MAP-based Technology Study And Implement Of Image Super-resolution Restoration

The goal of Super-resolution restoration is to draw a high-resolution image from a low-resolution ratio image array. Super-resolution refer to removal the image system (out of focus blur, motion blur, non-ideal sampling, etc.) and recovery of clear image that got after resurning high-frequency information of the formation of image with lost optical system. Although classical linear image restoration (such as linear and cubic spline expansion) has been thoroughly studied and implemented, because the edge area of the image and discontinous area become fuzzy.So study in guarantee image Super- resolution ratio restore result and is it test prerequisite that the speed does not reduce to charge, how establish image model, reduce computer operation amount, is it have important meaning and value to recover as to image. This text realizes from design and algorithm of the model to this question that two respects have carried on deep research and realized. The groundwork and innovation of the thesis include:First, Super-resolution re ratio deficiency that exist in the method to single width image, design one test probability estimate (MAP ) under the frame because of being most heavy, utilize the high-resolution image of the array image of many frames to construct the method again. MAP estimates that has offered the good frame for the optimum estimation of the image.convenient mathematics is asked and solved, can obtain it than the image information with more single width image, construct out the resolution ratio with higher more image than any single frame again;Second, is it design one Gibbs model to recover to image, design his parameter choose the mode. Through Gibbs model, can analyse the whole composition of the image from some characteristic of the image, having good convergence efficiency, this has brought convenience for real-time processing;Third, to the characteristic different from background of edge of body in motion in the image, in order to fully utilize redundant information and high-frequency information in the image between the array image, design one based on object edge movement estimate method, adopt different sport is it differentiate rule recover more image details to estimate. Can ask for the movement information of the part of edge of body in motion of image more accurately, make the edge area of the image and discontinous area clearer, thus image got finally to makeresult kind, is it bring more facility to recover for ultra resolution ratio of image.Based on the technique above, we successfully reconstruct a single high-resolution image from image sequences and it is more efficient than other methods.