KR970010096B1 - Two layer motion estimation apparatus and method for image data compression - Google Patents

Abstract

The two layer movement deducing apparatus is comprised of (a) determining input image data is either a field or a frame data(201), one half picking out a searching region and a standard image of an original image in the horizontal directions upon the input image data being determined to be the field data(202,203), one half picking out the original image and the standard image in the horizontal and vertical directions upon the input image data being determined to be the frame data to generate the searching region and standard block data(204); (b) performing whole region searching with respect to the picked out data and generating a movement vector in unit of integer pixel(205) and adjusting the generated movement vector(206); and upon the adjustment of the movement vector value being completed for a half pixel searching region to exist within the whole searching region, finding the searching region data where the position of the half pixel is interpolated(207) from a coding part(10), performing the whole region searching to generate the movement vector in unit of the half pixel(208).

Description

Two-layer Motion Estimation Apparatus and Method for Compression Coding of Image Data

1 is a configuration diagram of an encoder system for general video compression.

2 is a control flowchart of the present invention.

3 is a block diagram of a motion estimation unit according to the present invention.

* Explanation of symbols for main parts of the drawings

1: Input image storage unit 2: Discrete cosine conversion unit

3: quantizer 4: inverse quantizer

5: Inverse discrete cosine converting unit 6: Playback image storage unit

7: motion compensation unit 8: motion estimation unit

9 variable length encoder 10 encoder

11, 25: horizontal jogger 12, 26: horizontal jogger

13 to 17: block matching unit 18, 20, 21, 23: half pixel estimation unit

19, 22, 27: comparison unit 28: horizontal and vertical interpolation unit

The present invention relates to a motion estimation apparatus and a method thereof, and in particular, to decipher the input image data in order to reduce the calculation amount of the motion estimation, and to perform the motion estimation of the deduced pixel unit primarily in a given search region. In order to compensate for the accuracy of the motion vector generated from the deduced image data, the image data is interpolated in the original pixel unit and the secondary motion tracking is performed in the local region to perform the final half pixel motion estimation. The present invention relates to a two-layer motion estimation apparatus for compression coding and a method thereof.

FIG. 1 illustrates a general image compression encoder system. The input digital image is stored in the input image storage unit 1 frame by frame, and the motion estimation unit 8 stores the input image storage unit 1. In the case of the past frame data or the bidirectional prediction, which is the playback data stored in the video data and the playback video storage unit 6, the motion vector is calculated in units of macro blocks using the past or future frame data.

As described above, the motion vector calculated by the motion estimator 8 is sent to the motion compensator 7 to compensate for the motion, and then the discrete cosine transform unit (DCT) together with the original image stored in the input image storage unit 1. L Discrete Cosine Transform) (2). Accordingly, the discrete cosine transforming unit (DCT) 3 converts the difference between the original image and the motion compensated image and outputs the frequency to the quantizer 3.

The quantizer 3 quantizes the output of the discrete cosine transforming unit (DCT) 3 according to the bit rate required in the fixed path, and outputs the result to the transmitting unit through the variable length encoder 9.

On the other hand, the output data of the quantizer 3 is combined with the video whose motion is compensated for by the inverse quantizer 4 and the inverse discrete cosine transform unit (ICT) 5 and the output of the motion compensator 7 is reproduced. It is stored in the storage unit 6.

In the encoder system for image compression as described above, the motion estimation unit 8 plays a major role in reducing the data rate to be transmitted by removing redundancy between successive image frames.

Conventional systems made based on MPEG-2 use field / frame adaptive motion estimation and compensation, which are global search, hierarchical search, or other efficient search method for integer motion estimation. We calculate the motion vector of the original image by using, and estimate the half-pixels for the surrounding 9 points using the result.

However, the global search method has a problem that the search area cannot be made large enough as necessary because an excessive amount of computation is required for the search. The hierarchical search and other efficient search methods reduce the calculation amount for the search, There was a problem that the estimation error caused by the motion estimation could not be compensated in the half-pixel unit estimation step.

Accordingly, the present invention has been made to solve the above problems, it is to deduce the image data that is applied to reduce the calculation amount of the motion estimation, and to estimate the motion of the estimated pixel unit in the given search region In order to compensate for the accuracy of the motion vector generated from the deduced image data, the image data is interpolated in the original pixel unit and the second motion estimation is performed in the local region to perform the motion estimation of the final half pixel unit. An object of the present invention is to provide a two-layer motion estimation apparatus for compressing and encoding data and a method thereof.

