than the best block. Here, the purpose of the fraction search (second stage) in the multistage search is to detect a fraction vector near the best integer position (vector_min). As described above, the block far from the best block is the next one. In the case of a point block, the direction from the best block to the next block (that is, the direction obtained by vector_min and vector_second) may be different from the direction from the best integer position (vector_min) to the actual best fractional position. In this case, since the fraction vector is obtained based on the fractional position that is not the best, the result of the fraction search may not be good.

Also, at the time of integer search, a difference from the current frame (encoding target frame) is taken for the integer positions of 4 points (or 8 points) in the vicinity of the block, and the block is moved in the direction of the integer position where a good result value is obtained. Thus, a method for searching for the best block without moving the block over the entire region of the search range is also known. On the other hand, in the method described in Patent Document 1, since ranking is performed in units of blocks in integer search, it is necessary to obtain evaluation values for all blocks within the search range. Therefore, it cannot be applied to the method for reducing the block movement amount as described above, and it is difficult to effectively reduce the calculation amount.
JP-A-10-247242

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a motion detection device and the like that can reduce the amount of calculation in the fraction search and obtain a good fraction search result.

In order to achieve the above object, a motion vector detection device according to a first aspect of the present invention includes:
A motion vector detection device for detecting a motion vector for a current block on a current frame by performing integer search and fractional search on a reference frame,
Results of the search of the integer unit, the difference between pixels of the reference frame and the pixel of the current block is an integer position is minimum as the first integer positions, among integer position near the periphery of the first integer position, the first a second integer position specifying means for specifying the difference is small integer position of the pixel and the pixel of the current block of the reference frame to the next integer position as the second integer position of,
A fractional position estimating means for estimating a fractional position with a predetermined accuracy based on the first integer position and the second integer position;
A fraction search means for performing a fraction search with higher accuracy than the predetermined accuracy based on the first integer position and the fraction position estimated by the fraction position estimation means;
With
The fraction search means includes:
Comparing the difference between the reference frame pixel at the first integer position and the current block pixel with the difference between the reference frame pixel at the fractional position estimated by the fractional position estimation means and the current block pixel; and A determination means for determining whether the position where the difference is small is either;
Search center designating means for designating a position determined by the determining means as having a smaller difference as a search center in a fractional search with higher accuracy than the predetermined accuracy;
Is further provided.

In the motion vector detection device,
It is desirable that the second integer position specifying unit specifies a second integer position from among integer positions adjacent to the first integer position.

in this case,
Preferably, the second integer position specifying means specifies a second integer position from integer positions adjacent to each other in eight directions around the first integer position.

In the motion vector detection device,
The fractional position estimating means preferably estimates an intermediate position between the first integer position and the second integer position as the fractional position with the predetermined accuracy.

In the motion vector detection device,
The second integer position specifying means, the difference between pixels of the first reference frame at integer positions pixel and the current block is, from the difference between the pixel of the pixel and the current block of the reference frame in the estimated fractional position If small, of the integer positions in the vicinity around the first integer position, the second integer positions the new difference is small integer positions of the pixels in the reference frame to the next pixel of the current block of the second integer positions It is good.

In order to achieve the above object, a program according to the second aspect of the present invention is:
A program for detecting a motion vector of a moving image on a computer,
On the computer,
An integer unit search is performed on the reference frame, and the integer position where the difference between the pixel of the reference frame and the pixel of the current block of the current frame is the minimum is set as the first integer position. of integer positions, a function of specifying the difference is small integer position of the pixel of the first next to the reference frame pixel and the current block of integer positions as a second integer positions,
A function of estimating an intermediate position between the first integer position and the second integer position as a fractional position with a predetermined accuracy;
The difference between the reference frame pixel at the first integer position and the current block pixel is compared with the difference between the reference frame pixel at the estimated fractional position and the current block pixel, and the difference is smaller. A function to determine whether or not
A function for performing a fractional search with higher accuracy than the predetermined accuracy, with the position determined by the determining means as having a smaller difference as a search center;
A function to calculate a motion vector for the current block based on the result of the fraction search;
It characterized in that to realize.

The above program
In the computer,
The difference between the pixel and the current block pixels in the reference frame in said first integer position, when the pixels of the reference frame in the estimated fractional positions less than the difference between the pixel of the current block, the first integer position among integer positions around the vicinity, functions as a second integer position difference new a small integer position of the pixel of the second of the next reference frame of integer positions pixel and the current block,
Is further realized.

  According to the present invention, the second integer position (next-point integer position) having the smallest difference from the current block next to the best integer position around the vicinity of the first integer position (best integer position) obtained by integer search. ) And an intermediate position (for example, 1/2 pel position) between the best integer position and the next integer position is estimated as a fractional position obtained by a fractional search with a predetermined accuracy (for example, 1/2 pel). Then, among the estimated fraction positions (provisional fraction positions) and the best integer positions, a fraction search with higher accuracy (for example, 1/4 pel) is performed with a position having a smaller difference from the current block as a search center. Therefore, for example, when a motion vector is detected by a multi-stage search such as “integer search (first stage) → 1/2 pel search (second stage) → 1/4 pel search (third stage)” Since the result in the second stage is estimated from the result of the integer search, the amount of computation in the second stage can be reduced. As a result, it is possible to increase the processing speed in the encoding of moving images, reduce the power consumption of the apparatus, and the like.

  Here, the next integer position is identified from the vicinity of the best integer position, more specifically, from the integer positions adjacent to the best integer position. For this reason, a position that does not correlate with an actual fractional motion vector, which can be generated by specifying the best integer position and the next integer position in units of blocks within the search range, does not become the next integer position. Furthermore, if the difference from the current block at the best integer position is smaller than the difference from the current block at the fractional position estimated based on the specified next integer position, the next point around the best integer position The position where the difference from the current block is next to the integer position is the next integer position. For this reason, even if the position with the next smallest difference after the best integer position is not highly correlated with the actual fractional motion vector due to various conditions or the like, a position having a high correlation with the actual fractional motion vector is searched. Therefore, high-speed processing can be performed stably.

(Embodiment 1)
Embodiments according to the present invention will be described below with reference to the drawings. In the present embodiment, a case where the present invention is applied to an encoding apparatus that encodes moving image data will be described below as an example. The encoding apparatus 100 according to the present embodiment uses a predictive differential encoding method that predicts a next image by motion compensation between images of a predetermined unit (for example, a frame or a field) constituting input moving image data. Assume that image data is encoded. In this embodiment, an input unit image is a “frame”. That is, the encoding apparatus 100 according to the present embodiment performs encoding by inter-frame prediction. Note that the image unit is not limited to a frame and is arbitrary, and may be, for example, a “field”.

