Motion compensation

Block motion compensation (BMC), also known as motion-compensated discrete cosine transform (MC DCT), is the most widely used motion compensation technique.[2] In BMC, the frames are partitioned in blocks of pixels (e.g. macro-blocks of 16×16 pixels in MPEG). Each block is predicted from a block of equal size in the reference frame. The blocks are not transformed in any way apart from being shifted to the position of the predicted block. This shift is represented by a motion vector.

To exploit the redundancy between neighboring block vectors, (e.g. for a single moving object covered by multiple blocks) it is common to encode only the difference between the current and previous motion vector in the bit-stream. The result of this differentiating process is mathematically equivalent to a global motion compensation capable of panning. Further down the encoding pipeline, an entropy coder will take advantage of the resulting statistical distribution of the motion vectors around the zero vector to reduce the output size.

It is possible to shift a block by a non-integer number of pixels, which is called sub-pixel precision. The in-between pixels are generated by interpolating neighboring pixels. Commonly, half-pixel or quarter pixel precision (Qpel, used by H.264 and MPEG-4/ASP) is used. The computational expense of sub-pixel precision is much higher due to the extra processing required for interpolation and on the encoder side, a much greater number of potential source blocks to be evaluated.

The main disadvantage of block motion compensation is that it introduces discontinuities at the block borders (blocking artifacts). These artifacts appear in the form of sharp horizontal and vertical edges which are easily spotted by the human eye and produce false edges and ringing effects (large coefficients in high frequency sub-bands) due to quantization of coefficients of the Fourier-related transform used for transform coding of the residual frames[4]

Block motion compensation divides up the current frame into non-overlapping blocks, and the motion compensation vector tells where those blocks come from (a common misconception is that the previous frame is divided up into non-overlapping blocks, and the motion compensation vectors tell where those blocks move to). The source blocks typically overlap in the source frame. Some video compression algorithms assemble the current frame out of pieces of several different previously-transmitted frames.

Frames can also be predicted from future frames. The future frames then need to be encoded before the predicted frames and thus, the encoding order does not necessarily match the real frame order. Such frames are usually predicted from two directions, i.e. from the I- or P-frames that immediately precede or follow the predicted frame. These bidirectionally predicted frames are called B-frames. A coding scheme could, for instance, be IBBPBBPBBPBB.

Further, the use of triangular tiles has also been proposed for motion compensation. Under this scheme, the frame is tiled with triangles, and the next frame is generated by performing an affine transformation on these triangles.[5] Only the affine transformations are recorded/transmitted. This is capable of dealing with zooming, rotation, translation etc.

Variable block-size motion compensation (VBSMC) is the use of BMC with the ability for the encoder to dynamically select the size of the blocks. When coding video, the use of larger blocks can reduce the number of bits needed to represent the motion vectors, while the use of smaller blocks can result in a smaller amount of prediction residual information to encode. Other areas of work have examined the use of variable-shape feature metrics, beyond block boundaries, from which interframe vectors can be calculated.[6] Older designs such as H.261 and MPEG-1 video typically use a fixed block size, while newer ones such as H.263, MPEG-4 Part 2, H.264/MPEG-4 AVC, and VC-1 give the encoder the ability to dynamically choose what block size will be used to represent the motion.

Overlapped block motion compensation (OBMC) is a good solution to these problems because it not only increases prediction accuracy but also avoids blocking artifacts. When using OBMC, blocks are typically twice as big in each dimension and overlap quadrant-wise with all 8 neighbouring blocks. Thus, each pixel belongs to 4 blocks. In such a scheme, there are 4 predictions for each pixel which are summed up to a weighted mean. For this purpose, blocks are associated with a window function that has the property that the sum of 4 overlapped windows is equal to 1 everywhere.

Studies of methods for reducing the complexity of OBMC have shown that the contribution to the window function is smallest for the diagonally-adjacent block. Reducing the weight for this contribution to zero and increasing the other weights by an equal amount leads to a substantial reduction in complexity without a large penalty in quality. In such a scheme, each pixel then belongs to 3 blocks rather than 4, and rather than using 8 neighboring blocks, only 4 are used for each block to be compensated. Such a scheme is found in the H.263 Annex F Advanced Prediction mode

Practical motion-compensated video compression emerged with the development of motion-compensated DCT (MC DCT) coding,[8] also called block motion compensation (BMC) or DCT motion compensation. This is a hybrid coding algorithm,[7] which combines two key data compression techniques: discrete cosine transform (DCT) coding[8] in the spatial dimension, and predictive motion compensation in the temporal dimension.[7] DCT coding is a lossy block compression transform coding technique that was first proposed by Nasir Ahmed, who initially intended it for image compression, in 1972.[9]

In 1974, Ali Habibi at the University of Southern California introduced hybrid coding,[10][11] which combines predictive coding with transform coding.[7][12] However, his algorithm was initially limited to intra-frame coding in the spatial dimension. In 1975, John A. Roese and Guner S. Robinson extended Habibi's hybrid coding algorithm to the temporal dimension, using transform coding in the spatial dimension and predictive coding in the temporal dimension, developing inter-frame motion-compensated hybrid coding.[7][13] For the spatial transform coding, they experimented with the DCT and the fast Fourier transform (FFT), developing inter-frame hybrid coders for both, and found that the DCT is the most efficient due to its reduced complexity, capable of compressing image data down to 0.25-bit per pixel for a videotelephone scene with image quality comparable to an intra-frame coder requiring 2-bit per pixel.[14][13]

In 1977, Wen-Hsiung Chen developed a fast DCT algorithm with C.H. Smith and S.C. Fralick.[15] In 1979, Anil K. Jain and Jaswant R. Jain further developed motion-compensated DCT video compression,[16][7] also called block motion compensation.[7] This led to Chen developing a practical video compression algorithm, called motion-compensated DCT or adaptive scene coding, in 1981.[7] Motion-compensated DCT later became the standard coding technique for video compression from the late 1980s onwards.[17][2]

The first digital video coding standard was H.120, developed by the CCITT (now ITU-T) in 1984.[18] H.120 used motion-compensated DPCM coding,[7] which was inefficient for video coding,[17] and H.120 was thus impractical due to low performance.[18] The H.261 standard was developed in 1988 based on motion-compensated DCT compression,[17][2] and it was the first practical video coding standard.[18] Since then, motion-compensated DCT compression has been adopted by all the major video coding standards (including the H.26x and MPEG formats) that followed.[17][2]