Font Size: a A A

Image Super-resolution Restoration Using POCS Algorithm Based On MAP

Posted on:2018-09-01Degree:MasterType:Thesis
Country:ChinaCandidate:L L WangFull Text:PDF
Abstract/Summary:PDF Full Text Request
Image super-resolution restoration technology is to obtain a high resolution image with more details and better visual effect than the original image using one low resolution image or the complementary information between multiple low resolution images through the corresponding restoration algorithm.At present,at home and abroad,the super-resolution restoration technologies generally use sequence images for restoration,that is,to remove the image degradation factors and then merge multiple images into one high resolution and high quality image.Image super-resolution restoration technology has currently been widely used in computer vision,image compression,medical image processing,remote sensing images,military images acquisition and collection,digital TV conversion,biometric identification and so on.Making a general survey of the research situation at home and abroad,the study on image super-resolution restoration has got certain outcomes,especially the widely used algorithms of projection onto convex sets and maximum a posteriori.However,there are still some problems for the existing methods such as the not-unique restoration result,the poor edge preserving ability and the lost image detail.Considering the existing problems,in order to further improve the quality of the super-resolution restoration image,we improve the algorithms of projection onto convex sets super-resolution and maximum a posteriori in the thesis.Innovation and research contents mainly include the following three aspects:?The super-resolution restoration algorithm of projection onto convex sets based on the modified point spread function is proposed.In order to maintain the image edge features and improve the anti noise ability,the traditional point spread function is modified by adding a weight factor.The point spread function is divided into eight directions,which makes the point spread function of edge direction unchanged,and the weight factor of edge orthogonal direction decreases so as to achieve the effect of enhancing the image edge preserving ability and improving the image quality.?An improved maximum a posteriori super-resolution restoration algorithm is proposed.The key content of this part is the selection of probability model.Different from the traditional way,the Gibbs random field model is selected as the priori probability model in this thesis.And in order to preserve the detail information of the image as much as possible,we modify the coefficients of the Gibbs term in the model.The conditional probability model is used to choose Gauss-Markov random field model.In order to reduce the image edge information loss caused by over-smoothness,the image edge information is extracted and enhanced firstly.Then,the image is added to the improved maximum a posteriori algorithm for super-resolution restoration,which further increases the detail information of the image.?The POCS super-resolution restoration algorithm based on maximum a posteriori(MAP)is proposed.The two algorithms of POCS and MAP are combined to fully suppress the shortcomings and play their own advantages.At first,we use the MAP super-resolution restoration algorithm to estimate several low resolution unprocessed images,and obtain an initial high resolution image which is used as the initial reference image of POCS super-resolution restoration.And it is preprocessed using the eight directions of Sobel operators and multi directional orphological filter.Then it is added to the POCS restoration algorithm based on the modified point spread function proposed in the first part to start iteratively calculating.The corresponding convex set projection operator is obtained by priori information such as energy boundedness.The improved POCS algorithm is used iteratively to get the solution until the set threshold or the set number of iterations is reached and the final super-resolution restoration image is obtained finally.Using the software simulation,we can obtain the results of different super-resolution restoration image under different algorithms.The quality of the restored super-resolution image is evaluated from several aspects: the subjective evaluation method and the objective evaluation of the peak signal to noise ratio,the mean square error and the structural similarity and other indicators.The experimental results show that the proposed algorithms can effectively improve the quality of super-resolution restoration image.
Keywords/Search Tags:Projection onto convex sets, Super-resolution restoration, Maximum a posteriori, Point spread function, Probability model
PDF Full Text Request
Related items