Research Of Image Super-resolution Reconstruction

In this paper, several major reconstruction algorithms and improved methods aiming atexistent problems are studied. The paper deals with the following tasks:Firstly, the origination and basic theory of super-resolution reconstruction technology areintroduced. On that basis, a general imaging degradation model about low resolution imagescapturing process is drawn into in this paper. Then, several major super-resolution reconstructionalgorithms are introduced and compared simply.Secondly, aiming at the fuzzy phenomenon of reconstruction image which is caused by theintroduction of regularization item, an improved MAP algorithm is proposed based onmorphological edge preserving. In the improved algorithm, a morphology edge preserving operatoris designed, according to the dilation and erosion mathematical morphology. Then the operator isdrawn into the regularization item to improve the clarity of the edge details and restrain smoothfuzzy phenomenon. Experimental results demonstrate that the improved method has a strong abilityof keeping edge and the reconstruction image has better visual effect.After that, aiming at the serious edge noise phenomenon of reconstruction image in originalPOCS algorithm, an improved algorithm based on projection modification is proposed in this paper.In the improved algorithm, a projection modification operator, designed by combining the edgeconstraint with the difference between prior and current reconstruction results, is used to modify theresidual threshold and point spread function, so that the POCS reconstruction process is modified,simultaneously. Experiments are conducted with test images with noise and the other test imageswithout noise, separately. Experimental results demonstrate that the improved algorithm achievessatisfactory performance both in subjective vision quality and objective data quantitative analysis.In the following, in order to suppress and reduce the ringing artifacts of the IBP superresolution algorithm, an improved IBP reconstruction algorithm is put forward in this paper. Afteranalyzing the reason for ringing artifacts, a self-adapting analysis is drawn into the back projectionprocess according to the variation on error projection matrix and local variance variation on errorprojection matrix. In the reconstruction process, the error projection matrix is modified based onthe analysis results. Experiments are conducted with test images with noise and the other testimages without noise, separately. Better visual effect and ringing artifacts suppression effectivenessare demonstrated by experiments. The results show that the algorithm achieves satisfactory performance.According to the adequate study on the super-resolution algorithms, a super-resolutionreconstruction experimental software based on OpenCV is designed. Three classical reconstructionalgorithms, including MAP, IBP and POCS, are implemented in the software. Besides that, theparameters in three algorithms can be controlled also.