Key frame extraction using rough set theory for video retrieval

Abstract

Key frame is a representative frame which contains the entire information of the shot. It is used for indexing, classification, analysis and retrieval of video. The existing algorithms generate relevant key frames but they also generate a few redundant key frames. Some of them are not able to represent the entire shot since relevant key frames are not extracted. We have proposed a more effective algorithm based on DC coefficients and Rough Sets to prevail over the rest. It extracts the most relevant key frames by eliminating the vagueness of the selection of key frames. It can be applied for compressed MPEG videos hence decompression is not required. The performance of this algorithm shows its effectiveness.