JP2006042114A - Motion vector detecting device, and detecting program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a motion vector detecting device which is used for motion compensation prediction encoding to highly efficiently encode animation information of an interlace structure, so as to reduce arithmetic throughput, heighten the speed of a processing, and raise efficiency in encoding, and to provide a motion vector detecting program. <P>SOLUTION: A motion vector search range control means 103 sets a frame motion vector search range to be normally smaller than a field motion vector search range with respect to a field motion vector detecting means 104 and a frame motion vector detecting means 105. The field motion vector detecting means 104 and the frame motion vector detecting means 105 respectively and independently detect a motion vector. A motion vector selecting means 106 compares a field motion vector with a frame motion vector, selects the motion vector with efficiency in encoding, having a small error evaluation value, as the motion vector of a motion vector detection object block, and then, outputs it to a motion vector output means 107. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a motion vector detection apparatus and a motion vector detection program for detecting a motion vector necessary for motion compensation prediction of an interlaced structure image that is transmitted by dividing a frame image into two field images.

  In compression coding of a moving image typified by MPEG, motion compensation predictive coding that compresses a code amount using a correlation between consecutive images is often used. Motion compensation predictive coding is performed by dividing an image into blocks of 16 × 16 pixels, for example, and sequentially setting each block as a motion detection target block.

  Motion compensated predictive coding requires a motion vector that indicates how much the corresponding block area in the reference image used for coding the motion detection target block is shifted. is there. When the motion vector that minimizes the error evaluation value between the motion detection target block and the corresponding block of the reference image is detected and used for motion compensation prediction, the coding efficiency becomes the best.

The error evaluation value expresses the degree of difference between corresponding blocks, and is obtained by block matching.
The block matching is a technique for obtaining an error evaluation value by calculating a matching evaluation function by matching a motion detection target block with an image within a detection range in a reference image at every position in pixel units. As a matching evaluation function, for example, for all pixels in a block, a sum of absolute values of pixel value differences for each pixel (absolute value difference sum), and a difference value of pixel values for each pixel is squared. A sum (sum of squares of difference), a code amount of a motion vector calculated and added to the absolute difference sum or the sum of squares of difference, and a code amount of a DCT coefficient calculated or predicted are used.

  There is a full search method as an example of a motion vector detection method with the smallest error evaluation value. In the full search method, a motion that gives the smallest error evaluation value by checking the error evaluation value obtained by performing block matching on all positions within a predetermined search range on the reference image corresponding to the motion detection target block. A vector is detected. In the full search method, a motion vector having the smallest error evaluation value can be detected within the search range, but there is a problem that a huge amount of calculation is required.

One technique for reducing the amount of calculation for motion vector detection is a hierarchical motion vector detection method. The detection method will be described below with reference to FIG.
In the reduced image 701 with reduced resolution, the motion vector 703 of the reduced image corresponding block 702 is obtained, and the obtained motion vector 703 is expanded based on the reduction ratio of the reduced image 701 so as to correspond to the original image 704. A hierarchical motion vector 706 is obtained. Then, in the original image 704, the hierarchical motion vector 706 is set as the search initial point 707, the motion vector search range 708 is determined based on the search initial point 707, and the motion vector is searched in the motion vector search range 708. is there.

  Here, the reduced image corresponding block 702 is obtained by associating the encoded block 705 in the original image 704 with the reduced image 701. Since this motion vector detection apparatus can obtain the hierarchical motion vector 706 as a search initial point with a small amount of computation, the computation amount is reduced, but the hierarchical motion vector 706 that can capture only global motion is the true motion. If they are far apart, there is a problem that the search cannot catch up and the coding efficiency is lowered.

  Another method for reducing the amount of motion vector detection is a peripheral block reference method. The peripheral block reference method searches for a motion vector using at least one of the motion vectors of the peripheral blocks as a search initial point. As shown in FIG. 8, the peripheral block is a block for which a motion vector has already been obtained among blocks 801 to 804 and blocks 806 to 809 adjacent to the encoding block 805 to be encoded.