  A configuration of the encoding apparatus 100 according to the present embodiment will be described with reference to FIG. FIG. 1 is a functional block diagram illustrating a functional configuration of the encoding device 100 according to the present embodiment.

  As illustrated, the encoding apparatus 100 according to the present embodiment includes a transform unit 110, a quantizer 120, an encoder 130, an inverse quantizer 140, an inverse transform unit 150, and a restored image memory 160. And a motion detection unit 170 and a motion compensation unit 180.

  The conversion unit 110 performs data compression by performing conversion encoding on the input frame and difference data. Here, for example, the moving image data is converted into the spatial frequency domain by performing integer transform of 4 × 4 blocks. That is, data compression is performed by biasing the transform coefficient in the low frequency range using the correlation with adjacent pixels.

  The quantization unit 120 quantizes the value of the transform coefficient obtained by the transform unit 110 by, for example, representing a multiple of the quantization step width, and performs data compression. That is, the conversion coefficient used for conversion in the conversion unit 110 is mapped to a discrete level. Such quantization reduces the amount of high frequency code.

  The encoding unit 130 outputs, for example, entropy encoding the data quantized by the quantization unit 120 and encodes it using a code such as a VLC (Variable Length Coding) method. To do. More specifically, for example, Exponential Golomb Coding or CAVLC (Context-based Adaptive Variable Length Coding) is used. Furthermore, in order to reduce the amount of codes to be generated, encoding may be performed using an arithmetic code such as CABAC (Context-based Adaptive Binary Arithmetic Coding). Data encoded by the encoding unit 130 (hereinafter referred to as “encoded bit stream”) is output to a predetermined transmission medium, recording medium, or the like. The output encoded bit stream is decoded by a predetermined decoding device to reproduce moving image data.

  The inverse quantization unit 140 performs inverse quantization on the data quantized by the quantization unit 120, and the inverse transform unit 150 performs quantization by performing inverse transform on the data quantized by the inverse quantization unit 140. The input data (quantized frame) is restored.

  The restored image memory 160 is a storage device configured by, for example, a semiconductor storage element, and is a so-called frame memory that stores a frame restored by the inverse conversion unit 150 (hereinafter referred to as “restored frame”). That is, the pre-encoding data for the output encoded frame is stored. The restored frame stored in the restored image memory 160 is used as a “reference frame” referred to in inter prediction.

  The motion detector 170 detects the motion of the moving object in the image by detecting the difference between the current input frame (current frame) and the frame before the current frame (reference frame), and the motion vector (motion Vector). In this embodiment, a macro block (Macro Block: MB) composed of a predetermined number of pixels is set on the current frame, and a block (reference block) having the smallest difference from the macro block is searched on the reference frame. Then, motion detection is performed by a so-called “block matching method” in which a vector indicating a difference between these is detected as a motion vector.

  Here, the motion detection unit 170 of the present embodiment detects motion vectors up to a fractional unit by so-called “multi-stage search”, in which a search is performed in units of fractional pixels after searching for motion vectors in units of integer pixels. Here, the search in units of integer pixels is referred to as “integer search”, and the search in units of fractional pixels is referred to as “fractional search”.

  The motion compensation unit 180 performs motion compensation on the previous frame based on the motion vector detected by the motion detection unit 170. That is, a prediction frame is created based on the detected motion vector. Then, the difference between the input frame and the prediction frame is an encoding target.

  The motion detection unit 170 according to the embodiment of the present invention is a multi-stage search such as “integer search (first stage) → 1/2 pel search (second stage) → 1/4 pel search (third stage)”. When detecting a motion vector, the processing amount is increased and the frame rate is improved by reducing the amount of calculation in the second stage. The configuration of the motion detection unit 170 that performs such an operation will be described with reference to FIG.

  FIG. 2 is a block diagram illustrating a configuration of the motion detection unit 170. As shown in the figure, the motion detection unit 170 includes an integer search unit 171, a best integer position determination unit 172, a surrounding integer position evaluation unit 173, a 1/2 pel calculation unit 174, a search center determination unit 175, a 1/4 pel calculation unit 176, A vector calculation unit 177;

  The integer search unit 171 includes a pixel value of a macroblock (hereinafter referred to as “current block”) designated as an encoding target block in the input current frame, a pixel value of a search block on the input reference frame, and Is obtained by integer search. In the present embodiment, a predetermined search range is set on the reference frame, and the search block is shifted by an integer number of pixels within the search frame, and the difference between the pixel value of each search block and the pixel value of the current block is, for example, Then, the sum of absolute differences (SAD) is obtained, and the result of each operation is output as an evaluation value for each search block.

  The best integer position determination unit 172 determines, based on the evaluation value output from the integer search unit 171, which block has the smallest difference from the current block (hereinafter referred to as “best integer block”). And identified as the best integer position (first integer position).

  The surrounding integer position evaluation unit 173 has the best evaluation value among the integer blocks adjacent in the eight directions around the best integer block determined by the best integer position determination unit 172 based on the evaluation value output by the integer search unit 171. A block (hereinafter referred to as “next-point integer block”) is determined and specified as the next-point integer position (second integer position).

  The 1 / 2pel calculation unit 174 is obtained by a 1/2 pixel (1 / 2pel) unit search (1 / 2pel search) based on the next integer block and the best integer block determined by the surrounding integer position evaluation unit 173. A fractional position to be estimated and a fractional position estimated (hereinafter referred to as “provisional fractional position”), a pixel value of a search block (hereinafter referred to as “provisional fractional block”) at the provisional fractional position, and a pixel of the current block A difference from the value is calculated by, for example, a difference absolute value sum calculation, and the calculation result is output as an evaluation value.

  The search center determination unit 175 compares the evaluation value of the next integer block and the evaluation value of the provisional fractional block, and determines which evaluation value is better, that is, whether the difference is smaller. The search center determination unit 175 determines the search center of the subsequent fraction search (1/4 pixel accuracy) based on the determination result.

  The 1 / 4pel calculation unit 176 performs a fractional search (1 / 4pel search) in units of 1/4 pixel (1 / 4pel) based on the search center determined by the search center determination unit 175. Here, while the search block is shifted by 1/4 pixel, the difference between the pixel value at each shift position and the pixel value of the current block is obtained by, for example, the sum of absolute differences, and the best evaluation value is obtained. Block (hereinafter referred to as “best fractional block”).

