Single-frame image super-resolution reconstruction method based on Euclidean subspace group double mapping

Abstract

The invention belongs to the technical field of image processing, and discloses a single-frame image super-resolution reconstruction method based on Euclidean subspace group double mapping. The methodcomprises the following steps: firstly, generating an initial sample training set by utilizing natural images, and performing simplification and classification on the initial sample training set to obtain a screened sample training set; secondly, extracting features from the screened sample training set, generating a feature training set, and clustering the feature training set to form an Euclidean subspace group; sequentially calculating a first remapping coefficient from the low-resolution image block to the high-resolution image block in each local subspace of the Euclidean subspace groupby using a linear regression model; applying the first remapping coefficient to each low-resolution image block in the Euclidean subspace group to reconstruct a to-be-selected high-resolution image block, and calculating a second remapping coefficient from the to-be-selected high-resolution image block to the real high-resolution image block; and finally, generating a test sample set by the test image, and outputting a high-resolution reconstructed image by applying the double mapping coefficients. The algorithm software and hardware of the method are low in cost; independent and mobile imageprocessing platforms are also applicable.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a single-frame image super-resolution reconstruction method based on double mapping of Euclidean subspace groups. Background technique [0002] In the information age, images, as an important carrier of information, have received more and more attention. The higher the resolution of the image, the richer the information it can provide. However, in practical applications, due to the physical conditions of the imaging system and the impact of the objective environment, the imaging process is often affected by factors such as motion blur, downsampling, and noise, which often make the resolution of the final image cannot meet the requirements. In order to solve this contradiction, the image super-resolution reconstruction algorithm came into being; this algorithm refers to using a single frame or multiple frames of low-resolution images of the same scene to estimate a high-...

AI Technical Summary

Problems solved by technology

[0008] To sum up, the existing problems in the existing technology are: the existing model does not divide the image blocks in the training set, and inappropriate feature matching may occur; Some models describe the mapping relationship between high-resolution and low-resolution images too simply, which may affect the recognition of details in super-resolution reconstructed images and reduce the quality of reconstruction

Images