Image Super-resolution Via Centralized Sparse Representation

In the process of acquiring the image, the image quality will be reduced. Although thetraditional image restoration technology will improve the quality of the acquired image on acertain extent. Since this method can not restore the image details that have been lost in theprocess of image. So a new solution has been proposed, that is super-resolution restorationtechniques. Image super-resolution recovery techniques is aim to try to restore thehigh-frequency details that have been lost in the imaging system, so as to improve the quality ofthe image.In this paper, we generally discuss the image super-resolution algorithm based onconcentrated sparse representation. Firstly, we introduced a new image noise, which is statisticalsparse representation of the image noise. By using the statistical data, it is found that thedistribution of noise is very similar to the Laplace distribution in the image. The restored imageis closer to the original high resolution image by using this kind of noise. Then, a new priorimodel is introduced that is an AMRF prior model. The new priori model can effectively reducethe oscillation stripes and noise in the restored image. Finally, a new adaptive PCA dictionary isintroduced, which is based on image block. The new dictionary is not only considering the localcharacteristics of the image, but also considered a global characteristic of the image.The new restoration algorithm not only guarantees the image sparse representation of therationality and fidelity, but also efficiently use the similar image block. The experimental resultsindicate that the new image restoration algorithm has good subjectivity and objectivity.