KR100207399B1 - Motion estimation apparatus - Google Patents

Abstract

The present invention relates to a motion estimation apparatus for determining a motion vector representing a displacement between a current frame and a previous frame of an image signal, and determining a motion vector by panning of a camera, wherein the current frame has a plurality of pannings having the same size. A motion estimation apparatus comprising a plurality of candidate blocks having a size equal to that of the panning search block, wherein a previous frame is divided into a search block: sequentially receiving predetermined motion vectors of each search block adjacent to the search block, When the panning search block is moved by each of the motion vectors on the previous frame, the panning search block is defined as a search area including a predetermined peripheral pixel section around a candidate block of the previous frame matching the moved panning search block. A search region forming unit 160; A candidate block providing unit for generating candidate blocks from the search region corresponding to each motion vector defined by the search region forming unit 160; A panning predicting unit (180) for calculating an error function between the panning search block and the blocks in the search area to generate a first minimum error value having the smallest double error; A memory 190 for storing first minimum error values corresponding to each motion vector; A panning prediction vector detection unit (200) for comparing the first minimum error values corresponding to each motion vector stored in the memory (190) with each other and determining a motion vector corresponding to a second minimum error value having a minimum error value; A comparison unit (210) for comparing the motion vector determined by the panning prediction vector detector (200) with a preset threshold value; If the comparison is determined by the comparator 210 and the motion vector determined by the panning prediction vector detector 200 is smaller than a preset threshold, the motion vector provided from the panning prediction vector detector 200 is converted into a motion vector by panning. It is characterized by comprising a control unit 220 for determining.

Description

Motion estimation device

1 is a detailed block diagram showing a preferred embodiment of the present invention.

2 and 3 are views for explaining a motion vector detection process according to an embodiment of the present invention.

* Explanation of symbols for main parts of the drawings

110, 120: frame memory 130: motion compensation unit

140: calculator 150: motion estimation unit

160: panning search region forming unit 170: candidate block providing unit

180: panning prediction unit 190: memory

200: panning prediction vector detection unit 210: comparison unit

220: control unit

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion estimation apparatus for performing inter-frame coding by estimating motion of an image generated according to panning of a camera.

In general, it is well known that digital image transmission can maintain a much better image quality than an analog transmission, but in the case of a high-definition TV, a frame-by-frame image When expressed in this digital form, a large amount of digital data must be transmitted.

However, in the transmission channel for transmitting digital data, the available frequency range is limited. Therefore, in order to transmit a large amount of digital data, it is necessary to reduce the amount of transmission by compressing the transmitted data. In addition, hybrid coding schemes combining spatial compression techniques are known to be the most efficient.

Most hybrid coding techniques use motion compensated DPCM (Differential Pulse Code Modulation), two-dimensional Discrete Cosine Transform (DCT), quantization of DCT coefficients, VLC (variable length coding), and the like. In particular, the motion compensation DPCM determines the motion of the object between the current frame and the previous frame, and predicts the current frame according to the motion of the object to produce a differential signal representing the difference between the current frame and the predicted value. Ericsson's Fixed and Adaptive Predictors for Hybrid Predictive / Transform Coding, IEEE Transactions on Communication, COM-33, NO.12 (1985, December), or A motion Compensated Interframe Coding Scheme for Television Pictures, IEEE Transactions by Ninomiy and Ohtsuka It is well described in on Communication, COM-30, NO.1 (January, 1982).

On the other hand, the type of the picture is a forward motion estimation / compensation from an I-frame to which only interframe coding is applied, a previous I-frame or a P-frame according to a motion estimation method. P-frames that are predictively coded, and B-frames that are bi-directional motion-compensated predictive coded from two frames, i.e., P-frames before and after, and used as a reference for predictive coding other frames. Frames, i.e. and P-frames, are called anchor frames.

Therefore, various methods for estimating the displacement, or motion vector, of an object using the current frame and the anchor frame have been proposed in the related art. The conventional motion vector estimation method has two types, namely, a pixel repetition algorithm and a block matching algorithm (JRJain). Displacement Measurement and Its Application in Interframe Image Coding, IEEE Transactions on Communication, COM-29, NO.12 (see December, 1981).

