Algorithm and architecture of prediction core in stereo video hybrid coding system

Abstract

3D video will become noticeable video technology in the next generation. In this paper, a stereo video coding system is proposed from algorithm level to hardware architecture level. We propose a novel stereo video coding system by exploiting joint block compensation scheme to achieve high coding efficiency. It is also suitable for hardware implementation. Due to more than twice computational complexity relative to mono video coding systems, a new hardware architecture based on hierarchical search block matching algorithm (HSBMA) with some modification is proposed. With special data flow, no bubble cycles exist during block matching process. Proposed architecture also adopts near overlapped candidates reuse scheme (NOCRS) to save heavy burden of data access. Besides, by the proposed new scheduling, both on-chip memory requirement and offchip memory bandwidth can be reduced. A prototype chip can achieve real-time requirement under the operating frequency of 81 MHz for 30 D1 frames per second (fps) in left and right channel simultaneously, with ME/DE search range of [-64, +63] in horizontal direction and [-32, +31]/[-16, +15] in vertical direction. Compared with the hardware requirement for implementation of full search block matching algorithm (FSBMA), only 11.5% on-chip SRAM and 1/30 amount of PEs are needed. It shows that the hardware cost is quite small.