In order to achieve the above object, according to the present invention, in the image compression encoder system, field 1 and field 2 (the field data of the input original image and the reference image are respectively deduced by appropriate values in the horizontal direction). even field) first frame means for generating the search region and the reference block data, and frame data of the input original image and the reference image are deduced by appropriate values in the horizontal and vertical directions to generate the frame search region and the reference block data. Generating a motion vector by performing a global search by receiving the second extraction means, the field 1, the field 2 search region and the reference block data, and the frame search region and the reference block data output from the first extraction means and the second extraction means, respectively. Block matching means, interpolation means for interpolating a reproduced image to an appropriate value in the horizontal and vertical directions, and the block image interpolated with the reproduced image interpolated by the interpolation means. A half-pixel estimating means for receiving a motion vector output from the means for global search and calculating an estimated error, and comparing the estimated error obtained from the half-pixel estimating means to calculate a motion vector and an estimated error for fields 1 and 2; A first comparison means and a second comparison means for comparing a motion vector with respect to the fields 1 and 2 calculated from the first comparison means and an estimated error and a motion vector with respect to the frame and an estimated error to produce a final motion vector. And a motion estimating unit.

In the video compression encoder system, the first step of determining whether the input image data is the field data or the frame data, and if the image data input in the step is the field data, input the image data. A second step of generating a search region and reference block data by deducting the proper value in the horizontal direction and inputting the image data to the proper value in the horizontal and vertical directions if the input image data is the frame data; A third step of generating an integer pixel unit motion vector through global search using the searched detection area and the reference block data, a fourth step of adjusting the motion vector generated in the step not to exceed the allowable range, and interpolating the playback image. Compose half search unit and search globally to generate half pixel unit motion vector. The motion estimation is performed through the fifth step.

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

The present invention is a two-stage hierarchical search method applied to an interlaced image, and performs motion estimation in units of integer and pixel units for each macroblock and half pixel unit for each macroblock.

FIG. 2 is a control flow diagram of the present invention, and it is determined whether the input image data is a field or a frame (201). If the field data, that is, the odd or even field data, the search region and the reference image of the original image are determined. The search area and the reference block data are made by halving in the horizontal direction (202, 203). If the input image data is frame data, the search area and the reference block data are reduced by halving the original image and the reference image in the horizontal and vertical directions. (204)

That is, for the field data, the original image and the reference image are delineated 1/2 in the horizontal direction by the following equation.

f (x, y) = [f (2x, y) + f (2x + 1, y)] / 2

Here, x is the horizontal coordinate of the pixel, y is the vertical coordinate of the pixel, and f (x, y) is the pixel value at the coordinate position (x, y).

For the frame data, the original image is rounded down to 1/4, i.e., in the horizontal and vertical directions, respectively, by the following equation.

f (x, y) = [f (2x, 2y) + f (2x + 1, 2y) + f (2x, 2y + 1) + f (2x + 1, 2y + 1)] / 4

The global search is performed on the data deduced as described above to generate an integer pixel-based motion vector (205).

On the other hand, the search area to be subjected to the global search has the same size in the vertical and horizontal directions with respect to the field and the frame, and the size of the reference block is the same, so that all the field frames are (8x8) pixels.

If the motion vector value generated as described above is located at the edge of the allowed motion search region, the half pixel search region may be out of the entire search region, and thus the motion vector value is adjusted to prevent this.

When the adjustment of the motion vector value is completed so that the half-pixel search region exists in the entire search region, the search region data obtained by interpolating the half-pixel position is obtained from the data reproduced by the encoder 10 (207). Generate a pixel-by-pixel motion vector.

Interpolation of the reproduced image data is performed by the following equation.

f (x + 0.5, y) = [f (x, y) + f (x + 1, y)] / 2

f (x, y + 0.5) = [f (x, y) + f (x, y + 1)] / 2

f (x + 0.5, y + 0.5) = [f (x, y) + f (x + 1, y) + f (x, y + 1) + f (x + 1, y + 1)] / 4

The search area interpolated as above is -4 to +3.5 pixels in both vertical and horizontal, and the reference block size is 16x8 pixels in the field and 16x16 pixels in the frame.

An embodiment of the motion estimation unit using the above algorithm is illustrated in FIG. 3.

First, search area data of field 1 and field 2 are generated by horizontally extracting the search area of the original image through the horizontal 1/2 copy unit 11 for the first step search. In addition, the original image is reduced by one half in the horizontal and vertical directions through the horizontal and vertical half-folding units 12 to generate a frame search area.

On the other hand, the reference image is also similarly deduced in the horizontal direction through the horizontal 1/2 deduction unit 25 to generate field 1 and field 2 reference block data, and the reference image is horizontally and vertically 1/2 deduction unit ( 26), frame reference block data is generated by decrementing each half in the horizontal and vertical directions.