In the more widely used block matching algorithm, the current frame is divided into a number of search blocks, and the size of each search block is generally in the range of 8x8 to 32x32 pixels. In order to determine the vector, the similarity is calculated between each candidate block included in the search region of the previous frame and the search block of the current frame having the same size, and averaged for calculating the similarity between the search block of the current frame and the candidate block of the search region. Error functions such as absolute error and mean square error are used.

In addition, the motion vector is defined as the displacement between the search block and the candidate block generating the smallest error function value.

However, since the block matching algorithm calculates the similarity between each candidate block having the same size and the search block of the current frame included in the search region of the previous frame, the accuracy of the motion prediction is high, while the amount of computation for the motion prediction is high. Too many systems were cumbersome and inefficient in time.

The present invention has been made to solve the above problems, by determining the motion vector according to the movement of the screen caused by the panning of the camera to perform the motion prediction, to reduce the amount of motion prediction calculation according to the block matching algorithm The object is to provide an estimating apparatus.

In order to achieve the above object, the present invention provides a motion estimation apparatus for determining a motion vector indicating a displacement between a current frame and a previous frame of an image signal, and determining a motion vector by panning of a camera, wherein the current frame is the same. A motion estimation apparatus comprising a plurality of candidate blocks having a size equal to the panning search block, wherein the previous frame is divided into a plurality of panning search blocks having a size: A predetermined motion of a search block adjacent to the panning search block. When the panning search block is moved by each of the motion vectors on the previous frame, the predetermined peripheral pixel intervals are centered around the candidate block of the previous frame that matches the moved panning search block. Navigation area, including navigation area Forming part; A candidate block providing unit generating candidate blocks from the search region corresponding to each motion vector defined by the search region forming unit; A panning predictor configured to calculate an error function between the panning search block and candidate blocks in the search area, respectively, to generate a first minimum error value having the smallest double error, and to obtain first minimum error values corresponding to each motion vector; Memory for storing; A panning prediction vector detector for comparing the first minimum error values corresponding to each motion vector stored in the memory with each other and determining a motion vector corresponding to a second minimum error value having a minimum error value; A comparison unit comparing the motion vector determined by the panning prediction vector detector with a preset threshold; And a controller configured to determine a motion vector provided from the panning prediction vector detector as a motion vector by panning when the motion vector determined by the panning prediction vector detector is smaller than a preset threshold. It is done.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

In general, in performing inter-frame encoding by using time-domain correlation of image data, image data of a frame to be encoded between adjacent frames and a frame for motion prediction includes movements of objects in the image, and Can be divided into panning movements.

If the camera is fixed and the objects are moving, the background without motion will be motionless in the next frame. Also, if the camera pans, the backgrounds will have a constant motion vector and only the moving objects will have a different vector.

Therefore, when determining the motion vector according to panning, motion prediction is not necessary in most of the images without motion such as the background, and the motion is searched only in the region to which the object belongs.

FIG. 1 is a detailed block diagram illustrating a preferred embodiment of the present invention, and includes a frame memory 110 storing image data of a current frame, a frame memory 120 storing image data of a previous frame, and an object between a current frame and a previous frame. The motion compensator 130 that predicts the motion of the current frame using the motion vector corresponding to the displacement and then provides a predicted frame between the current frame of the frame memory 110 and the predicted frame of the motion compensator 130. The operation unit 140 which detects the difference and provides the data for the discrete cosine transform, provides a predetermined motion vector of the search block adjacent to the panning search block of the current frame to determine the motion vector by panning, and moves by panning. When the vector is provided from the outside, the vector is provided to the motion compensator 130, and the panning search block moves by panning. If it is not determined, the motion vector is determined by the block matching algorithm and is provided to the motion compensation unit 130. The search region forming unit 160 sequentially receiving the predetermined motion vectors and defining a search region including ± 2 pixel sections around the candidate block of the previous frame corresponding to the position away from the panning search block by the received motion vector. ), A candidate block provider 170 generating candidate blocks from a search region corresponding to each motion vector defined by the search region forming unit 160, a panning search block of the current frame, and a search filter corresponding to the vector. A panning predictor 180 which calculates an error function between candidate blocks in the region and provides a minimum error value having the smallest double error, The memory 190 storing the minimum error value corresponding to the motion vector and the minimum error value corresponding to each motion vector stored in each memory 190 are compared with each other, and the motion vector corresponding to the error value having the minimum error value is compared. The comparison result of the panning prediction vector detector 200, the panning prediction vector detector 200, which is determined by the comparison unit 210, and the comparison unit 210. If the motion vector provided from 200 is greater than a preset threshold, the motion estimation unit 150 controls to determine the motion vector by the block matching algorithm, and the motion vector provided from the panning prediction vector detector 200 is preset. The controller 2 controls the motion vector provided from the panning prediction vector detector 200 to the motion estimator 150 when the threshold value is smaller than the threshold. 20).