  The vector calculation unit 177 calculates and outputs a “motion vector” for the current block based on the result obtained by the integer search by the integer search unit 171 and the 1/4 pel search by the 1 / 4pel operation unit 176. To do.

  Each configuration of the motion detection unit 170 is controlled by a predetermined control signal. Such a control signal can be supplied to each unit from a control unit (not shown) of the encoding device 100 including, for example, a CPU (Central Processing Unit). It may be supplied from an external device or the like.

  The encoding apparatus 100 configured as described above encodes moving image data in the following procedure.

  The moving image data input to the encoding device 100 is sent to the conversion unit 110 and the motion detection unit 170 in units of frames. The motion detection unit 170 detects a motion vector by detecting a difference between the input frame and the reference frame. The motion compensation unit 180 compensates the reference frame based on the motion vector to generate a prediction frame. Then, the difference between the input frame and the predicted frame is taken, and the difference data is encoded by the encoding unit 130 after being converted by the conversion unit 110 and quantized by the quantization unit 120. The motion vector data detected by the motion detector 170 is also encoded by the encoder 130, and both data are multiplexed and output.

  On the other hand, the difference data quantized by the quantization unit 120 is restored by being inversely quantized by the inverse quantization unit 140 and inversely transformed by the inverse transform unit 150. The restored difference data and the predicted frame generated by the motion compensation unit 180 are added and stored in the restored image memory 160 as a “restored frame” (restored image). The restored frame stored in the restored image memory 160 is input to the motion detection unit 170 as a “reference frame” (reference image). The motion detection unit 170 detects a motion vector indicating a difference between the input reference frame and the current frame.

  The motion detection unit 170 according to the present embodiment obtains a difference vector between the current frame and the reference frame by a so-called block matching method. That is, a block that approximates a macroblock on the current frame is searched on the reference frame, and a difference vector between the searched block (reference block) and the macroblock is detected as a motion vector.

  Here, in this embodiment, motion vectors are obtained by “multi-stage search” in which search is performed in the order of “integer search (first stage) → 1/2 pel search (second stage) → 1/4 pel search (third stage)”. To detect. The operation of the motion detector 170 in this case (“motion vector detection process 1”) will be described with reference to the flowchart shown in FIG. This “motion vector detection process 1” is started when the current frame is input to the motion detection unit 170.

  When the current frame is input to the motion detection unit 170, an integer search is performed by the integer search unit 171 (step S101).

  In the present embodiment, in the case of integer search, as shown in FIG. 4A, a block having the same position and size as the current block CB specified on the current frame CF is defined on the reference frame RF as a search block SB1_INT. To do. In the present embodiment, an integer search, that is, a search block used in the first stage search is denoted as a search block SB1. Note that the block sizes of the current block CB and the search block SB1 are arbitrary, and are hereinafter referred to as “search block size”.

  The operation of integer search will be described with reference to FIG. The integer search unit 171 defines a search range SA1 of a predetermined size centered on the search block SB1_INT set on the reference frame, and within this search range SA1, starts from the search block SB1_INT (block of the same size (search block) While SB1) is shifted by one integer pixel, the difference between the pixel value of the current block CB and the pixel value of each search block SB1 is calculated, for example, by calculating the sum of absolute differences, and the calculation result is evaluated for each search block SB1. The value is output to the best integer position determination unit 172 as a value.

  In the present embodiment, one of the pixels in the block is used as a reference point, and the position of the reference point is designated to determine the position of the block. Here, as shown in FIG. 4B, the integer position at the upper left corner of the block is assumed to be a reference point, and the reference point for the search block SB1 used in the integer search is expressed as a reference point OP1.

  The best integer position determination unit 172 determines which of the evaluation values output from the integer search unit 171 is the best evaluation, that is, the block having the smallest difference from the current block (best integer block SB_min). Then, it is specified as the best integer position (first integer position) (step S102). That is, the block having the smallest difference from the current block in integer units among the blocks in which the integer search range SA1 is shifted by one integer pixel starting from the search block SB1_INT is defined as the best integer block SB1_min (FIG. 4 ( b)).

  The best integer position determination unit 172 outputs position information indicating the position of the identified best integer block SB1_min to the surrounding integer position evaluation unit 173 and the vector calculation unit 177. In this case, for example, information indicating the coordinates of the reference point OP1_min of the best integer block SB1_min is output as position information. The best integer position determination unit 172 also outputs the evaluation value of each search block SB1 acquired from the integer search unit 171 to the surrounding integer position evaluation unit 173.

  The surrounding integer position evaluation unit 173 determines which block (next-point integer block SB1_min2) has the smallest difference from the current block next to the best integer block SB1_min in the vicinity of the best integer block SB1_min obtained by the integer search. The next point integer position (second integer position) is determined and determined (step S103). In the present embodiment, the next integer block SB1_min2 is specified from among the blocks adjacent in the eight directions around the best integer block SB1_min.

  When the best integer block SB1_min is identified based on the position information of the best integer block SB1_min acquired from the best integer position determination unit 172, the surrounding integer position evaluation unit 173 includes blocks adjacent in the eight directions around the best integer block SB1_min. A “surrounding block search range SA1 ′” as shown in FIG. 5A for defining the range is set. As shown in FIG. 5B, the surrounding block search range SA1 'is a range of 3 × 3 integer pixels adjacent in the eight directions around the reference point OP1_min of the best integer block SB1_min. In this case, integer pixels (integer positions) adjacent in the eight directions around the reference point OP1_min are set as integer pixels A to H (integer positions A to H).

  The surrounding integer position evaluation unit 173 specifies a search block SB1 in which each integer position in the surrounding block search range SA1 'is a reference point. That is, these search blocks SB1 are search blocks SB1a to SB1h adjacent to the best integer block SB1_min in the eight directions as shown in FIGS. 6 (a) to (h).

  When such search blocks SB1a to SB1h are in the integer search range SA1, the evaluation values of these blocks are obtained in the integer search in step S101. Therefore, the surrounding integer position evaluation unit 173 extracts the evaluation values of the search blocks SB1a to SB1h from the acquired evaluation value of the search block SB1, and identifies which block has the smallest difference among them. .

  Here, when the best integer block SB1_min is located at the edge of the integer search range SA1, for example, some of the search blocks SB1a to SB1h may not be calculated in the integer search. In such a case, the surrounding integer position evaluation unit 173 outputs the position information of the corresponding block to the integer search unit 171 and instructs the integer search unit 171 to perform an integer search of a necessary part at any time, and search blocks SB1a to SB1h All evaluation values of are obtained.

