Single-image super-resolution reconstruction method

Abstract

The invention discloses a single-image super-resolution reconstruction method used for solving the technical problem that image reconstruction precision is low in an existing image super-resolution reconstruction method. According to the technical scheme, firstly, high / low-resolution image blocks are extracted through a great number of high-resolution images and serve as a dictionary; then, according to input image blocks, the low-resolution image blocks in the dictionary are selected to conduct calculation of a deformation field; finally, the corresponding high-resolution image blocks in the dictionary are deformed. A final high-resolution image is obtained through local restriction and global restriction. By means of the deformable image blocks, the expression capability of the dictionary is greatly enhanced, and therefore the final reconstruction effect is improved. The 30000 7*7 image blocks are selected to serve as the dictionary. When the extraction step length of the image blocks satisfies the equation that S=1, and super-resolution reconstruction with the enlargement factor being 3 is conducted on a 256*256 standard test image which is a Lena image, reconstruction precision satisfying the equation that PSNR=31.53 can be achieved and is higher than the reconstruction precision satisfying the equation that PSNR=29.68 in documents.

Description

technical field

[0001] The invention relates to an image super-resolution reconstruction method, in particular to a single image super-resolution reconstruction method. Background technique

[0002] High-resolution images have important application value in criminal investigation, behavior monitoring, target recognition, and medical image processing. On the premise of not changing the existing image sensor and imaging equipment, it is of great significance to improve the resolution of the image by using the method of super-resolution reconstruction. The existing single image super-resolution reconstruction methods mainly include: methods based on interpolation, methods based on manifold learning, methods based on sparse coding and methods based on self-similar image blocks.

[0003] The document "Single-frame and multi-frame image super-resolution reconstruction based on locally constrained linear coding, Journal of Jilin University (Engineering Science Edition), 2013, Vol....

Images