Operation of the preferred embodiment of the present invention configured as described above is as follows.

In general, in performing motion estimation, the current frame is divided into search blocks of equal size in M * N (M, N: positive integer) pixels. Accordingly, referring to FIG. 2, eight search blocks 11, 12, 13, 14, 15, 16, 17, and 18 are adjacent to a predetermined panning search block 10 of the current frame.

Further, in case of motion estimation for each search block of a frame divided into a plurality of search blocks, motion estimation of the search block above the frame is first performed, and motion estimation of the search block below the frame is performed later. However, motion estimation of the search block on the left in the same horizontal position on the frame is performed first, and motion estimation of the search block on the right is performed later.

Accordingly, referring to FIG. 2, in order to estimate the motion of the panning search block 10, reference numerals 11, 12, 13, and 14 of the search blocks 11 to 18 adjacent to the panning search block 10 are used. The search blocks shown as are already search blocks for which motion estimation has been completed, and the other search blocks 15, 16, 17, and 18 are blocks for which motion estimation has not yet been made.

In the preferred embodiment of the present invention, when the motion for the panning search block 10 is to be estimated, the motion vectors of the adjacent search blocks 11, 12, 13, and 14, which have already been estimated, and the magnitude of 0 are zero. Use motion vectors

As shown in FIG. 2, the motion estimation unit 150 may search for blocks 11, 12, 13, and 14 in which motion vectors are determined among the search blocks 11 to 18 adjacent to the panning search block 11 of the current frame. When the motion vectors A1, A2, A3, and A4 are sequentially provided to the panning vector detector 160, the panning search region forming unit 160 searches for panning by the motion vectors provided from the motion estimation unit 150. A search region is defined as an area including ± 2 pixel intervals around the candidate block of the previous frame corresponding to the position away from the block.

The candidate block provider 170 provides each of the candidate blocks in the search region defined by the panning search region forming unit 160 to the panning predictor 180, and the panning predictor 180 sends the frame memory 110 to the frame memory 110. An error function is calculated through block matching with the candidate blocks provided by the panning search block 10 and the candidate block providing unit 170, and the double minimum error value is provided to the memory 190.

That is, when the motion estimation unit 150 provides the motion vector A1 to the panning search region forming unit 160, as shown in FIG. 3, the panning search region forming unit 160 corresponds to the motion vector A1. A search region 21 is defined as a region including ± 2 pixel intervals around the candidate block 20 of the previous frame corresponding to the position away from the panning search block 10. Accordingly, the candidate block provider 170 provides each candidate block in the search area 21 to the panning predictor 180, and the panning predictor 180 provides the panning search block 10 provided from the frame memory 110. ) And the error function are calculated through block matching with the candidate blocks provided from the candidate block provider 170 and then stored in the double minimum error value memory 190.

In addition, when the motion estimation unit 150 provides the motion vector A2 to the panning search region forming unit 160, the panning search region forming unit 160 performs the panning search block 10 by the motion vector A2 as described above. The search region 21 defines a region including a ± 2 pixel section around the candidate block 20 of the previous frame corresponding to the position away from the cross-section. Accordingly, the candidate block provider 170 provides each candidate block in the search area 21 to the panning predictor 180, and the panning predictor 180 provides the panning search block 10 provided from the frame memory 110. The error function is calculated through block matching with the candidate blocks provided from the candidate block provider 170 and the double minimum error value is stored in the memory 190.