  Here, the motion vector position in the fractionally interpolated image has a very high correlation with the position having the second smallest difference in the vicinity of the best integer position obtained by the integer search. For this reason, based on the position of the best integer block SB1_min specified in step S102 and the position of the next integer block SB1_min2 specified in step S103, the motion vector position (fractional position) obtained by the fractional search is temporarily estimated. be able to.

  That is, the 1/2 pel calculation unit 174 estimates the fractional position obtained by the 1/2 pel search based on the best integer block SB1_min identified in step S102 and the next integer block SB1_min2 identified in step S103 (step S104). That is, in a normal search operation, the search is performed in eight directions (or four directions) around the search center to obtain the fractional position. Here, the fractional position is estimated by one calculation. The estimated fractional position is a fractional position (hereinafter referred to as “provisional fractional position HP”) that is a search center candidate for a higher-accuracy fractional search (ie, 1 / 4pel search).

  Here, for example, when the next integer block SB1_min2 is the search block SB1c (FIG. 6C), as shown in FIG. 7A, the reference point OP1_min of the best integer block SB1_min and the next integer block SB1_min2 , That is, the intermediate position (1/2 position) between the surrounding block search range SA1 ′ and the integer position C is defined as the provisional fraction position HP. In this case, the block having the smallest difference from the current block in integer units is the best integer block SB1_min, and the block having the next smallest difference next to the best integer block SB1_min (next-point integer block SB1_min2) around the best integer block SB1_min. ) Is the search block SB1c.

  Since a motion vector with a fractional pixel accuracy (fractional motion vector) tends to be highly correlated with such a next integer block, the actual fractional motion is based on the position of the next integer block SB1_min2 with respect to the best integer block SB1_min. A vector can be estimated. In the case of FIG. 7A, since the next integer block SB1_min2 is the search block SB1c, a vector indicating the fractional position between the reference point OP1_min of the best integer block SB1_min and the reference point of the search block SB1c is an actual fraction. It can be estimated as a motion vector.

  When the provisional fraction position HP is specified, as shown in FIG. 7B, the provisional fraction position HP becomes a reference point (hereinafter referred to as “reference point OP2”) (hereinafter referred to as “provisional fraction block SB2”). Stipulate). The block size of the provisional fractional block SB2 is the same as the search block size. The 1 / 2pel calculation unit 174 performs pixel values at intermediate positions (1/2 positions) between the integer pixels by using an interpolation method such as “6tap FIR” for the pixels included in the range of the provisional fractional block SB2. ½ pixel interpolation is performed (see FIG. 8A), and a difference calculation with the current block (1/2 pel calculation) is performed (step S105).

  In other words, the 1 / 2pel calculation unit 174 calculates the pixel value of the provisional fractional block SB2 from the reference frame, and calculates the difference from the pixel value of the current block CB by, for example, the difference absolute value sum calculation. The 1 / 2pel calculation unit 174 outputs the obtained difference value to the search center determination unit 175 as the evaluation value of the provisional fractional block SB2.

  The search center determination unit 175 acquires the evaluation value of the provisional fractional block SB2 from the 1 / 2pel calculation unit 174 and also acquires the evaluation value of the best integer block SB1_min from the best integer position determination unit 172. Then, the evaluation value of the provisional fractional block SB2 and the evaluation value of the best integer block SB1_min are compared, and which block has a better evaluation value, that is, which block has a smaller evaluation value. Determine if there is. That is, it is determined which of the provisional fraction block SB2 based on the provisional fraction position HP and the block showing the smaller difference value is the best integer block SB1_min (step S106).

  The search center determination unit 175 determines a position to be used as the search center of a fractional search (1/4 pel search) to be executed next in 1/4 pixel units according to the determination result. That is, when the difference value of the provisional fraction block SB2 is smaller (step S106: Yes), the search center determination unit 175 sets the search center of the 1/4 pel search as the provisional fraction block SB2 (step S107).

  On the other hand, when the difference value of the best integer block SB1_min is smaller (step S106: No), the search center determination unit 175 sets the search center for fractional search as the best integer block SB1_min (step S108).

  The search center determination unit 175 outputs position information indicating the position as the search center of the 1/4 pel search to the 1/4 pel calculation unit 176. Here, when the provisional fraction block SB2 is designated as the search center, information indicating the coordinates of the reference point OP2 of the provisional fraction block SB2 is output as position information, and when the best integer block SB1_min is designated as the search center, Information indicating the coordinates of the reference point OP1_min of the best integer block SB1_min is output as position information.

  The 1/4 pel calculation unit 176 performs a fractional search (1/4 pel search) in units of 1/4 pixels based on the position information indicating the search center position determined by the search center determination unit 175 (step S109). Here, a search range whose search center is the position indicated by the position information output by the search center determination unit 175 (hereinafter referred to as “search range SA2”) is defined, and the search block size corresponding to the search block size is defined within this search range SA2. While shifting a block (hereinafter referred to as “search block SB3”) by ¼ pixel, the difference between the pixel value of each block and the pixel value of the current block CB is calculated.

  In performing the fractional search in units of 1/4 pixel, the 1 / 4pel calculation unit 176 performs 1/4 pixel interpolation on at least the range in which the search block SB3 moves according to the search range SA2. In this case, for each of the pixels subjected to ½ pixel interpolation as shown in FIG. 8A, for example, by using an interpolation method such as “linear interpolation”, each ½ as shown in FIG. Interpolation is performed so that a pixel value at an intermediate position (1/4 position) between the positions is obtained.

  The operation of the fraction search when the provisional fraction position HP is set as the search center of the fraction search in step S107 will be described below. FIG. 9A is a diagram illustrating the provisional fractional block SB <b> 2 interpolated by ¼ pixel. The 1/4 pel calculation unit 176 defines a search range for 1/4 pel search on such a provisional fractional block SB2. In this case, for example, as shown in FIG. 9B, the 1 / 4pel calculation unit 176 has a range including 1/4 positions adjacent to each other in eight directions around the reference point OP2 of the provisional fractional block SB2. The search range is SA2.

  The 1 / 4pel calculation unit 176 sequentially designates blocks in which each fractional position in the search range SA2 is a reference point (that is, the top left position of the block) as a search block SB3 for 1 / 4pel search. Difference calculation (1/4 pel calculation) between the pixel value of SB3 and the pixel value of the current block CB is performed. Here, a 1/4 pel search is performed starting from the provisional fractional block SB2 as shown in FIG.

