JP2000165883A - Motion vector detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the power consumption by decreasing an unnecessary arithmetic operation in the case of detecting a motion vector with accuracy in the unit of a half pixel. SOLUTION: A border discrimination section 15 gives a command to an address control section 16 so as not to generate an address for pixel data at the outside of a border when the border discrimination section 15 discriminates that a retrieval area for detecting a half pixel precision motion vector includes an area at the outside of a frame or slice border. An interpolation pixel generating section 13 reads pixel data from a retrieval area pixel data memory 11 on the basis of an address generated by the address control section 16 to generate only required interpolation pixels spaced by a half pixel interval.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for detecting a motion vector used in motion compensation coding, which is a technique for compression coding moving picture data.

[0002]

2. Description of the Related Art In order to transfer and store a moving image signal using a communication line having a small capacity, an image compression technique for reducing the data amount of an image signal having a huge amount of data is indispensable. Generally, an image signal contains high redundancy (information that can be reduced), and there is an inter-frame predictive coding method as one of the methods for reducing the redundancy. However, this method can achieve high image code compression for an image with a small motion, but has a small compression effect on an image with a large motion because the correlation between frames is low. As a method for solving this problem, there is a motion compensation predictive coding method using a motion vector. This motion vector can be detected by a block matching method that searches for a block having the highest correlation with the block to be encoded in the reference image. In the block matching method, generally, a cumulative sum of absolute differences and a cumulative sum of square errors are used as evaluation values of the degree of correlation.

As a coding method of image information, MPEG
(Moving Picture coding Experts Group) is known. In MPEG, motion compensation prediction with half-pixel accuracy (0.5-pixel accuracy) is performed by interpolating between pixels of a reference image to improve prediction accuracy. The half-pixel precision motion compensation has a function of loosely blurring an image as well as improving prediction accuracy.
For example, a motion vector detecting device disclosed in Japanese Patent Application Laid-Open No. 7-95585 detects such a motion vector with half-pixel accuracy.

[0004] One of the half-pixel precision motion vector detections by the block matching method is to move a candidate block in a predetermined search area in a reference image on a pixel-by-pixel basis and to select a candidate block having the highest correlation with an encoded block. A method of obtaining a block (best match block with integer pixel precision), detecting a motion vector with integer pixel precision, and moving this candidate block in half pixel units in the vicinity thereof to detect a motion vector with half pixel precision. is there.

In this method, after detecting a motion vector with integer pixel precision, the best match block with integer pixel precision and pixel data adjacent to the best match block are used to interpolate interpolated pixels at half pixel intervals inside and around this block. Generate Then, an area obtained by enlarging this block by half pixel width outside each side becomes a search area for detecting a half-pixel precision motion vector. FIG. 3 shows an interpolated pixel (half pixel) H for an integer pixel P when the size of a candidate block is 3 pixels × 3 pixels.
3 shows an example of a search area S for detecting a half-pixel accuracy motion vector.

As shown in FIG. 3, after a search with integer pixel precision, if an interpolated pixel H is generated in a search area S for detecting a half pixel precision motion vector, a half pixel precision motion vector can be detected. it can.

[0007]

However, the conventional half-pixel accuracy motion vector detecting apparatus has the following problems.

When a half-pixel precision motion vector is obtained by the block matching method, if the search area for detecting a half-pixel precision motion vector includes an area outside a frame boundary or an area outside a specific slice boundary, the detection is performed. There are motion vectors that cannot be used. Of course, there is no need to generate an interpolated pixel for detecting such a motion vector. Nevertheless, the conventional half-pixel accuracy motion vector detecting apparatus has a problem that such an unnecessary calculation for generating an interpolated pixel is also performed.

In this case, the difference absolute value calculation unit as the correlation degree calculation unit in the block matching method has a plurality of difference absolute value accumulators corresponding to each of the half-pixel precision motion vectors that can be detected. Is
There is a problem that the absolute difference accumulator corresponding to the motion vector that cannot be detected also operates and consumes power.

SUMMARY OF THE INVENTION It is an object of the present invention to prevent a calculation relating to an undetectable motion vector from being performed when detecting a motion vector with half-pixel accuracy, and to suppress the time required for the calculation and the power consumption.