The panning predictor 170 sequentially performs such a process from the motion vector A1 to the motion vector A4 and stores the minimum error values corresponding to the motion vectors A1 to A4 in the memory 190, respectively. .

The panning prediction vector detector 200 compares the minimum error values corresponding to each of the motion vectors A1 to A4 stored in the memory 190 with each other, and has a minimum error value having a minimum error value among the minimum error values corresponding to each motion vector. The motion vector corresponding to the value is provided to the comparator 210.

At this time, for example, luminance values are compared between the panning search block 10 and the corresponding pixels in the candidate blocks of the search region corresponding to the motion vectors A1 to A4, thereby causing an error for each candidate block. The function will be generated, and the error function will indicate the degree of similarity between the panning search block and each candidate block.

The comparison unit 210 compares the vector value provided from the panning prediction vector detection unit 200 with a preset threshold value, and provides the comparison result to the control unit 220.

The controller 220 determines that the similarity between the panning search block and the candidate block corresponding to the motion vector provided from the panning prediction vector detection unit 200 is low when the vector value provided from the panning prediction vector detection unit 200 is greater than the preset threshold. The motion estimation unit 150 is then controlled to perform motion vector determination on the panning search block by a block matching algorithm.

However, if the vector value provided from the panning prediction vector detector 200 is smaller than the preset threshold, the controller 220 may have a similarity between the panning search block and the candidate block corresponding to the motion vector provided from the panning prediction vector detector 200. After determining that is high, the motion estimator 150 is controlled to control the motion vector provided from the panning prediction vector detector 200 to the motion compensator 130.

On the other hand, the motion estimation unit 150 stores the motion vector by panning from the panning prediction vector detection unit 200 under the control of the control unit 220 and stores the motion vector, and provides the same to the motion compensation unit 130. If the motion vector due to panning is not provided from the vector detector 200, the motion vector is estimated by estimating the motion between the search block of the current frame and the candidate blocks of the previous frame, the motion vector is determined and stored, and provided to the motion compensator 130. do.

Therefore, the motion compensator 130 predicts the motion of the current frame using a motion vector corresponding to the displacement of the object between the current frame and the previous frame, and then provides the predicted frame to the calculator 140, and the calculator 140. Detects the difference between the current frame of the frame memory 110 and the predicted frame of the motion compensator 130 and provides the data for the discrete cosine transform.

As described above, the present invention determines the motion vector according to the panning of the camera using a predetermined motion vector before performing the motion estimation with an existing algorithm such as a block matching algorithm. have.

Claims (2)

A motion estimation apparatus for determining a motion vector representing a displacement between a current frame and a previous frame of an image signal, wherein the current frame is divided into a plurality of panning search blocks having the same size. And the previous frame includes a plurality of candidate blocks having the same size as the panning search block: sequentially receiving predetermined motion vectors of a search block adjacent to the panning search block, and performing the panning search. When a block is moved by each of the motion vectors on the previous frame, a search region is defined including a predetermined pixel region around a candidate block of the previous frame matching the moved panning search block. Unit 160; A candidate block provider 170 generating candidate blocks from the search region corresponding to each motion vector defined by the search region forming unit 160; A panning predicting unit (180) for calculating an error function between the panning search block and the blocks in the search area to generate a first minimum error value having the smallest double error; A memory 190 for storing first minimum error values corresponding to each motion vector; A panning prediction vector detection unit (200) for comparing the first minimum error values corresponding to each motion vector stored in the memory (190) with each other and determining a motion vector corresponding to a second minimum error value having a minimum error value; A comparison unit (210) for comparing the motion vector determined by the panning prediction vector detector (200) with a preset threshold value; If the comparison is determined by the comparator 210 and the motion vector determined by the panning prediction vector detector 200 is smaller than a preset threshold, the motion vector provided from the panning prediction vector detector 200 is converted into a motion vector by panning. And a control unit 220 for determining the motion estimation device. The motion estimation apparatus of claim 1, wherein a predetermined peripheral pixel section around the candidate block of the previous frame is: ± 2 pixel section around the candidate block of the previous frame.