That is, as shown in FIG. 10B, the 1 / 4pel calculation unit 176 performs a 1 / 4pel search for a position shifted by one fractional pixel (one quarter pixel) from the reference point OP2 of the provisional fractional block SB2. of a reference point OP3 ₁ search block SB3 ₁ as a matching target, the reference point OP3 ₁ specifies the search block SB3 ₁ as the upper left apex position of the block. The block size of the search block SB3 is equivalent to the search block size.

1 / pel operation unit 176, the search block SB3 _1, performs difference calculation between the current block CB. That is, to get the pixel values of the search block SB3 ₁ from the reference frame, calculating a difference between the pixel value of the current block CB. In this case, the 1 / pel operation unit 176, for example, obtains a difference value by the difference absolute value sum calculation, a difference value obtained as the evaluation value of the search block SB3 _1.

When the evaluation value for the search block SB31 is acquired, the position obtained by shifting the reference point OP3 ₁ by one fractional pixel is set as the next reference point OP3 _2, and the difference calculation is performed on the search block SB3 ₂ as in the case of the search block SB3 _1. . In this way, the 1 / 4pel calculation unit 176 sequentially specifies the search blocks SB3 _{1 to} SB3 _{8 in} which each fractional position in the search range SA2 is a reference point, starting from the reference point OP2 of the provisional fractional block SB2. Each search block SB3 is matched with the current block CB. In this case, for example, the search blocks SB3 _{1 to} SB3 ₈ are sequentially specified in the order indicated by the arrows in FIG.

The 1 / 4pel calculation unit 176 has the best evaluation value among the obtained evaluation values of the search blocks SB3 _{1 to} SB3 ₈ , that is, the block having the smallest difference value (hereinafter referred to as “search block SB3_min”). And position information indicating the position of the search block SB3_min is output to the vector calculation unit 177. In this case, for example, information indicating the coordinates of the reference point OP3_min of the search block SB3_min is output as position information.

  The above is the 1/4 pel search when the evaluation of the provisional fraction block SB2 based on the provisional fraction position HP is better than the best integer block SB1_min.

  On the other hand, if the evaluation of the best integer block SB1_min is better than that of the provisional fractional block SB2, a 1/4 pel search is performed with the best integer block SB1_min as the search center. The operation in this case will be described below.

  FIG. 11A is a diagram illustrating the best integer block SB1_min subjected to 1/4 pixel interpolation. The 1 / 4pel calculation unit 176 defines a search range for 1 / 4pel search on the best integer block SB1_min. In this case, as shown in FIG. 11B, a range including quarter positions adjacent in the eight directions around the reference point OP1_min of the best integer block SB1_min is set as a search range SA2.

  The 1 / 4pel calculation unit 176 sequentially specifies the search block SB3 in which each fractional position in the search range SA2 thus defined is the reference point OP3, and calculates the difference between each search block SB3 and the current block CB. This operation is the same as the search operation when the provisional fractional block SB2 is used as the search center (see FIG. 10).

That is, the 1 / 4pel calculation unit 176 obtains the evaluation values of the search blocks SB3 _{1 to} SB3 ₈ and determines which _one of the search values SB3_min is the best evaluation value, that is, the difference value is the smallest. The position information indicating the position of the search block SB3_min is output to the vector calculation unit 177. In this case, for example, information indicating the coordinates of the reference point OP3_min of the search block SB3_min is output as position information.

  The vector calculation unit 177 calculates a “motion vector” for the current block CB based on the integer position obtained by the integer search and the fraction position obtained by the fraction search (step S110). That is, the vector indicating the difference between the search block SB1_INT corresponding to the current block CB and the search block SB3_min in which the difference between the current block CB and the fractional search is minimized is the “motion vector” in the block. Accordingly, the vector calculation unit 177 indicates position information indicating the coordinates of the reference point OP1_INT of the search block SB1_INT acquired from the integer search unit 171 and the coordinates of the reference point OP3_min of the search block SB3_min acquired from the 1 / 4pel calculation unit 176. Based on the position information, a “motion vector” is calculated. That is, a vector indicating a deviation amount and direction with respect to the reference point OP3_min is calculated as a “motion vector” with the reference point OP1_INT as the origin.

  Note that the vector calculation unit 177 may individually obtain an integer unit motion vector (integer motion vector) and a fractional unit motion vector (fractional motion vector) in accordance with the processing related to motion compensation. Here, since the “integer motion vector” is a vector indicating a deviation between the search block SB1_INT and the best integer block SB1_min, the vector calculation unit 177 determines the position of the reference point OP1_INT of the search block SB1_INT and the best integer block SB1_min. An “integer motion vector” can be calculated based on the position of the reference point OP1_min. Further, since the “fractional motion vector” is a vector indicating a deviation between the best integer block SB1_min and the search block SB3_min, the vector calculation unit 177 determines the position of the reference point OP1_min of the best integer block SB1_min and the reference of the search block SB3_min. A “fractional motion vector” can be calculated based on the position of the point OP3_min.

  The vector calculation unit 177 outputs the “motion vector” obtained in this way to the motion compensation unit 180 and ends the process. The motion compensation unit 180 performs motion compensation based on the motion vector output from the motion detection unit 170 and creates a prediction frame. Then, only the difference between the predicted frame and the current frame is encoded by the encoding unit 130 and output as an encoded bit stream.

  As described above, according to the motion detection unit 170 according to the first embodiment, the position of the best integer block SB1_min (first integer position) detected by the integer search and the periphery of the best integer block SB1_min Next, the fractional position obtained by the 1/2 pel search is estimated based on the position (second integer position) of the next-point integer block SB1_min2 with good evaluation. That is, the actual fractional motion vector obtained by the fractional search tends to be highly correlated with the direction of the next best evaluated integer position around the best position in integer units. The 1/2 position between and is estimated as the fractional position obtained by the 1 / 2pel search (provisional fractional position). Thereby, since the difference calculation performed by a 1 / 2pel search is performed only about a provisional fraction position, the amount of calculation can be reduced compared with the case where the difference calculation for eight directions is performed.

  Here, according to the first embodiment, the next-point integer block SB1_min2 is obtained from blocks adjacent to the periphery of the best integer block SB1_min. As a result, it is possible to avoid a problem that may occur in the conventional technique for obtaining the best integer position and the next integer position for each block within the search range of the integer search.