By combining the respective search areas and the reference block data, five block matching units, that is, the first block matching unit 13, receives data from the field 1 search area and the field 1 reference block to perform block matching and perform the second block. The matching unit 14 receives the data of the field 2 search area and the field 1 reference block to perform block matching, and the third block matching unit 15 receives the data of the field 1 search area and the field 2 reference block and performs block matching. The fourth block matching unit 16 receives the data of the field 2 search area and the field 2 reference block to perform block matching. The fifth block matching unit 17 applies the frame search area and the frame reference block data. It performs block matching and outputs a motion vector.

The reconstructed image search area, which is the interpolation of the motion vector and the horizontal and vertical 1: 2 interpolation unit 1 and 2 in the horizontal and vertical directions, respectively, is a first to fifth half pixel value. Approved to the governments 18, 20, 21, 23, and 24, the half-pixel estimator performs a half-pixel search.

That is, the first half-pixel estimation unit 18 receives the motion vector output from the first block matching unit 13 and the reproduced image interpolated by the horizontal and vertical 1: 2 interpolation unit 28, and performs global search to estimate the error. The second half-pixel estimator 20 receives the motion vector output from the second block matcher 14 and the reproduced image interpolated by the horizontal and vertical 1: 2 interpolator 28, and performs a global search. The third half-pixel estimator 21 receives the motion vector output from the third block matcher 15 and the reproduced image interpolated by the horizontal and vertical 1: 2 interpolator 28 to search for the error. The fourth half-pixel estimation unit 23 receives the motion vector output from the fourth block matching unit 13 and the reproduced image interpolated by the horizontal and vertical 1: 2 interpolation unit 28, and then performs a global search. The fifth half-pixel estimator 24 calculates an estimated error, and the fifth half-pixel estimator 24 horizontally and numbers the motion vector output from the fifth block matcher 17. 1:02 to receive beam is applied to the reproduced image from the interpolation shaft 28 full search to calculate the estimated error.

By comparing the estimation error obtained as a result of the half-pixel search, the first comparison unit 19 compares the estimation error obtained by the first half-pixel estimator 18 and the second half-pixel estimator 20 to field 1 The second comparison unit 22 compares the estimation error obtained by the third half-pixel estimator 21 and the fourth half-pixel estimator 23 with respect to the motion vector and the estimation error with respect to. And an estimation error, and the third comparison unit 27 finally compares the motion vector and the estimation error for the field 1, the field 2, and the frame to determine the field or frame prediction mode, and calculates the final motion vector value accordingly. Output

As described above, the present invention reduces the computation amount for motion estimation by deciding the input image data, and performs motion estimation in units of half pixels by interpolating local area data based on the estimated position at the front end. This has the effect of further improving accuracy.

Claims (8)