In general, the order in which motion vectors are obtained is blocks 801, 802, 803,.
Since the order is 9, when the motion vector of the coding block 805 is searched, the blocks whose motion vectors have already been obtained are 801 to 804. Since this motion vector detection device does not require the amount of computation for detecting the initial search point, the amount of computation is reduced. However, when a new motion that does not exist in the neighboring blocks occurs, the search cannot catch up and the coding efficiency is increased. There is a problem that decreases.

  In order to solve the problems of the hierarchical motion vector detection method and the peripheral block reference method, there is a motion vector detection method that combines the hierarchical motion vector detection method and the peripheral block reference method. This motion vector detection method uses the motion vector obtained by the hierarchical motion vector detection method and the motion vector obtained by the peripheral block reference method as the motion vector, which has a smaller error evaluation value.

  In this motion vector detection method, when the wrong motion vector is obtained in the reduced image, which is a problem of the hierarchical motion vector detection method, the motion vector obtained by the peripheral block reference method is adopted. Coding efficiency does not decrease. Also, when a new motion that does not exist in the peripheral block, which is a problem of the peripheral block reference method, is generated, the motion vector obtained from the hierarchical motion vector is adopted, so that the encoding efficiency does not decrease. Thus, by combining the hierarchical motion vector detection method and the peripheral block reference method, the problems of the respective motion vector detection methods can be solved.

  However, the calculation amount of this motion vector detection method involves the calculation amount by the hierarchical motion vector detection method and the calculation amount by the peripheral block reference method, which increases the calculation amount.

  By the way, in an image having an interlace structure in which two field images are overlapped to form one frame image, two frames located at the same spatial coordinates in the frame / field of the encoding target image are used in order to improve encoding efficiency. A method is used in which a motion vector is detected in each of the motion vector detection target blocks, and a smaller error evaluation value is selected. Therefore, since it is necessary to detect a motion vector in each frame / field, the amount of motion vector detection processing for an interlaced image is about twice that of a progressive structure image that transmits one image as it is. Cost.

As a conventional example 1, an interlaced image motion vector detection device using a combination of a hierarchical motion vector detection method and a peripheral block reference method will be described with reference to FIG.
As shown in FIG. 2, the motion vector detection apparatus of the conventional example 1 includes a reference image memory 201, an encoding target image memory 202, a field motion vector detection means 204, a frame motion vector detection means 205, and a motion vector selection. Means 206, motion vector output means 207, peripheral block field motion vector memory 208, and peripheral block frame motion vector memory 209.

The reference image memory 201 stores reference image data.
The encoding target image memory 202 stores encoding target image data.
The field motion vector detection unit 204 detects a field motion vector, and supplies the detected field motion vector and the corresponding error evaluation value to the motion vector selection unit 206.

Hereinafter, the procedure of the field motion vector detection process will be described.
First, using the hierarchical motion vector detection method, a certain fixed range (for example, ± 16 pixels centered on the origin) is searched to obtain a hierarchical field motion vector of a field motion vector detection target block.

Next, an error evaluation value EhFi in the hierarchical field motion vector is obtained.
Next, error evaluation values EsFi1 to EsFi4 are obtained using peripheral block field motion vectors corresponding to blocks at positions corresponding to 801 to 804 in FIG. 8 stored in the peripheral block field motion vector memory 108.

  Next, a point having a minimum error evaluation value among EhFi and EsFi1 to EsFi4 is set as a search initial point. A field motion vector having a minimum error evaluation value is detected in an area of, for example, ± 2 pixels around the determined initial search point. Then, the detected field motion vector and the corresponding error evaluation value are supplied to the motion vector selection means 206.

  The frame motion vector detection unit 205 detects a frame motion vector and supplies the detected frame motion vector and an error evaluation value corresponding thereto to the motion vector selection unit 206.