  That is, when the best position and the next point position are determined in block units within the search range, a block far from the block including the best integer position may be the next point. Here, since the fractional motion vector to be obtained indicates a fractional level error between the best integer position and the actual best position, the fractional search vector based on the next integer position that is not in the vicinity of the best integer position is used. Even if the search center is determined, it may not be correlated with the actual fractional motion vector. Therefore, by determining the next integer block SB1_min2 from the blocks adjacent in the eight directions around the best integer block SB1_min as in the first embodiment, it is possible to detect the motion vector in fractional units quickly and accurately. .

(Embodiment 2)
As described above, the fractional motion vector detected by the fractional search tends to have a high correlation with the next integer block near the best integer block detected by the integer search. Tentative fractional positions were determined based on various trends. However, depending on various conditions, even the next integer block around the best integer block may not be highly correlated with the actual fractional motion vector. Even in such a case, the operation of the motion detection unit 170 that detects a fractional motion vector accurately and quickly will be described below as a second embodiment. Note that the configuration of the motion detection unit 170 according to the present embodiment is the same as the configuration of the motion detection unit 170 of the first embodiment. In addition, the notation relating to the block and the like is the same as in the first embodiment. Therefore, the same or similar configurations as in the first embodiment will be described with the same reference numerals.

  The operation of the motion detection unit 170 according to the present embodiment (“motion vector detection process 2”) will be described with reference to the flowcharts shown in FIGS. This “motion vector detection process 2” is started when the current frame is input to the motion detection unit 170, similarly to the “motion vector detection process 1” according to the first embodiment.

  When the processing is started, “integer search” is executed by the integer search unit 171 (step S201). The integer search here is the same operation as in the first embodiment. That is, the integer search unit 171 sets a predetermined search range SA1 centered on the reference point OP1_INT of the search block SB1_INT, obtains a difference from the current block CB while shifting the search block SB1 by one integer pixel, Is output to the best integer position determination unit 172 and the search center determination unit 175 as the evaluation value. Based on the evaluation value, the best integer position determination unit 172 identifies the best integer block SB1_min (first integer position) that minimizes the difference from the current block CB.

  When the best integer block SB1_min is specified, the surrounding integer position evaluation unit 173 specifies search blocks SB1a to SB1h adjacent in the eight directions around the best integer block SB1_min as in the case of the first embodiment. Evaluation values for two search blocks SB1 are obtained.

  Here, the surrounding integer position evaluation unit 173 ranks the search blocks SB1a to SB1h based on the obtained evaluation values in the order of good evaluation, that is, in ascending order of the difference values (step S202). In this case, for example, the surrounding integer position evaluation unit 173 has a predetermined storage area, and stores information indicating the position of each block (for example, the coordinates of the reference point) and the rank in association with each other. To do. This storage area may be configured in a portion other than the surrounding integer position evaluation unit 173.

  The surrounding integer position evaluation unit 173 sets a rank pointer for designating the target rank to an initial value “1” indicating “first rank” (step S203), and the reference point OP1 of the search block SB1 with the rank indicated by the rank pointer A half position between the (second integer position) and the reference point OP1_min of the best integer block SB1_min is specified as the provisional fraction position HP (step S204).

  The 1 / 2pel calculation unit 174 calculates the difference between the pixel value of the provisional fractional block SB2 based on the specified provisional fraction position HP and the pixel value of the current block CB by the same operation as in the first embodiment, and evaluates the evaluation value. Is output to the search center determination unit 175 (step S205).

  The search center determining unit 175 compares the evaluation value of the best integer block SB1_min with the evaluation value of the provisional fractional block SB2, and determines whether the evaluation value of the provisional fractional block SB2 is better, that is, the best integer block It is determined whether or not the difference value between the provisional fractional block SB2 and the current block CB is smaller than the difference value between the SB1_min and the current block CB (step S206).

  When the evaluation of the provisional fraction block SB2 is better (step S206: Yes), the search center determination unit 175 sets the provisional fraction block SB2 as the search center of the 1/4 pel search (step S207). In this case, a 1/4 pel search is executed by the same operation as in the first embodiment (step S208), and the search block SB3_min having the smallest difference from the current block is specified. Then, the search vector calculation unit 177 calculates a “motion vector” for the current block based on the specified search block SB3_min, outputs it to the motion compensation unit 180 (step S209), and ends the process.

  On the other hand, when the evaluation value of the best integer block SB1_min is better than the evaluation value of the provisional fractional block SB2 (step S206: No), the surrounding integer position evaluation unit 173 determines the rank from the rank currently designated by the rank pointer. It is determined whether there are still lower blocks (step S210: FIG. 13).

  When a block lower than the current rank exists (step S210: Yes), the surrounding integer position evaluation unit 173 increments the rank pointer by 1 (step S211), and proceeds to step S204 (FIG. 12). That is, the search block SB1 of the next rank is set as a new next-point integer block SB1_min2 (second integer position), and a 1/2 position between the reference point OP1 of this block and the reference point OP1_min of the best integer block SB1_min. Is a new provisional fractional position HP. Then, the evaluation value of the provisional fraction block SB2 based on the newly designated provisional fraction position HP is compared with the evaluation value of the best integer block SB1_min (steps S204 to 206).

  On the other hand, when there is no block lower than the current rank, that is, when the current rank pointer indicates “8” indicating “8th place” (step S210: No), the surrounding integer position evaluation unit 173 , Information indicating that is output to search center determination section 175. In this case, the search center determination unit 175 sets the best integer block SB1_min as the search center of the 1/4 pel search (step S220), and proceeds to step S208 (FIG. 12).

  In this case, by the same operation as in the first embodiment, a 1/4 pel search is performed with the best integer block SB1_min as the search center, and the vector calculation unit 177 calculates a motion vector based on the search block SB3_min with the smallest difference. To the motion compensation unit 180.

  In other words, in the motion detection unit 170 according to the second embodiment of the present invention, the search blocks SB1a to SBh adjacent to the periphery of the best integer block SB1_min are ranked based on the respective evaluation values (difference values from the current block). The provisional fractional position HP is obtained from the block having a higher rank. Then, the evaluation value of the provisional fraction block SB2 based on the calculated provisional fraction position HP is compared with the evaluation value of the best integer block SB1_min, and if the evaluation value of the provisional fraction block SB2 is better, the provisional fraction position HP To the search center of 1 / 4pel search.

  On the other hand, if the evaluation of the best integer block SB1_min is better than that of the provisional fraction block SB2, the provisional fraction position HP is determined based on the search block SB1 of the next rank. Similarly, the best integer block SB1_min Compare. In this way, when the provisional fraction block SB2 based on the provisional fraction position HP determined in order of the search blocks SB1a to SB1h is better evaluated than the best integer block SB1_min, the provisional fraction block SB2 is used as a search center. Perform a 1 / 4pel search.

