Improved H.264/AVC and H.264/SVC with Adaptive GOP Detection Based on Content Variations

• Main Authors: Jun-Ren Ding, 丁俊仁
• Other Authors: Jar-Ferr Yang
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/69710384245931881269

博士 === 國立成功大學 === 電機工程學系碩博士班 === 96 === Video compression technology can result in different compression performance for different videos due to different VCV (video content variation) in the same target bitrate. It is important to adaptively adjust the GOP (group of pictures) for different VCV. Besides, the compression performance will be affected while a scene change occurs in a GOP. Therefore, in this paper, we suggest a joint AGD (adaptive GOP detection) and SCD (scene change detection) coding (JASC) algorithms to improve the H.264-based advance video coding (H.264/AVC) and scalable video coding (H.264/SVC). Simulation results show that H.264/SVC with JASC saves the computational time of 59% on average, and obtains the SCD detection rate of 89% than CE2 (core experiment 2); the H.264/AVC with JASC improves the PSNR of 0.62dB than the initial H.264/AVC, and obtains the SCD detection rate of 98%. Although H.264/AVC and H.264/SVC are the newest video compression standards, the compression performance of H.264-based lossless (H.264-LS) image coding is dissatisfactory. In order to implement lossy, lossless and near lossless bestreams simultaneously, we suggest adopting the H.264-based lossy image coding which is regarded as first layer. The quantization error of the first layer is coded by existing CABAC which is regarded as different layer. Simulation results show that proposed two-layer H.264-based lossless (Two-layer H.264-LS) image coding has better compression performance than the past coders. Finally, we suggest two simple methods to implement a fast set partitioning in hierarchical trees (SPIHT) algorithm. Simulation results show that proposed method decreases a lot of computational time and memory.