  First, using the hierarchical motion vector detection method, a fixed range (for example, ± 16 pixels centered on the origin) is searched to obtain a hierarchical frame motion vector of a frame motion vector detection target block.

Next, an error evaluation value EhFl in the hierarchical frame motion vector is obtained.
Next, error evaluation values EsFl1 to EsFl4 are obtained using peripheral block frame motion vectors corresponding to blocks at positions corresponding to 801 to 804 in FIG. 8 stored in the peripheral block frame motion vector memory 109.

  Next, a point having the minimum error evaluation value in EhFl and EsFl1 to EsFl4 is set as a search initial point. A frame motion vector that minimizes the error evaluation value is detected in an area of, for example, ± 2 pixels centering on the determined initial search point. Then, the detected frame motion vector and the corresponding error evaluation value are supplied to the motion vector selection means 206.

  The motion vector selection unit 206 receives the field motion vector and the corresponding error evaluation value from the field motion vector detection unit 204, and receives the frame motion vector and the corresponding error evaluation value from the frame motion vector detection unit 205. The smaller motion vector is supplied to the motion vector output means 207.

The motion vector output means 207 outputs the determined motion vector to the outside.
In the conventional example 1, by using a combination of the hierarchical motion vector detection method and the peripheral block reference method, the processing amount of the motion vector in the interlaced image is compared with the processing amount when the full search method is used. Although reduced, since frame motion vector detection and field motion vector detection are performed by the same equivalent process, the processing amount of frame motion vector detection is required to be approximately the same as the processing amount of field motion vector detection. Therefore, there is a problem that the processing amount of motion vector detection for an interlaced image requires about twice as much as the processing amount of motion vector detection for a progressive image.

  In order to reduce the processing amount of motion vector detection of an interlaced image, for example, a motion vector detection device as in Patent Document 1 has been proposed. This motion vector detection device detects a field motion vector prior to a frame motion vector, and in the frame motion vector detection, an already detected field motion vector is set as a search initial point, and a frame motion vector detection range is set around the search initial point. By making the area narrow, the frame motion vector detection processing amount is reduced, and as a result, the processing amount required for motion vector detection of the entire interlaced image is reduced.

Hereinafter, as a conventional example 2, a motion vector detection device described in Patent Document 1 will be described with reference to FIG.
As shown in FIG. 3, the motion vector detection device of the conventional example 2 includes a reference image memory 301, an encoding target image memory 302, a field motion vector detection means 304, a frame motion vector detection means 305, and a motion vector selection. It comprises means 306 and motion vector output means 307.

The reference image memory 301 stores reference image data.
The encoding target image memory 302 stores image data to be actually encoded.

  The field motion vector detection unit 304 detects the field motion vector of the motion detection target block on the encoding target image data using the reference image data, and the detected field motion vector is sent to the frame motion vector detection unit 305 or The field motion vector and the corresponding error evaluation value are supplied to the motion vector selection means 306.

  The frame motion vector detection unit 305 detects a frame motion vector from the reference image data and the encoding target image data using the field motion vector input from the field motion vector detection unit 304 as a search initial point, and detects the detected frame motion vector. And an error evaluation value corresponding thereto are supplied to the motion vector selection means 306. As the frame motion vector, a frame motion vector having a minimum error evaluation value is detected in an area of, for example, ± 2 pixels centering on the search initial point.

Here, the operation of the frame motion vector detection means 305 will be described with reference to the flowchart of FIG.
First, the field motion vector input from the field motion vector detection unit 304 is set as a search initial point (S101). Next, a frame motion vector having a minimum error evaluation value is detected in an area of, for example, ± 2 pixels centering on the search initial point (S102). Then, the detected frame motion vector and the corresponding error evaluation value are supplied to the motion vector selection means 306 (S103).

