鲁棒可区分的压缩视频感知哈希算法研究

Robust and discriminative perceptual hash algorithm in compressed video

运用比较宏块互异数方法得到视频关键帧,提出基于Gabor小波分解的视频感知特征快速提取算法,针对小波分解后得到的特征矩阵,给出基于负熵目标函数的FastICA优化降维量化策略,并运用中位值量化方法得到哈希位串.采用标准格式视频验证的结果显示,该算法对亮度变化、噪音污染等常规内容操作具有良好的鲁棒性能,对感知内容不同的视频序列也有较好的区分性能.研究成果可为视频版权、视频安全和视频篡改检测提供理论支撑和技术支持.

The method of comparing the different number of macro-block to get the video key frame was applied and a fast video perception features extraction algorithm based on Gabor wavelet decomposition was proposed in this paper. On account of the feature matrix resulting from wavelet decomposition, we propound the quantitative policy FastlCA, baoed on the objective function of negative entropy, to optimize dimension reduction. At last, we get the Hash bunch from the method of the medium value quantization. According the standard format video vali- dation, we can conclude that the algorithm have a good Robust performance on the normal operation of the chan- ges of lightness, noise pollution, etc. It also has a good distinguishable performance on perceiving the list in vide- os that have different contents. The restrlt of the research can provide technical support in the area of video copy- right, video security, tamper detection of video, etc.