Compressive Sensing-based Video Recovery Using the Multidirectional Total Variation Minimization

Video processing that using compressive sensing involves acquiring some of the linear measurements of the video as its samples and recovering the video utilizing the samples. In the process, the video is generally regarded as a time varying signal that the correlation of its adjacent frames is recommended to be used, for the purpose of improving the quality of the compressive sensing recovery results. Based on the fact that the two dimensional total variation minimization method may cause the loss of the texture of an image, the diagonal and back-diagonal directional differentiation are added in the calculation of the two dimensional total variation in this paper. Then this method is utilized to minimize the three dimensional total variation for the recovery of compressive sensing video, and the experiments on the recovery of the compressive sensing videos illustrate that better performance can be obtained, compared with when the classical three dimensional total variation minimization method is used.