Returning to the description of FIG.
The motion vector selection unit 306 receives the field motion vector and the corresponding error evaluation value from the field motion vector detection unit 304, and receives the frame motion vector and the corresponding error evaluation value from the frame motion vector detection unit 305. The smaller motion vector is supplied to the motion vector output means 307.

The motion vector output means 307 outputs the determined motion vector to the outside.
In the conventional example 2, the field motion vector is used as an initial search point for frame motion vector detection, so that the frame search range can be narrowed, so that the processing amount of the frame motion vector can be reduced.

  However, since the field motion vector is used as the initial search point when obtaining the frame motion vector, if an incorrect motion vector is detected in the field motion vector detection, the frame motion vector also detects the incorrect motion vector. Therefore, the error evaluation value does not become small. Therefore, there exists a subject that encoding efficiency will fall.

In addition, since the frame motion vector must be detected after detecting the field motion vector, there is a problem that the processing speed cannot be increased because the parallel processing of the field motion vector detection and the frame motion vector detection cannot be performed.
"JP-A-7-107485"

  In the motion vector detection apparatus that performs frame motion vector detection and field motion vector detection by the same processing as shown in the conventional example 1, the processing amount of frame motion vector detection needs to be about the same as the processing amount of field motion vector detection. It becomes. Therefore, there is a problem that the processing amount of motion vector detection for an interlaced image requires about twice as much as the processing amount of motion vector detection for a progressive image.

  Further, if the motion vector detecting device shown in the conventional example 2 is used, the processing amount of the frame motion vector can be reduced. However, since the field motion vector is used as the initial search point when obtaining the frame motion vector, if an incorrect motion vector is detected in the field motion vector detection, the frame motion vector also detects the incorrect motion vector. Therefore, the error evaluation value does not become small. Therefore, there exists a subject that encoding efficiency will fall.

  In addition, since the frame motion vector must be detected after detecting the field motion vector, there is a problem that the processing speed cannot be increased because the parallel processing of the field motion vector detection and the frame motion vector detection cannot be performed.

The present invention has been made in view of the above problems.
The present invention sets the frame motion vector detection search range to be always smaller than the field motion vector detection search range and performs motion vector detection, thereby reducing the frame motion vector detection processing amount without impairing the accuracy of frame motion vector detection. The purpose is to reduce the amount of motion vector detection processing of the entire interlaced image.

  In addition, the present invention processes the frame motion vector detection and the field motion vector detection independently, so that even if an erroneous motion vector is detected in the field motion vector detection, the frame motion vector detection is not affected. It is another object of the present invention to improve coding efficiency by detecting an accurate motion vector without reducing the error and reducing the error evaluation value.

  Furthermore, another object of the present invention is to enable parallel processing of motion vector detection by independently processing frame motion vector detection and field motion vector detection, thereby realizing high-speed processing.

