Video Super-resolution Reconstruction Technique And Its Application

The Video resolution in real application cannot often satisfy the high-resolutionrequirements, due to the limitations of the imaging device, data compression or the bandwidth.To solve this problem, there is proposed a super resolution algorithm, which is estimating asingle image of high quality and resolution from a set of low-resolution images.This paper begins with a brief analysis of the background and significance of the superresolution of video, and investigation of the current research related technologies. Then basedon the Bayesian estimation, the MAP super-resolution algorithm was discussed and achievedin compressed video. At the same time, the following related work carried out:1) Establish image noise probability distribution model, image motion estimation model.2) Due to the computational complexity of the MAP algorithm, three improvements wasproposed in this paper:a) Select an inexact line search with simple inexact to identify the step length, whichavoids the computation of the Hessian matrix.b) Computes the increment of the MAP objective function as the component of thegradient directly.c) Dimension reduction, parallel design are used in the actual calculation process.Experimental results shows that the improvements of the algorithm not only reduce thecomputation time, but also guarantee the solution’s convergence and maintain the similarimage quality.