本文提出了一种基于H.264/AVC压缩域的GOP(Group of Pictures)级视频场景转换检测算法.该算法利用H.264/AVC基本档次码流中的帧内预测模式、运动矢量和宏块编码类型等可用信息,提出了基于子块的色度模式差异、累积运动值和累积帧内宏块数等三个判决准则,然后综合利用这三个判决准则,提出了一种GOP级的视频场景转换检测算法.实验结果表明,与现有的一个COP级场景检测算法对比,本文提出的GOP级视频场景转换检测算法可以获得更好的检测性能.
提出了一种直接在JPEG图像压缩域进行肤色检测的算法.该算法首先在熵解码后的DCT系数中提取图像块的颜色特征和纹理特征,然后利用数据挖掘建立用于表征压缩域图像特征和肤色检测结果之间关系的肤色模型,并利用该模型进行初步肤色检测,最后利用区域生长的方法分割出图像中的肤色区域.实验结果表明,与像素域的SPM(Skin Probability Map)肤色检测算法相比,本文方法可以获得更高的检测准确率和更快的检测速度.
针对无线视频传感器网络(Wireless video sensor networks,WVSN)对视频编码算法的具体需求,提出一种基于运动检测的低复杂度视频编码算法。该算法只对当前编码帧中的运动对象进行编码,并且以面向对象的结构输出码流。实验结果表明,与H.264全I帧编码相比,本文提出的算法编码速度提高了约3倍,编码性能提高了约2 dB。与H.264基本档次相比,虽然编码性能略有下降,但是编码速度平均提高了8倍左右。本文提出的算法可以在编码效率和编码速度之间获得很好的折衷,在一定程度上可以满足WVSN的需求。
In this paper, the complexity of intra coding is first analyzed so as to achieve a weight of complexity measurement for each intra mode. Then, a new complexity scalable control algorithm for intra coding in H. 264 is proposed, based on the rearrangement of the order of candidate modes and an efficient complexity allocation and control (CAAC) scheme at the macroblock (MB) level. The candidate modes of each MB are rearranged according to the local-edge information. Experimental results show that our proposed algorithm can make an appropriate cut-off point of the candidate modes sequence adaptively according to the current energy condition of a mobile device, so as to adjust the complexity at any level while maximizing the video quality, which can prolong the operational lifetime of the battery with minimum degradation in video quality.