Therefore, in order to solve the above problems, the present invention provides the following apparatus and program.
(1) In a motion vector detection device for detecting a motion vector used for encoding an interlaced structure image that constitutes one frame image by superimposing two temporally continuous field images,
The field input image is divided into blocks for a predetermined number of pixel units, each block in the field input image is sequentially set as a field motion detection target block, and a search range is set in the field reference image according to each field motion detection target block A field motion vector that compares a pixel in each field motion detection target block with a pixel in the search range in the field reference image and obtains a field motion vector for each field motion detection target block by block matching Detection means;
A frame input image is divided into blocks of a predetermined number of pixels, each block is sequentially set as a frame motion detection target block in the frame input image, and a search range is set according to each frame motion detection target block in a frame reference image A frame motion vector that compares a pixel in each frame motion detection target block with a pixel in the search range in the frame reference image to obtain a frame motion vector for each frame motion detection target block by block matching Detection means;
The field motion vector and the frame motion vector detected in the field motion detection target block and the frame motion detection target block located at the same spatial coordinates in the encoding target image are respectively compared, and the encoding efficiency is compared. A motion vector selection means for selecting and outputting a motion vector having a smaller error evaluation value, which is a degree indicating
A motion vector detection apparatus comprising: search range setting means for setting the search range for detecting the frame motion vector as a range that is always narrower than the search range for detecting the field motion vector.
(2) In a motion vector detection program for detecting a motion vector used for encoding an interlaced structure image that constitutes one frame image by superimposing two temporally continuous field images,
The field input image is divided into blocks for a predetermined number of pixel units, each block in the field input image is sequentially set as a field motion detection target block, and a search range is set in the field reference image according to each field motion detection target block A field motion vector that compares a pixel in each field motion detection target block with a pixel in the search range in the field reference image and obtains a field motion vector for each field motion detection target block by block matching Detection function,
A frame input image is divided into blocks of a predetermined number of pixels, each block is sequentially set as a frame motion detection target block in the frame input image, and a search range is set according to each frame motion detection target block in a frame reference image A frame motion vector that compares a pixel in each frame motion detection target block with a pixel in the search range in the frame reference image to obtain a frame motion vector for each frame motion detection target block by block matching Detection function,
The field motion vector and the frame motion vector detected in the field motion detection target block and the frame motion detection target block located at the same spatial coordinates in the encoding target image are respectively compared, and the encoding efficiency is compared. A motion vector selection function for selecting and outputting a motion vector with a smaller error evaluation value, which is a degree indicating
A motion vector detection program for causing a computer to realize a search range setting function for setting the search range for detecting the frame motion vector as a range that is always narrower than the search range for detecting the field motion vector .

  According to the present invention, frame motion vector detection is performed without degrading the accuracy of frame motion vector detection by setting the search range for frame motion vector detection to be always smaller than the search range for field motion vector detection. The processing amount can be reduced, and the motion vector detection processing amount of the entire interlaced image can be reduced.

  Also, by independently processing frame motion vector detection and field motion vector detection, even if an erroneous motion vector is detected in field motion vector detection, frame motion vector detection is not affected. Since accurate motion vector detection is possible, the error evaluation value can be reduced, and the encoding efficiency can be improved.

  Furthermore, by independently processing the frame motion vector detection and the field motion vector detection, the motion vector detection can be performed in parallel, and the motion vector detection processing can be speeded up.

In the present invention, characteristics at the time of motion vector detection in an interlaced image described below are used.
First, a detailed method for detecting a field motion vector will be described with reference to FIG. As shown in the figure, the field image is composed of a pair of a top field and a bottom field.

  Therefore, the motion vector needs to be detected in each of the top field and the bottom field. A pair of the detected two motion vectors forms a field motion vector, which is used in subsequent motion compensation prediction processing.

The top field motion vector is detected by the following procedure.
The top field of the encoding target image is compared with the top field and the bottom field of the reference image to detect two motion vectors of top-t and top-b, respectively.

The detected top-t and top-b error evaluation values are compared, and the smaller one is used as the top field motion vector.
The bottom field motion vector is detected by the following procedure.

  By comparing the bottom field of the encoding target image with the top field and bottom field of the reference image, two motion vectors of bottom-t and bottom-b are detected.

The detected bottom-t and bottom-b error evaluation values are compared, and the smaller one is used as the bottom field motion vector.
The error evaluation value for comparison with the frame motion vector is obtained by adding the error evaluation values of the top field motion vector and the bottom field motion vector detected above.

  Here, consider the case of an image with a large movement. In an image with a large movement, the correlation between the images becomes small and the error evaluation value becomes large even in a short time. Therefore, since a difference appears even in a short time between fields, the error evaluation value becomes smaller as the time between images used for motion vector detection is shorter.

  Consider the top field motion vector. The top-b detected between the top field of the encoding target image and the bottom field of the reference image is used for motion vector detection than the top-t detected between the top field of the encoding target image and the top field of the reference image. The time between field images is short. Therefore, there is a high possibility that the top-b side is selected as the top field motion vector.

