A system and method for image super-resolution reconstruction utilize standardized group sparsity regularization

Abstract

The invention discloses an image super-resolution reconstruction system and a method which are regularized by sparse standardization group, and relates to the technical field of image restoration. Theinvention firstly adopts bilinear interpolation method to obtain initial estimation value of image super-resolution reconstruction; Then, by using the normalized sparse prior of image block group, the quality of super-resolution image reconstruction is improved by iterating the adaptive soft threshold filtering in PCA domain and the regularized least squares in pixel domain of image block group.The system comprises an input low-resolution image (00), an initialization module (10), a routing module (20), an image filtering module (30), an image reconstruction module (40) and an output high-resolution image (50). The image reconstruction module (40) also interacts with the routing module (20). The invention can improve the restoration effect of the high-frequency detail of the image and effectively improve the quality of the super-resolution image reconstruction. The invention is Suitable for video surveillance, medical imaging and other applications.

Description

technical field [0001] The present invention relates to the technical field of image restoration, in particular to an image super-resolution reconstruction system and method using normalization group sparse regularization. Background technique [0002] Image super-resolution aims to restore high-resolution images from low-resolution images, and it has a wide range of applications in medical imaging, digital photography, and computer vision. The traditional interpolation-based super-resolution method is based on the assumption of image continuity, and uses the prior knowledge of similarity of adjacent pixels to generate the gray value of the interpolated pixel by weighting and summing the gray values ​​of adjacent pixels. Interpolation-based super-resolution methods have the advantage of being computationally simple. Traditional interpolation methods include: nearest neighbor interpolation, bilinear interpolation, bicube interpolation, etc. [See literature: [1] Keys R., 'Cu...

AI Technical Summary

Problems solved by technology

Previous methods usually assume that the sparse transformation coefficients have zero-mean statistical properties. However, due to the general non-stationary properties of image signals, the statistical properties of non-locally similar image blocks in different regions will be significantly different. The sparse representation based on the zero-mean assumption limits the Efficient reconstruction of image high-frequency texture detail information

Images