Novel Method for Fast 3D DCT for Video Compression

The new fast algorithm for compression of intra and residual frames of video sequences is a transform coding algorithm, based on a three-dimensional Cosine transform with a fixed and variable size of the transformation matrix. The task of eliminating the temporal redundancy is by far one of the central tasks of the process of video compression. It has been shown that the efficiency of removing temporal redundancy algorithms can be improved by increasing the efficiency of inter prediction algorithms by the way of increasing the efficiency of transformation coding. Particular attention is paid to computational complexity indicator of compression process. In the basis of the research of existing algorithms it was found that division of the computational domain into subdomains will significantly reduce the level of computational cost while maintaining the same performance of video procession. Also there is presented an adaptive coding method based on the algorithm of fast three-dimensional Cosine transform, invariant to the properties of moving objects in video sequences such as dynamics of movement, size and accessory to foreground and significantly reduces the computational cost while increasing the quality of the final video sequences. For the evaluation and testing of the developed method a number of video sequences for the standard H.265 from the test set JCT-VC has been used.