  Consider a bottom field motion vector. The bottom-b detected between the bottom field of the encoding target image and the bottom field of the reference image is used for motion vector detection than the bottom-t detected between the bottom field of the encoding target image and the top field of the reference image. The time between field images is short. Therefore, there is a high possibility that the bottom-b side is selected as the bottom field motion vector.

  Consider a case where the field motion vector and the frame motion vector of the same encoding target image are compared. At this time, in the top-b which is the top field motion vector, the time between the field images used for the top field motion vector detection is shorter than the time between the frame images used for the frame motion vector detection. Also, bottom-b, which is a bottom field motion vector, has a time between field images used for bottom field motion vector detection equal to a time between frame images used for frame motion vector detection.

  Therefore, when compared in total, it can be considered that the field motion vector composed of a pair of the top field motion vector and the bottom field motion vector has a shorter time between images used for motion vector detection than the frame motion vector. Therefore, in a large motion image, the field motion vector is more likely to have a smaller error evaluation value than the frame motion vector, and the probability that the field motion vector is selected as the motion vector of the motion detection target block is higher. Is expensive.

  On the other hand, in the case of an image that does not move too much, the decrease in correlation over time is not a problem for a short time between fields. Rather, the magnitude of the error evaluation value depends on the characteristics of the picture pattern. Will be determined. The motion vector having the smallest error evaluation value is selected as the motion vector of the final motion detection target block. Therefore, in an image that does not have a very large motion, the field motion vector is generally selected as the motion vector of the motion detection target block if it is more likely that the error evaluation value is smaller than the frame motion vector. It is not necessarily true that the probability that the image is higher than the field motion vector and the frame motion vector.

  In addition, in the case of an image with a large motion, if the motion vector search range in the motion detection target block is set wider, it is possible to prevent erroneous detection because the motion vector is outside the search range. desirable. On the other hand, in the case of an image that does not have a very large motion, even if the motion vector search range is set wider than necessary, the accuracy does not change compared to the case where it is set narrower. Further, since the amount of processing required for the search increases, it is desirable to set the search range as narrow as possible.

  Therefore, in the case of an interlaced image with a large motion, it is necessary to set a wide motion vector search range, and in this case, as described above, the field motion vector has an error evaluation value more than the frame motion vector. There is a high probability that the field motion vector is selected as the motion vector of the motion detection target block.

  On the other hand, in the case of an interlaced image that does not move much, the motion vector search range may be narrow. In this case, it cannot be said that the probability that a field motion vector is selected is high, and the motion vector having the smaller error evaluation value is selected as the motion vector of the motion detection target block. The

Thus, the frame motion vector may be selected in the case of an image that does not have a large amount of motion, and is unlikely to be selected in the case of an image that has a large amount of motion.
Therefore, even if the frame motion vector search range is set to be as wide as the field motion vector search range in order to increase the motion vector selection accuracy for an image with large motion, it is selected as the motion vector of the motion detection target block. Therefore, searching for a frame motion vector search region exceeding a predetermined range is not effective. Therefore, even if the frame motion vector search range is always set narrower than the field motion vector search range, the frame motion vector search range is almost the same as the field motion vector search range. An unchanging motion vector can be detected.

  Therefore, according to the present invention, the frame motion vector detection search range is always set to be smaller than the field motion vector detection search range and the motion vector detection is performed, so that the frame motion vector detection accuracy is not impaired. The amount of detection processing can be reduced, and the amount of motion vector detection processing for the entire interlaced image can be reduced.

Embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram of an embodiment of a motion vector detection apparatus of the present invention.
In this embodiment, the reference image memory 101, the encoding target image memory 102, the motion vector search range control means 103, the field motion vector detection means 104, the frame motion vector detection means 105, and the motion vector selection means 106 , A motion vector output means 107, a peripheral block field motion vector memory 108, and a peripheral block frame motion vector memory 109.