In the video compression encoder system, field data of the input original image and the reference image are deduced by appropriate values in the horizontal direction, respectively, to generate field 1 (odd field), field 2 (even field) search area, and reference block data. First pushing means (11,25); Field 1 and Field 2 search output from the second deduction means 12 and 26 generating frame search area and reference block data by deducting the frame data of the input original image and the reference image to appropriate values in the horizontal and vertical directions. Block matching means (13 to 17) for receiving the region and the reference block data and the frame search region and the reference block data, respectively, to search globally to generate a motion vector; Interpolation means 28 for interpolating a reproduced image applied from an external reproduced image storage unit 6 to an appropriate value in the horizontal and vertical directions; Half-pixel estimation means (18, 20, 21, 23) for global search and calculation of the estimated error by receiving the interpolated video interpolated by the interpolation means (28) and the motion vectors output from the block matching means (13 to 17); 24); First comparison means (19,22) for comparing motions estimated by the half-pixel estimation means and calculating motion vectors and estimation errors for fields 1 and 2; A motion vectorization estimation error for the field 1 and the field 2 calculated by the first comparison means 19 and 22 and a motion vector and the estimation error for the frame calculated by the block matching means to output a final motion vector; A two-layer motion estimation apparatus for compressing and encoding image data, characterized in that it comprises a motion estimating unit comprising two comparison means (27). 2. The apparatus of claim 1, wherein the block matching unit comprises: a first block matching unit (13) for generating a motion vector by performing block matching by receiving data of the field 1 search area and the field 1 reference block; A second block matching unit 14 which receives data of the field 2 search area and the field 1 reference block and performs block matching to generate a motion vector; A third block matching unit 15 which receives data of the field 1 search area and the field 2 reference block and performs block matching to generate a motion vector; A fourth block matching unit 16 receiving data of the field 2 search area and the field 2 reference block to perform block matching to generate a motion vector; And a fifth block matching unit (17) for generating a motion vector by performing block matching upon receiving the frame search area and the frame reference block data. The half-pixel estimating means receives a motion vector output from the first block matching unit 13 and a reproduction image interpolated by the horizontal and vertical 1: 2 interpolation unit 28. A first half pixel estimator 18 for searching and calculating an estimated error; The second half-pixel estimator 19 calculates an estimated error by receiving the motion vector output from the second block matcher 14 and the reproduced image interpolated by the horizontal and vertical interpolation unit 28 and performing global search. ; The third half-pixel estimator 21 calculates an estimated error by receiving a motion vector output from the third block matcher 15 and a reproduced image interpolated by the horizontal and vertical interpolator 28, and performing global search. ; A fourth half-pixel estimator 23 that receives a motion vector output from the fourth block matcher 13 and a reproduced image interpolated by the horizontal and vertical interpolator 28, and globally searches and calculates an estimation error. ; A fifth half-pixel estimator 24 which receives a motion vector output from the fifth block matcher 17 and a reproduced image interpolated by the horizontal and vertical interpolator 28 to search globally and calculate an estimation error Two-layer motion estimation apparatus for image data compression coding, characterized in that consisting of. 4. The method of claim 3, wherein the first comparing means compares the estimation errors obtained by the first half pixel estimator 18 and the second half pixel estimator 20 to calculate a motion vector and an estimation error for field 1. A first comparison unit 19; And a second comparison unit 22 which calculates a motion vector and an estimation error for the field 2 by comparing the estimation errors obtained by the third half pixel estimator 21 and the fourth half pixel estimator 23. A two-layer motion estimation apparatus for video data compression coding. First estimation means (11, 23) for generating field 1 (odd field), field 2 (even field) search areas, and reference block data by deducting the field data of the input original image and the reference image to appropriate values in the horizontal direction, respectively. Second sampling means (12,26) for generating a frame search region and reference block data by deducting the frame data of the input original image and the reference image to an appropriate value in the horizontal and vertical directions; A block for generating a motion vector by global search by receiving field 1, field 2 search area and reference block data, frame search area, and reference block data output from (11,25) and second extraction means (12,26), respectively. Matching means (13 to 17), interpolation means (28) for interpolating the reproduced image to an appropriate value in the horizontal and vertical directions, and the reproduced image and the block matching means (13 to 17) interpolated by the interpolation means (28). Receives the motion vector output from The half-pixel estimation means (18, 20, 21, 23, 24) which search the entire area and calculate the estimation error and the estimation error obtained by the half-pixel estimation means are compared with the motion vector and the estimation error for the field 1 and the field 2. First and second comparison means (19, 22) for calculating?, Motion vectors and estimated errors for fields 1 and 2 calculated by the first comparison means (19, 22), and for frames calculated by the block matching means; In a motion estimation method applied to an image compression encoder system including a motion estimation unit including a second comparison means 27 for comparing a motion vector and an estimation error and outputting a final motion vector, the input image data is field data. A first step 201 of determining whether the data is frame data; If the image data input in the first step is field data, the input image data is deduced to an appropriate value in the horizontal direction (202, 203). If the input image data is frame data, the input image data is an appropriate value in the horizontal and vertical directions. A second step of generating a search region and reference block data by inferring (204); A third step (205) of generating an integer pixel unit motion vector by performing a global search through the search area and the reference block data generated in the second step; A fourth step (206) of adjusting the motion vector generated in the third step not to exceed an allowable range; And a fifth step (207, 208) of constructing a half pixel unit search region by interpolating the playback image applied from the external playback image storage unit 6 and generating a half pixel unit motion vector by global search. A two-layer motion estimation method for image data compression coding. 6. The image of claim 5, wherein if the data to be deduced in the second step is field data, the image is deduced by f (x, y) = [f (2x, y) + f (2x + 1, y)] / 2. Two-layer motion estimation method for data compression coding. The method according to claim 5, wherein if the data to be deduced in the second step is frame data, f (x, y) = [f (2x, 2y) + f (2x + 1, 2y) + f (2x, 2y) + 1) + f ( A two-layer motion estimation method for video data compression coding, which is rounded by 2x + 1, 2y + 1)] / 4. The method of claim 5, wherein the interpolated playback image data in the fifth step f (x + 0.5, y) = [f (x, y) + f (x + 1, y)] / 2 f (x, y + 0.5) = [f (x, y) + f (x, y + 1)] / 2 f (x + 0.5, y + 0.5) = [f (x, y) + f (x + 1, y) + f (x, y + 1) + f (x + 1, y + 1)] / 4 A two-layer motion estimation method for video data compression encoding, characterized by comprising: