Residual-based ultra-resolution image reconstruction method

Abstract

The invention relates to a residual-based ultra-resolution image reconstruction method, which specifically comprises the following steps of: first calculating residuals between original high-resolution images and images obtained by performing interpolation amplification on low-resolution images; then establishing sample pairs by using the characteristics of low-resolution image samples and corresponding image residuals, classifying the sample pairs by taking the low-resolution image samples as references and adopting K-averaging, and training each type of sample pair by adopting a K-singular value decomposition (K-SVD) method to obtain dictionary pairs of the low-resolution image samples and the image residuals; and finally selecting a dictionary pair according to a Euclidean distance between a test sample and a type center, calculating the weighted sum of image residuals reconstructed by each type with similar Euclidean distances with the test sample as a final reconstructed image residual, and obtaining a high-resolution image by combining interpolation results of the low-resolution images. Only the image residuals are required to be reconstructed, and the high-resolution image can be reconstructed by combining the interpolated images, so that edge detail reconstruction results of the high-resolution image are improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a method for performing super-resolution reconstruction on an image, in particular to a residual-based image super-resolution reconstruction method. Background technique [0002] Image super-resolution reconstruction refers to the method of reconstructing a high-resolution image from one or more input low-resolution images, which makes full use of the obtained resources and uses high-performance hardware to obtain high-resolution images. Compared with other methods, it has a lower cost and has broad application prospects in various fields such as video surveillance, medical imaging, and high-definition video. [0003] Super-resolution reconstruction is a typical ill-conditioned inverse problem. When there is only a single frame of low-resolution images, it is transformed into a problem of solving underdetermined equations. In order to improve the reconstruction performanc...

AI Technical Summary

Problems solved by technology

In order to improve the reconstruction performance, prior knowledge can be considered in the solution process to eliminate many artifacts, but satisfactory results cannot be achieved when reconstructing images with rich details; in order to reconstruct visually pleasing high-resolution images , a lot of prior knowledge is applied to super-resolution reconstruction. The typical method is regularized super-resolution reconstruction. The well-known regularization prior is to use full variation. This method can obtain better results for piecewise smooth edge structures. Good reconstruction results, but there are usually stepped artifacts for edges with rich details; the recently proposed prior knowledge is that the image has sparsity, and the image is expanded into a sparse expression by the atoms of the dictionary, which, as the name implies, is most of the coefficients of the expansion coefficient is close to 0, and the learned dictionary is used for super-resolution reconstruction. The typical method is to learn the high-resolution and low-resolution dictionaries of the image, and for the low-resolution image sub-block of the test, solve its Represent the coefficients, and then use the manifold consistency between the low-resolution sub-block and the high-resolution sub-block to reconstruct the high-resolution image sub-block using the representation coefficient of the low-resolution sub-block, but the details at the edge of the image will also be reconstruction noise

Images