The reference image memory 101 stores reference image data.
The encoding target image memory 102 stores encoding target image data.
The motion vector search range control means 103 determines a frame motion vector search range that is always narrower than the field motion vector search range, and supplies each search range to the field motion vector detection means 104 and the frame motion vector detection means 105. It is. For example, when the field motion vector search range is ± 16 pixels centered on the origin, the frame motion vector search range is ± 8 pixels centered on the origin.

  The field motion vector detection means 104 detects a field motion vector within the search range set by the motion vector search range control means 103, and detects the detected field motion vector and the corresponding error evaluation value as a motion vector selection means. 106 is supplied.

Here, the operation of the field motion vector detection means 104 will be described with reference to the flowchart of FIG.
First, the motion vector search range control means 103 sets the field motion vector detection search range to ± 16 pixels, for example (S201).

Next, a hierarchical motion vector is obtained using the set field motion vector detection search range as a search range (S202).
A hierarchical motion vector detection method will be described with reference to FIG. In the reduced image 701 with reduced resolution, the motion vector 703 of the reduced image corresponding block 702 is obtained, and the obtained motion vector 703 is expanded in consideration of the reduction rate of the reduced image 701, whereby the original image 704 is restored. A corresponding hierarchical motion vector 706 is obtained. Here, the reduced image corresponding block 702 is obtained by associating the encoded block 705 in the original image 704 with the reduced image 701.

Returning to the flowchart of FIG.
Next, error evaluation values EhFi with hierarchical motion vectors and error evaluation values EsFi1 to EsFi4 with peripheral motion vectors are obtained (S203). Here, the peripheral motion vector is a block whose motion vector has already been determined among the blocks 801 to 804 and 806 to 809 adjacent to the encoding block 805 to be encoded as shown in FIG. And In general, the order of obtaining the motion vector is 801 → 809, and therefore the blocks having the motion vector are 801 to 804.

  Next, a point having a minimum error evaluation value among EhFi and EsFi1 to EsFi4 is set as a search initial point (S204). A field motion vector having a minimum error evaluation value is detected in an area of, for example, ± 2 pixels around the determined initial search point (S205). Here, the search method is ± 2 pixels centered on the initial search point. However, any search method using the initial search point may be used.

Then, the detected field motion vector and the corresponding error evaluation value are supplied to the motion vector selection means 106 (S206).
Returning to the description of the block diagram of FIG.

  The frame motion vector detection means 105 detects a frame motion vector within the search range set by the motion vector search range control means 103, and detects the detected frame motion vector and the corresponding error evaluation value as a motion vector selection means. 106 is supplied.

Here, the operation of the frame motion vector detection means 105 will be described with reference to the flowchart of FIG.
First, the motion vector search range control means 103 sets the frame motion vector detection search range to ± 8 pixels, for example (S301). Note that the search range for frame motion vector detection is ± 8 pixels, which is narrower than the search range for ± 16 pixels for field motion vector detection. Thereby, the processing amount of frame motion vector detection can be reduced from the processing amount of field motion vector detection.

Next, using the set frame motion vector detection search range as a motion vector search range control means, a hierarchical motion vector is obtained (S302).
Subsequently, error evaluation values EhFl with hierarchical motion vectors and error evaluation values EsFl1 to EsFl4 with peripheral motion vectors are obtained (S303).

  Next, a point having a minimum error evaluation value among EhFl and EsFl1 to EsFl4 is set as a search initial point (S304). For example, a frame motion vector having a minimum error evaluation value is detected in an area of ± 2 pixels centering on the determined initial search point (S305). Here, the search method is ± 2 pixels centered on the initial search point. However, any search method using the initial search point may be used.

Then, the detected frame motion vector and the corresponding error evaluation value are supplied to the motion vector selection means 106 (S306).
Returning to the description of the block diagram of FIG.

  The motion vector selection unit 106 receives the field motion vector and the corresponding error evaluation value from the field motion vector detection unit 104, and receives the frame motion vector and the corresponding error evaluation value from the frame motion vector detection unit 105. A motion vector having a small evaluation value is supplied to the motion vector output means 107.

The motion vector output means 107 outputs the determined motion vector to the outside.
The peripheral block field motion vector memory 108 stores the peripheral motion vector of the field of the encoded block.

The peripheral block frame motion vector memory 109 stores the peripheral motion vector of the frame of the encoded block.
Furthermore, the present invention includes a program for causing a computer to realize the functions of the motion vector detection device described above. These programs may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

It is a block diagram which shows the Example of the motion vector detection apparatus of this invention. It is a block diagram of the conventional motion vector detection apparatus. It is a block diagram of the other conventional motion vector detection apparatus. 10 is a flowchart of conventional frame motion vector detection means 204. It is a flowchart of the field motion vector detection means 104 of a present Example. It is a flowchart of the frame motion vector detection means 105 of a present Example. It is a figure of the conventional hierarchical motion vector detection. It is a figure of the conventional peripheral block. It is explanatory drawing of the conventional field structure.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... Reference image memory 102 ... Encoding object image memory 103 ... Motion vector search range control means 104 ... Field motion vector detection means 105 ... Frame motion vector detection means 106 ... Motion vector Selection means 107 ... motion vector output means 108 ... peripheral block field motion vector memory 109 ... peripheral block frame motion vector memory

Claims (2)

In a motion vector detection device that detects a motion vector used to encode an image of an interlace structure that constitutes one frame image by superimposing two temporally continuous field images,
The field input image is divided into blocks for a predetermined number of pixel units, each block in the field input image is sequentially set as a field motion detection target block, and a search range is set in the field reference image according to each field motion detection target block A field motion vector that compares a pixel in each field motion detection target block with a pixel in the search range in the field reference image and obtains a field motion vector for each field motion detection target block by block matching Detection means;
A frame input image is divided into blocks of a predetermined number of pixels, each block is sequentially set as a frame motion detection target block in the frame input image, and a search range is set according to each frame motion detection target block in a frame reference image A frame motion vector that compares a pixel in each frame motion detection target block with a pixel in the search range in the frame reference image to obtain a frame motion vector for each frame motion detection target block by block matching Detection means;
The field motion vector and the frame motion vector detected in the field motion detection target block and the frame motion detection target block located at the same spatial coordinates in the encoding target image are respectively compared, and the encoding efficiency is compared. A motion vector selection means for selecting and outputting a motion vector having a smaller error evaluation value, which is a degree indicating
A motion vector detection apparatus comprising: search range setting means for setting the search range for detecting the frame motion vector as a range that is always narrower than the search range for detecting the field motion vector. In a motion vector detection program for detecting a motion vector used to encode an image of an interlace structure that constitutes one frame image by superimposing two temporally continuous field images,
The field input image is divided into blocks for a predetermined number of pixel units, each block in the field input image is sequentially set as a field motion detection target block, and a search range is set in the field reference image according to each field motion detection target block A field motion vector that compares a pixel in each field motion detection target block with a pixel in the search range in the field reference image and obtains a field motion vector for each field motion detection target block by block matching Detection function,
A frame input image is divided into blocks of a predetermined number of pixels, each block is sequentially set as a frame motion detection target block in the frame input image, and a search range is set according to each frame motion detection target block in a frame reference image A frame motion vector that compares a pixel in each frame motion detection target block with a pixel in the search range in the frame reference image to obtain a frame motion vector for each frame motion detection target block by block matching Detection function,
The field motion vector and the frame motion vector detected in the field motion detection target block and the frame motion detection target block located at the same spatial coordinates in the encoding target image are respectively compared, and the encoding efficiency is compared. A motion vector selection function for selecting and outputting a motion vector with a smaller error evaluation value, which is a degree indicating
A motion vector detection program for causing a computer to realize a search range setting function for setting the search range for detecting the frame motion vector as a range that is always narrower than the search range for detecting the field motion vector .

Images