利用反向投影的flash场景自适应视频编码算法  

Adaptive Video Coding Algorithm for Flash Scene Using Back Projection

作　　者：梁传君[1] 徐亮[1] 

机构地区：[1]新疆工程学院计算机工程系,乌鲁木齐830011

出　　处：《计算机测量与控制》2017年第7期243-245,共3页Computer Measurement &Control

基　　金：新疆维吾尔自治区高校科研计划青年教师科研启动基金项目(XJEDU2014S074);新疆工程学院科研基金项目(2015xgy101712)

摘　　要：针对闪光造成的光照变化会导致视频帧之间巨大的强度差异问题,提出利用反向投影的flash场景自适应视频编码算法;根据直方图差异提取闪光和非闪光帧,相应地为每个帧分配适当的编码类型,并在加权预测(WP)参数集确定中采用运动向量导数,通过反向投影保证flash场景的全局一致性;实验结果显示,提出的算法在Lena、Peppers、Building、Baboon、Nestling 5个视频上的峰值信噪比(PSNR)值分别可高达32.31dB、34.14dB、34.76dB、34.94dB、35.05dB,非常接近原始图像的PSNR;相比其他几种加权预测算法,提出的算法在PSNR及计算复杂度方面均获得了更加优越的编码性能。Aiming at the huge intensity difference between video frames caused by the illumination change of flash, adaptive video coding algorithm based on flash scene is proposed in this paper. The flash and non flash frames are extracted according to the histogram difference, corresponding coding types are allocated for each frame correspondingly, and the motion vector derivative is used in the weighted prediction （WP） parameter set determination. And the back projection is used to keep the global consistency of flash scene. The experimental results show that the peak signal to noise ratio （PSNR） achieved by the proposed algorithm on the five videos Lena, Peppers, Building, Baboon and Nestling can arrive at 32.31 dB, 34.14 dB, 34.76 dB, 34.94 dB and 35.05 dB, respectively, which is nearly to PSNR of the primary images. And compared with the traditional weighted prediction algorithm, the proposed algorithm achieves better coding performance in terms of PSNR and computational complexity.

关 键 词：Flash场景 自适应编码 加权预测 峰值信噪比 反向投影 计算复杂度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]