  If any of the provisional fractional blocks SB2 based on the search blocks SB1a to SB1h is not evaluated better than the best integer block SB_min, a 1/4 pel search is performed with the best integer block SB1_min as the search center.

  As described above, in the motion detection unit 170 according to the second embodiment of the present invention, the provisional fractional block SB2 based on the block with the next smallest difference (next integer position) in the vicinity of the best integer block SB_min (best integer position). If the evaluation is not better than the best integer block SB1_min, it is based on the position of the next rank of the next integer position around the best integer position, that is, the block having the smallest difference from the current block next to the next integer position. The provisional fractional position HP is obtained. For this reason, even if the correlation between the next integer block and the actual fractional motion vector is not high due to various conditions, etc., a fractional search is performed by finding a provisional fractional position that has a high correlation with the actual fractional motion vector. can do.

  In this case, the calculation for obtaining the provisional fractional position is performed from the search block SB1a-h adjacent to the best integer block SB1_min in the higher order, and therefore, 1/2 when the better evaluation value is obtained. Pixel interpolation calculation and evaluation value calculation can be terminated. Therefore, even when the correlation between the next-point integer position and the actual fractional motion vector is not high, it is possible to efficiently obtain the 1/2 position as the provisional fractional position.

  In the second embodiment, the order pointer is incremented by 1 until there is no lower block. However, this operation may be terminated before the lower block disappears, depending on the amount of calculation to be allowed. In the second embodiment, provisional fraction positions are designated one by one. However, a plurality of provisional fraction positions may be designated at a time. That is, if it is assumed that the next integer position does not necessarily affect the correlation with the actual fraction level motion vector, in addition to the first position, the provisional fraction position based on the next rank is also set as a candidate in advance. The search center of the fraction search may be determined by comparing the difference between the three points of the two provisional fraction positions and the best integer position and the current block. Here, if the number of computations of 1/2 pel search is less than 8 performed for the surrounding 8 points, the amount of computation will be reduced, so the number of provisional fractional positions that are candidates within this range can be determined. That's fine.

  As described above, according to the motion detection unit 170 according to each of the above embodiments, for example, when a motion vector is detected by a multistage search in the order of “integer search → 1/2 pel search → 1/4 pel search”. Next, the best integer position obtained by the integer search (first integer position) and the integer positions in the vicinity of the best integer position, the position with the smallest difference following the best integer position (second integer position) Based on the above, the fractional position at 1 / 2pel is estimated and narrowed down, so that the amount of computation of 1 / 2pel search operation is reduced. As a result, it is possible to accurately detect a motion vector in a fractional unit and reduce the amount of calculation as a whole. That is, it is possible to achieve both improvement of the frame rate by the search with fractional accuracy, speeding up of processing, and power saving.

  It should be noted that the part that performs motion detection in the decoding device that decodes the encoded bitstream output from the encoding device 100 may have the same configuration and operation as the motion detection unit 170 of the encoding device 100. it can. In other words, in the coding or decoding of moving image data, the portion that performs motion detection by block matching is configured in the same manner as the motion detecting unit 170, thereby improving the frame rate and processing in moving image coding and decoding. Both speeding up can be achieved.

  Note that the function of the motion detection unit 170 according to the above embodiment may be realized by software processing. In this case, for example, the function of the motion detection unit 170 can be realized by executing a program for realizing these functions on a general-purpose device such as a personal computer or a moving image recording / reproducing device.

That is, the program is installed in the storage unit of the general-purpose device, and a control unit such as a CPU executes the program. In this case, the function of the motion detection unit 170 according to the embodiment of the present invention can be realized by a general-purpose device by executing the following program.
(1-1) “Integer search program”: a program for searching the search range in units of integer pixels and detecting an integer position (first integer position) at which the difference from the current block is minimum (1-2) ) "Ambient point evaluation program": Evaluates the difference with the current block around the best integer position obtained by integer search, and next integer position (second integer position) with the next smallest difference after the best integer position (1-3) “fractional position estimation program”: Specify next-point integer position or a predetermined-order peripheral point, and specify the best integer position Program (1-4) “Comparison determination program” for estimating a fractional position obtained by a fractional search (1 / 2pel search) with a predetermined search accuracy based on the surrounding position and designating a provisional fractional position: best In integer positions Program for comparing the difference between the block and the difference between the current block at the provisional fraction position and determining which difference is smaller (1-5) “next integer position designation program”: provisional fraction When the difference at the best integer position is smaller than the difference at the position, the next next integer position (second integer position) is set to the surrounding point of the next rank after the next integer position (second integer position). Program (1-6) for designating as: “Fraction search program”: Predetermined search accuracy (1/2 pel) with the position determined to have a smaller difference between the best integer position and the provisional fraction position as the search center A program (1-7) “vector calculation program” for performing a higher-accuracy fractional search (1 / 4pel search) and detecting a fractional position where the difference from the current block is minimum: the difference from the current block is minimum Integer places And based on the fractional position, a program for calculating a motion vector in the current block

  When the above program is executed by a general-purpose device, a control device such as a CPU performs an integer search unit 171, a best integer position determination unit 172, a surrounding integer position evaluation unit 173, a 1/2 pel calculation unit 174, a search center determination unit 175, 1 For example, a storage device such as a main storage device or an external storage device functions as a predetermined storage area or the like.

  These programs can be provided by being incorporated in a general-purpose device in advance, or the programs themselves may be provided independently. The distribution method of the program in this case is arbitrary. For example, the program can be distributed by being stored in a recording medium such as a CD-ROM or a memory card, or can be distributed via a communication medium such as the Internet. Good. Then, the function of the motion detection unit 170 can be realized by installing the distributed program in a general-purpose computer device or the like and executing it in cooperation with an OS (Operation System: basic software).

  The program for realizing the function according to the present invention may be operated in cooperation with an existing program for encoding / decoding moving images. In other words, by providing a so-called update program for the existing video encoding / decoding program, the motion detection operation as described above is added to the existing video encoding / decoding process, and the frame rate is improved. Processing speed can be increased.

  The descriptions of the above embodiments are examples for applying the present invention, and the scope of the present invention is not limited by the above embodiments. That is, the present invention may be applied in various forms other than the above-described embodiments, and these are included in the scope of the present invention.

