Image super-resolution reconstruction method of visual vocabularies and based on texture context constraint

Abstract

The invention discloses an image super-resolution reconstruction method of visual vocabularies and based on texture context constraint and belongs to the field of image super-resolution research. The low-resolution visual vocabulary to the high-resolution visual vocabulary one-to-one mapping geometrical characteristic symbiotic prior knowledge is obtained through combining and aggregating to lead the similarity of geometrical characteristics to keep between each pair of low-resolution visual vocabulary and high-resolution visual vocabulary, the postulated condition based on an instance method is met, and detail accuracy of restored high-resolution images is ensure; and a two-step image super-resolution reconstruction frame is built, the frame effectively accesses the visual vocabulary which the real high-resolution images belong to through a texture context characteristic in a maximum a posteriori method, and high-resolution image details are constructed through low-resolution images by aid of subspace constraint and reconstruction constraint. Compared with the existing method, the high-resolution images extracted in the image super-resolution reconstruction method are real and clear in details and accord with the practical conditions.

Description

technical field [0001] The invention relates to the field of image super-resolution research, in particular to an image super-resolution reconstruction method based on visual vocabulary constrained by texture context. Background technique [0002] In practical applications, due to the long distance between the camera and the scene of interest, or the low resolution of the captured image due to hardware reasons, the number of pixels of the scene of interest in the image is very small, resulting in a lack of detailed information, resulting in It is difficult to identify and extract regions of interest from images. Image super-resolution research technology has attracted widespread attention because it can solve the above problems. [0003] The traditional image super-resolution reconstruction method is to estimate the high-resolution image through multiple low-resolution images of the same scene, but this method is difficult for the alignment of multiple low-resolution images...

AI Technical Summary

Problems solved by technology

Traditional interpolation methods and reconstruction methods need to cooperate with prior knowledge, but the commonly used prior knowledge about edges cannot construct high-resolution edge details from most smooth image regions of low-resolution images; instance-based methods although Lost high-resolution image detail information can be recovered by learning high-resolution and low-resolution image patch pairs, but instance-based methods assume that similarity between low-resolution images implies similarity between corresponding high-resolution images Not necessarily true in practice, so the accuracy of the restored high-resolution image details cannot be guaranteed

Images