Generating algorithm for integer DST radixes in video coding

• Published in: Journal of visual communication and image representation Vol. 25; no. 8; pp. 1918 - 1921
• Main Authors: Wang, Zhongyuan, Chang, Jun, Hu, Ruimin, Zhong, Rui
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Inc 01.11.2014; Elsevier
• Subjects: Algorithms; Applied sciences; Artificial intelligence; Chinese AVS2; Coding; Coding, codes; Computer science; control theory; systems; Exact sciences and technology; H.265; Information, signal and communications theory; Integer DST; Integer transform; Integers; Intra coding; Orthogonality; Pattern recognition. Digital image processing. Computational geometry; Searching; Signal and communications theory; Strategy; Tasks; Telecommunications and information theory; Transform radix; Transforms; Type-7 DST; Video coding; Video coding; Type-7 DST; Transform radix; Integer transform; H.265; Chinese AVS2; Integer DST; Intra coding; Capability index; Search strategy; Arithmetics; Discrete transformation; Orthogonality; Integer matrix; Constrained optimization; Experimental result; Image coding; Heuristic method; Chinese; Matrix method; Metric; ISO standard
• ISSN: 1047-3203; 1095-9076
• DOI: 10.1016/j.jvcir.2014.08.008

•A cost function comprising orthogonality metric and accuracy metric is established.•A generic method is proposed to generate complete integer DST radixes.•A basis superior over HEVC’s is revealed for the first time. Recently released H.265 is the new generation video coding standard established by ITU and ISO jointly. H.265 is the first to apply discrete sine transform (DST) to transform unit in intra prediction block, with it implemented in an integer transform approach. Given that the integer transform radixes of DST are not unique, it is a key task to seek for them possessing strong de-correlation capability and simple arithmetic calculation as well. This paper presents a generic generating algorithm for integer DST transform radixes through a deep insight into the principle of integer DST transform. Subject to three essential constraints on integer transform matrix, this method establishes a cost function comprising orthogonality metric and accuracy metric with respect to entries of a 4×4 matrix, and then seeks transform radixes via a heuristic search strategy. Experimental results show that our method can find out a set of DST radixes after dozens of search steps. The revealed DST radixes not only cover the one used by H.265 but also include a basis superior over H.265’s in terms of orthogonality and precision.