Near-duplicate keyframes detection in broadcast domain based on Bayesian information criterion
基于BIC的新闻视频近似重复帧检测方法

Near-duplicate keyframes detection was an important part in content-based broadcast video search. After the definition of the acoustic change detection in audio segmentation was introduced, a new method based on Bayesian Information Criterion (BIC) algorithm was proposed for detecting near-duplicate keyframes, which result from the feature of salient parts extracted from a frame. Experimental results show that this method without threshold and learning has good performance in recall and precision.