摘要
提出了一种基于三维小波的视频片断检索的新方法,该方法克服了传统视频检索方法中未考虑的时空关系。首先利用低频子图小波系数的标准方差均值对目标视频库进行初步筛选,然后对利用八叉树算法提取细节特征,进一步的把细节特征和近似特征组合起来对图像进行精确的检索。实验证明,该方法具有更好的检索精度。
A new approach for video clip retrieval based on 3-D wavelet is presented. It overcome the relations between temporal and spatial. First, thetargetvideo library is filteredbytheaverageo of wavelet-coefficientstandarddeviation of theapproximation sub-image.Then, the detail feature is extracted by the octree algorithm. Finally, retrieving the video clip accurately by combining the detail features with the approximate features. Experimental results indicated that the proposed approach is efficient in retrieving video clips.
出处
《计算机工程与设计》
CSCD
北大核心
2007年第5期1100-1101,1178,共3页
Computer Engineering and Design