作者: Yun Zhang , Sam Kwong , Guangjun Zhang , Zhaoqing Pan , Hui Yuan
关键词: Random access 、 Video quality 、 Real-time computing 、 Scalable Video Coding 、 Coding tree unit 、 Quantization (signal processing) 、 Computer science 、 Coding (social sciences)
摘要: INTRA video coding is essential for high quality mobile communication and industrial applications since it enhances quality, prevents error propagation, facilitates random access. The latest high-efficiency (HEVC) standard has adopted flexible quad-tree-based block structure complex angular prediction to improve the efficiency. However, these technologies increase complexity significantly, which consumes large hardware resources, computing time power cost, an obstacle real-time applications. To reduce save we propose a fast unit (CU) depth decision method based on statistical modeling correlation analyses. First, analyze spatial CU with different textures present effective strategies predict most probable range among CUs. Since may fail image boundary transitional areas between textural smooth areas, then model-based approach in adaptive early termination thresholds are determined updated rate-distortion (RD) cost distribution, content, quantization parameters (QPs). Experimental results show that proposed can by about 56.76% 55.61% average various sequences configurations; meanwhile, RD degradation negligible.