[0011]

Means for Solving the Problems To solve the above-mentioned problems, a means according to the present invention is a motion vector detecting device for detecting a motion vector with half-pixel accuracy, which is used for encoding in an encoding target image. A coded block pixel data memory for storing pixel data of the block, a search area pixel data memory for storing pixel data of the reference image, and a half-pixel precision motion vector detection based on a motion vector of integer pixel precision. In a search area, an interpolation pixel generation unit that generates an interpolation pixel using the pixel data of the search area pixel data memory; an interpolation pixel data generated by the interpolation pixel generation unit; and a pixel of the encoding block pixel data memory. A difference absolute value calculation unit for calculating a cumulative sum of a difference absolute value with the data. Further, a boundary determining unit that determines whether the search area for detecting the half-pixel accuracy motion vector includes an area outside the boundary of the reference image, and a search area for detecting the half-pixel accuracy motion vector by the boundary determination unit. When it is determined that includes an area outside the boundary, an address control unit that controls a read address of pixel data from the search area pixel data memory so that interpolation pixels outside the boundary are not generated. It was done.

According to a second aspect of the present invention, in the motion vector detecting device according to the first aspect, a boundary of the reference image is defined by a frame.

[0013] Further, the measures taken by the invention of claim 3 are as follows:
2. The motion vector detecting device according to claim 1, wherein a boundary of the reference image is defined by a slice.

According to the first to third aspects of the present invention, the boundary judging section determines whether or not the search area for detecting a half-pixel accuracy motion vector includes an area outside the frame boundary of the reference image or outside the boundary of a specific slice. That is, it is determined whether there is a half-pixel precision motion vector that cannot be detected, and the determination result is given to the address control unit. Then, the address control unit controls the pixel data read address of the search area pixel data memory storing the pixel data of the reference image so as not to read data outside the boundary,
The pixel data outside the boundary is not provided to the interpolation pixel generation unit, and the calculation for unnecessary interpolation pixel generation is not performed. Therefore, it is possible to reduce the number of operation cycles for detecting a motion vector with half-pixel accuracy.

The measures taken by the invention of claim 4 are as follows:
The motion vector detecting device according to claim 1, wherein the absolute difference value calculating unit includes a plurality of absolute difference value accumulators, and the search area for detecting a half-pixel accuracy motion vector has a boundary. A clock control unit that controls to stop supplying a clock to a difference absolute value accumulator relating to half-pixel precision motion vector detection that requires an interpolated pixel outside the boundary when it is determined to include an outside region. It is what it was.

According to the fourth aspect of the present invention, the clock supply to the difference absolute value accumulator corresponding to the undetectable half-pixel precision motion vector is stopped, so that the operation of the difference absolute value accumulator is stopped. In addition, power consumption related to half-pixel accuracy motion vector detection can be reduced.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration of a motion vector detecting device according to the present invention. The search area pixel data memory 11 stores the data of the pixels of the reference image, which are necessary for the half-pixel accuracy motion vector detection. The encoding block pixel data memory 12 stores the pixel data of the encoding block in the encoding target image. The interpolation pixel generation unit 13 receives pixel data from the search area pixel data memory 11 and generates interpolation pixels at half pixel intervals. The difference absolute value calculation unit 14 serving as a correlation degree calculation unit calculates interpolation pixel data in the candidate block of the reference image generated by the interpolation pixel generation unit 13 and pixels of the coding block read from the coding block pixel data memory 12. The absolute value of the difference with the data is integrated over the entire block to obtain a cumulative sum. The boundary determination unit 15 issues a command to the address control unit 16 in accordance with the relationship between the search area for half-pixel accuracy motion vector detection and the frame boundary of the reference image. The address control unit 16 generates an address of the pixel data of the reference image necessary for generating the interpolation pixel, and supplies the generated address to the search area pixel data memory 11. According to this address,
The pixel data is sent to the interpolation pixel generation unit 13.

FIG. 2 is a diagram showing a configuration of the absolute difference calculating section 14. As shown in FIG. The difference absolute value calculator 14 has a plurality of difference absolute value accumulators 20. Each difference absolute value accumulator 20
A difference absolute value calculator 21 for calculating a difference absolute value between the interpolation pixel data in the candidate block obtained from the reference image and the pixel data of the coding block; a register 23 for storing the accumulated value; And an adder 22 for adding the absolute value and the accumulated value.
Reference numeral 23 operates in synchronization with an externally applied clock CLK.

The size of the coding block is 16 pixels × 1
In many cases, the number of pixels is 6 pixels.
A case where a motion vector is detected with half-pixel accuracy for a block having a size of 3 × 3 pixels will be described with reference to FIGS.

