Study Of Keyframe Extraction Algorithm Based On SIFT Features

With the rapid development of multimedia technology and information technology, video data are becoming increasingly rich. Hence how to manage and use these mass videos efficiently has become an urgent problem. Using keyframes can not only greatly reduce the amount of video retrieval data and video storage, but also lay the foundation for video summary and video retrieval.Common methods of keyframe extraction include the way based on color,texture features,shot boundary and the clustering. However, their fidelities or compression ratios are not good. SIFT feature descriptor, a scale-invariant feature transform descriptor, is invariant to uniform scaling, orientation, and partially invariant to affine distortion and illumination changes. The effect of keyframe extraction can be improved by SIFT features.By analyzing the algorithm of SIFT features extraction, a new method of keyframe selection is proposed, and it is applied to the keyframe extraction algorithm which is based on SIFT in this paper. Then the experiment is carried out to verify the effectiveness of the algorithm on the videos of TRECVID database. In order to improve the real-time performance of the algorithm, a keyframe extraction algorithm based on PCA-SIFT and its CUDA version are designed in this paper. Finally, the two algorithms of keyframe extraction are tested by experiments. The experimental results show that the accelerated algorithm of keyframe extraction based on PCA-SIFT has a better real-time performance under the premise of guaranteeing a higher fidelity and compression ratio.