基于稀疏贝叶斯模型的视频恢复方法研究

Video Retrieval Method Using Sparse Bayesian Modeling

结合压缩感知的稀疏表示理论和贝叶斯模型,提出一种在数据库中进行快速有效的视频信息查询和视频恢复的方法。根据最小l1范数重构数据信息的要求,方法利用贝叶斯模型对参数进行估计,并通过稀疏特性对视频分类和特征提取,达到恢复视频信息的目的。实验结果表明,比较主成成份分析和随机投影算法,该文的实现方法具有更好的恢复性能。

Combining Compressed Sensing sparse theory with Bayesian modeling , a faster and more effective method of video information retrieval is proposed in database. Based on minimizing l1 principle for classification and feature extraction , the method implements video retrieval requirement by utilizing Bayesian modeling to automatically estimate the regularization parameters. Experimental results demonstrate that the proposed method performs better retrieval performances than PCA and RP algorithm.