First, a motion vector of a coded block is detected with integer pixel precision. By block matching method,
A candidate block having a size of 3 pixels × 3 pixels is moved in units of one pixel in a predetermined area in the reference image, and a block having the highest correlation with the coding block in the coding target image, that is, a block of integer pixel precision. Find the best match block B.
At this time, a motion vector with integer pixel precision is obtained from the position of the best match block B with integer pixel precision relative to the position of the coding block.

Next, the motion vector of the encoded block is detected with half-pixel accuracy. FIG. 3 shows eight kinds of motion vectors Va to Vh to be detected as a sum with a motion vector with integer pixel precision, an integer pixel P read from the search area pixel data memory 11, and an interpolation pixel generation unit 13. 6 shows an interpolation pixel (half pixel) H, a best match block B with integer pixel precision, and a search area S for half-pixel precision motion vector detection.

Here, the candidate block is moved by a half pixel in the direction of the motion vectors Va to Vh around the position of the best match block B with integer pixel precision, and the absolute difference between the candidate block and the coded block is calculated. The arithmetic unit 14 calculates the cumulative sum of the absolute values of the differences. Thereafter, although not particularly shown, a candidate block in which the cumulative sum is minimized is selected in the comparing unit, and a motion vector with half-pixel accuracy is obtained from the relationship with the position of the encoded block.

The interpolation pixel H is generated by interpolation between two integer pixels or between four integer pixels so as to be inside and adjacent to the best match block B with integer pixel accuracy. An area obtained by expanding the best match block B of the integer pixel precision by half pixel width outside each side is a search area S for detecting a half pixel precision motion vector.

Here, a case is considered where the search area S for detecting a half-pixel accuracy motion vector includes an area outside the frame boundary of the reference image. 4A shows a case where an out-of-boundary area is included on the left side of the boundary L, FIG. 4B shows a case where an out-of-boundary area is included above the boundary L, and FIG. 4D, FIG. 4D shows a case in which a region outside the boundary is included below the boundary L.

In such a case, within the boundary (including on the boundary)
Since the interpolation pixel H cannot be generated outside the boundary with only the data of the integer pixel P, the range in which the candidate block can be moved is limited. Therefore, a part of the motion vectors Va to Vh must be detected. Can not. For example, FIG.
4, the motion vectors Vd, Ve, Vf cannot be detected, and in the case of FIG. 4B, the motion vectors Vf, Vg, Vh cannot be detected.
Therefore, in such a case, there is no need to generate the interpolated pixel H for detecting these motion vectors outside the boundary, and the interpolated pixel generation unit 13 does not need to perform an operation for that.

Therefore, the boundary determination section 15 determines whether or not the search area S for detecting a half-pixel accuracy motion vector includes a region outside the frame boundary of the reference image. It instructs the address control unit 16 not to generate a data address.

More specifically, this determination is made, for example, by examining whether or not the best match block B with integer pixel accuracy includes pixels on the frame boundary, or by checking the integer pixel accuracy starting from the coded block. By checking whether or not the end point of the motion vector exists on a pixel on the frame boundary.

As a result, for example, as shown in FIG. 4B, when it is determined that the search area S for detecting a half-pixel accuracy motion vector includes a non-boundary area above the boundary L, the motion vectors Vf, Vg , Vh cannot be detected,
Even when there is data of an integer pixel P adjacent to the search area S for detecting a half-pixel accuracy motion vector, the boundary is set so that the interpolation pixel generation unit 13 does not read the data from the search area pixel data memory 11. The determination unit 15 instructs the address control unit 16. The interpolation pixel generation unit 13 calculates the pixel data read from the search area pixel data memory 11
Only the interpolation pixels H that can be generated in the search area S for detecting a half-pixel accuracy motion vector are generated.

At this time, the interpolation pixel generation unit 13 does not perform an operation for generating an unnecessary interpolation pixel H that does not contribute to vector detection, so that the number of operation cycles for detecting a half-pixel precision motion vector can be reduced.

FIG. 5 is a diagram showing another configuration of the motion vector detecting device according to the present invention. The clock controller 17 supplies the clock CLK to the absolute difference calculator 14. Further, the boundary determining unit 15 sends the clock C to the clock control unit 17.
A command is issued regarding supply and stop of LK.

