基于粒子滤波目标跟踪的FGS编码  

FGS Video Coding based on Motion Tracking Particle Filter

作　　者：周小黎[1] 林其伟[1] 

机构地区：[1]华侨大学信息科学与工程学院,福建厦门361021

出　　处：《通信技术》2013年第4期114-116,120,共4页Communications Technology

基　　金：福建省自然科学基金项目:基于视觉特性的鲁棒视频水印算法研究(No.2012J01275)

摘　　要：针对感兴趣区域优先编码的问题,提出了基于粒子滤波目标跟踪的精细可分级视频编码(FGS,Fine Granularity Scalable),该方法结合开源计算机视觉库(OpenCV),在增强层引入粒子滤波运动跟踪算法,人为定义出一个感兴趣的区域,对此区域进行跟踪,提取,位平面提升,优先传输和编码,并控制优先编码的位平面数,控制预测漂移,从而改善视频解码后的主观视觉效果。同时,在基本层引入单环算法来提高编码效率。实验结果表明,提出的算法在一定的码率范围内感兴趣区域(ROI)的PSNR值高于整体区域的PSNR值。For the preferential encoding region of interest, the FGS （Fine Granularity Scalable） coding strategy based on particle filter video tracking is improved. This strategy combines Open CV （Open Source Computer Vision Library） and introduces the particle filter video tracking algorithm into the enhanced layer. An area artificially defined as an area of interest is tracked and extracted, and hit-plane upgrade, preferential transmission and coding are also done, including the control of bit-plane number and prediction drift, thus to improve subjective visual effect after video decoding. Meanwhile a single-loop algorithm is introduced into the basic layer for improving the objective coding efficiency. The experiment shows that the PSNR value of ROI （region of interest） is higher than that of overall regional within certain hit-rate range.

关 键 词：粒子滤波 感兴趣区域 位平面 可分级编码 

分 类 号：TN919.8[电子电信—通信与信息系统]