A content-based approach for deciding the key frames in computer vision

The utilization of the key frames in computer vision would economize the storage capacity and also reduce the search amount of image information. In addition, visual processes can be recovered via the key frames. The content-based key frame in computer vision is defined, and a content-based approach for deciding the key frames in computer vision is proposed in this paper. Level-set-based geodesic active contours method is adopted in this approach to acquiring the accurate contours of the moving object. The normalized contours of the moving objects are transformed with wavelet. The whole shapes of the contours are described by scale coefficients of wavelet, and details of the contours are particularly described by the parameters of wavelet coefficients Gaussian density distribution. The contents of the vision are understood from the whole and details. The experiments show that the approach is effective and practicable.