The difference absolute value calculator 14 includes eight difference absolute value accumulators 20 as shown in FIG. 2 corresponding to eight types of vectors which can be obtained when detecting a half-pixel precision motion vector. Have. As described above, when detecting a half-pixel accuracy motion vector, the boundary determination unit 15 may determine that the search region S for detecting a half-pixel accuracy motion vector includes an area outside the frame boundary of the reference image. In such a case, since there is a motion vector that cannot be detected, the difference absolute value accumulator 20 corresponding to the motion vector is present.
The boundary determining unit 15 issues a command to the clock control unit 17 so as to stop the clock supply to the clock control unit 17. By doing so,
The operation of the absolute difference value accumulator 20 that does not contribute to the motion vector detection can be stopped, and the power consumption required for half-pixel accuracy motion vector detection can be reduced.

Further, even when the difference absolute value calculating section 14 has less than eight difference absolute value accumulators 20 and sequentially detects half-pixel precision motion vectors, it performs calculations on motion vectors that cannot be detected. The same applies if the clock supply to the difference absolute value accumulator 20 is stopped only when Therefore, the number of difference absolute value accumulators 20 is not limited to eight.

In the above embodiment, when a motion vector is detected for image data in a frame, whether the search area S for detecting a half-pixel accuracy motion vector includes an area outside the frame boundary of the reference image is determined. Although the description has been made assuming that the boundary determination unit 15 determines whether or not the motion vector is to be detected, the search area S for detecting half-pixel precision motion vectors May be determined by the boundary determination unit to determine whether or not a region outside the slice boundary is included.

[0034]

As described above, according to the present invention, when the search area for half-pixel accuracy motion vector detection includes an area outside the frame boundary or the slice boundary of the reference image, the interpolation pixel which does not contribute to the motion vector detection. By controlling the read address from the search area pixel data memory so as not to generate the motion vector, it is possible to reduce the number of operation cycles related to the detection of the half-pixel accuracy motion vector, and to calculate the absolute difference sum that does not contribute to the motion vector detection. By stopping the container, power consumption can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a motion vector detection device according to an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of a difference absolute value calculation unit.

FIG. 3 is an explanatory diagram showing a search area for detecting a half-pixel precision motion vector, an interpolation pixel, and a motion vector detected when a half-pixel precision motion vector is detected.

FIG. 4 is an explanatory diagram showing a case where a search area for detecting a half-pixel accuracy motion vector includes an area outside a frame boundary.

FIG. 5 is a block diagram illustrating another configuration of the motion vector detection device according to the embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 11 search area pixel data memory 12 coded block pixel data memory 13 interpolation pixel generation unit 14 difference absolute value calculation unit 15 boundary determination unit 16 address control unit 17 clock control unit 20 difference absolute value accumulator B Best match of integer pixel accuracy Block S Half pixel accuracy motion vector detection search area L Frame boundary P Integer pixel H Interpolated pixel (half pixel) Va-Vh Motion vector CLK Clock

Claims (4)

[Claims] 1. A motion vector detecting device for detecting a motion vector with half-pixel accuracy, comprising: a coded block pixel data memory for storing pixel data of a coded block in a picture to be coded; And a search area pixel data memory for storing a half-pixel accuracy motion vector set based on an integer pixel precision motion vector, and interpolating pixels using the pixel data of the search area pixel data memory. An interpolation pixel generation unit for generating; a difference absolute value calculation unit for calculating a cumulative sum of a difference absolute value between the interpolation pixel data generated by the interpolation pixel generation unit and the pixel data of the coded block pixel data memory; A boundary determination unit that determines whether or not the search region for pixel-accuracy motion vector detection includes a region outside the boundary of the reference image; When it is determined that the search area for detecting the half-pixel accuracy motion vector includes an area outside the boundary, the pixel data is read from the search area pixel data memory so that the interpolation pixel outside the boundary is not generated. A motion vector detection device comprising: an address control unit for controlling an address. 2. The motion vector detecting device according to claim 1, wherein a boundary of the reference image is defined by a frame. 3. The motion vector detecting device according to claim 1, wherein a boundary of the reference image is defined by a slice. 4. The motion vector detecting device according to claim 1, wherein the difference absolute value calculation unit includes a plurality of difference absolute value accumulators, and the search for detecting the half-pixel accuracy motion vector is performed. When it is determined that the region includes a region outside the boundary, a clock control unit that controls so as to stop supplying a clock to a difference absolute value accumulator relating to half-pixel precision motion vector detection requiring an interpolation pixel outside the boundary is provided. A motion vector detecting device further provided.