Shape-adaptive DCT with block-based DC separation and ΔDC correction

This paper refers to a shape-adaptive DCT algorithm (SA-DCT) originally proposed by Sikora and Makai (see ibid., vol.5, p.59-62, 1995). The SA-DCT has been developed in the framework of the ongoing MPEG-4 standardization phase of ISO/IEC, and has been included in the video verification model of MPEG-4. In this context, the focus of the paper is to emphasize a systematic performance limitation of conventional SA-DCT, and to propose an extended version (/spl Delta/DC-SA-DCT) which avoids this restriction in general. This modification considerably improves the efficiency of SA-DCT, and can easily be implemented in existing SA-DCT tools.