  For example, in the above embodiment, the case of multi-stage search of “integer search (first stage) → 1/2 pel search (second stage) → 1/4 pel search (third stage)” is illustrated, but with different search accuracy If the search is performed in stages, the search accuracy and the number of stages may be arbitrary. For example, the processing speed can be increased by applying the present invention to a multi-stage search such as “integer search (first stage) → 1/2 pel search (second stage)”. In this case, if the evaluation value of the provisional fraction block based on the provisional fraction position is better than the evaluation value of the best integer block, the vector indicating the provisional fraction block can be a 1/2 pel fractional motion vector. As a result, the difference calculation normally performed in the eight directions only needs to be performed once, so that the amount of calculation in the second stage is reduced and the processing can be speeded up.

  In the above embodiment, the next-point integer position is obtained from the eight directions around the best integer block, but may be obtained by thinning out, for example, in the four surrounding directions. The same applies to 1 / 4pel search.

It is a block diagram which shows the structure of the "encoding apparatus" concerning Embodiment 1 of this invention. FIG. 2 is a block diagram illustrating a configuration of a “motion detection unit” illustrated in FIG. 1. It is a flowchart for demonstrating "the motion vector detection process 1" concerning Embodiment 1 of this invention. It is a figure for demonstrating "integer search", (a) shows the example of a search block, (b) shows the example of the search range at the time of an integer search, and a search block. It is a figure for demonstrating the operation | movement which searches for a next point integer block, (a) shows the example of the position where a surrounding block search range is set, (b) shows the example of a surrounding block search range. It is a figure for demonstrating the operation | movement which determines a next-point integer block, (a)-(h) shows the example of the search block adjacent to the best integer block. It is a figure for demonstrating the operation | movement which specifies "provisional fraction position", (a) shows the example of the specified provisional fraction position, (b) is the provisional fraction block prescribed | regulated based on provisional fraction position. An example is shown. It is a figure for demonstrating fraction pixel interpolation, (a) shows the example of 1/2 pixel interpolation, (b) shows the example of 1/4 pixel interpolation. It is a figure for demonstrating 1 / 4pel search centering on a provisional fraction block, (a) shows the example of the provisional fraction block by which 1/4 pixel interpolation was performed, (b) is the search of 1 / 4pel search An example of a range is shown. It is a figure for demonstrating the operation | movement of a 1 / 4pel search, (a) shows the example of the temporary fraction block used as the starting point of a search, (b) shows the example of the search block shifted by 1 fraction pixel, ( c) shows an example of the search order. It is a figure for demonstrating 1 / 4pel search centering on the best integer block, (a) shows the example of the best integer block interpolated by 1/4 pixel, (b) is the search of 1 / 4pel search An example of a range is shown. It is a flowchart for demonstrating "the motion vector detection process 2" concerning Embodiment 2 of this invention. It is a flowchart for demonstrating "the motion vector detection process 2" concerning Embodiment 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Coding apparatus, 110 ... Conversion part, 120 ... Quantization part, 130 ... Encoding part, 140 ... Inverse quantization part, 150 ... Inverse transformation part, 160 ... Restoration image memory, 170 ... Motion detection part, 171 ... Integer ... search unit, 172 ... best integer position determination unit, 173 ... surrounding integer position evaluation unit, 174 ... 1 / 2pel calculation unit, 175 ... search center determination unit, 176 ... 1 / 4pel calculation unit, 177 ... vector calculation unit, 180 ... Motion compensation unit

Claims (7)

A motion vector detection device for detecting a motion vector for a current block on a current frame by performing integer search and fractional search on a reference frame,
Results of the search of the integer unit, the difference between pixels of the reference frame and the pixel of the current block is an integer position is minimum as the first integer positions, among integer position near the periphery of the first integer position, the first a second integer position specifying means for specifying the difference is small integer position of the pixel and the pixel of the current block of the reference frame to the next integer position as the second integer position of,
A fractional position estimating means for estimating a fractional position with a predetermined accuracy based on the first integer position and the second integer position;
A fraction search means for performing a fraction search with higher accuracy than the predetermined accuracy based on the first integer position and the fraction position estimated by the fraction position estimation means;
With
The fraction search means includes:
Comparing the difference between the reference frame pixel at the first integer position and the current block pixel with the difference between the reference frame pixel at the fractional position estimated by the fractional position estimation means and the current block pixel; and A determination means for determining whether the position where the difference is small is either;
Search center designating means for designating a position determined by the determining means as having a smaller difference as a search center in a fractional search with higher accuracy than the predetermined accuracy;
The motion vector detecting device further comprising: The second integer position specifying means specifies a second integer position from integer positions adjacent to the first integer position;
The motion vector detection apparatus according to claim 1 , wherein: The second integer position specifying means specifies a second integer position from among the integer positions adjacent in the eight directions around the first integer position.
The motion vector detection apparatus according to claim 2 , wherein The fractional position estimating means estimates an intermediate position between the first integer position and the second integer position as the fractional position of the predetermined accuracy;
The motion vector detection device according to any one of claims 1 to 3 , wherein The second integer position specifying means, the difference between pixels of the first reference frame at integer positions pixel and the current block is, from the difference between the pixel of the pixel and the current block of the reference frame in the estimated fractional position If small, of the integer positions in the vicinity around the first integer position, the second integer positions the new difference is small integer positions of the pixels in the reference frame to the next pixel of the current block of the second integer positions And
Motion vector detection device according to any one of claims 1 to 4, characterized in that. A program for detecting a motion vector of a moving image on a computer,
On the computer,
An integer unit search is performed on the reference frame, and the integer position where the difference between the pixel of the reference frame and the pixel of the current block of the current frame is the minimum is set as the first integer position. of integer positions, a function of specifying the difference is small integer position of the pixel of the first next to the reference frame pixel and the current block of integer positions as a second integer positions,
A function of estimating an intermediate position between the first integer position and the second integer position as a fractional position with a predetermined accuracy;
The difference between the reference frame pixel at the first integer position and the current block pixel is compared with the difference between the reference frame pixel at the estimated fractional position and the current block pixel, and the difference is smaller. A function to determine whether or not
A function for performing a fractional search with higher accuracy than the predetermined accuracy, with the position determined by the determining means as having a smaller difference as a search center;
A function to calculate a motion vector for the current block based on the result of the fraction search;
A program characterized by realizing. In the computer,
The difference between the pixel and the current block pixels in the reference frame in said first integer position, when the pixels of the reference frame in the estimated fractional positions less than the difference between the pixel of the current block, the first integer position among integer positions around the vicinity, functions as a second integer position difference new a small integer position of the pixel of the second of the next reference frame of integer positions pixel and the current block,
The program according to claim 6 , further realizing the above.

Images