Video Retrieval Based On Computational Intelligence

Recently, with the sharp development of computer networks and multimedia techniques, the amount and types of information over the Internet are growing day by day. How to get the desired information from the huge multimedia library is becoming a very important subject. For this purpose, it educes a new field about video processing: content-based video retrieval. This paper has done the following groundwork around this field:1 Key frame extraction. A new key frame extraction algorithm based on evolutionary artificial immune network is proposed in this paper. It turns video frames into points in multi-dimension space, and similar frames will be designated to the same cluster by evolutionary immune network clustering. Then the frame closest to the centroid of the cluster is extracted as a key frame. Experimental results have shown the reliable performance of the proposed algorithm. It can extract key frames automatically and doesn't require the number of clusters to be known beforehand. In addition, it can be applied to video sequence directly and doesn't require shot segmentation.2 motion estimation. Based on studying existing block motion estimation algorithms, we apply immune clonal selection to the searching strategy of block motion estimation and a block motion estimation based on immune clonal selection algorithm is proposed in this paper. Simulation results demonstrate the algorithm achieves accurate matching in low computational complexity.3 video classification. Based on the combined features of color and motion, we apply support vector machine to video classification. At the same time, a method called center distance ratio is utilized to select support vectors from given training examples, which can reduce the training samples and speed support vector machine. It is shown that thealgorithm is feasible and effective with the results of computer simulations.