【摘要】 提出了一种基于多特征融合的新闻节目高效检索方法。该方法充分利用媒资管理系统中新闻视频的文字描述和关键帧的图像特征,能够实现视频节目快速、准确的检索,提升新闻检索的性能。将文字描述、人脸和场景图像作为搜索项,通过Fisherfaces和LBP算子相结合的人脸识别算法,以及基于重要性加权的局部直方图匹配算法,实现样例图像与关键帧的匹配。实验结果表明,该方法能够提升媒资管理系统中新闻节目检索的准确度和效率,满足新媒体时代节目快速制作和发布的要求。更多还原

【Abstract】 In this paper,an efficient news retrieval method is proposed based on multiple features,which makes full use of word description and features of keyframes in MAM( Media Asset Management System),achieves fast and accurate retrieval,and improves the performance of news retrieval in MAM. We take the text description and images which contain important faces and scenes as input for video retrieval,and then match the input image with keyframes in MAM by applying the algorithm for face recognition based on Fisherfaces and LBP operators,and the local histogram match algorithm based on the weighted importance. Experimental results show that the proposed method can enhance the accuracy and efficiency of news retrieval in MAM,and meet the requirements for rapid production and release in new media age.更多还原