Sparse representation of image and video using easy path wavelet transform

摘要

This paper utilizes a wavelet based image compression technique, Easy Path Wavelet Transform (EPWT), for sparse representation of video. The algorithm has been extended for several aspects including Rigorous EPWT for three dimensional data compression and Relaxed EPWT for cost minimization. The proposed algorithm when applied on sampled frames of a video segment operates along permutations of frame indices exploiting the local correlations efficiently. These arrays of rearranged frame indices, in the case of redundant frames, are then uniquely stored while taking the repetition into consideration. The favorite direction criterion in the encoding process has also been modified to achieve enhanced sparsity for video segments while improving the existing quantitative results for images. The extended EPWT is effective and experimental results are comparable with existing state of the art techniques.