Sample learning-based rapid image super-resolution reconstruction method and device

Abstract

The invention discloses a sample learning-based rapid image super-resolution reconstruction method and device. The method comprises the following steps of: S1, obtaining training samples through a plurality of high-resolution images and low-resolution image with completely same contents to carry out model training so as to obtain a hierarchical clustering tree and a regression matrix correspondingto a plurality of image block sizes; and S2, carrying out local linear regression on self-adaptive multi-image blocks of the low-resolution images by utilizing the trained clustering tree and the regression matrix corresponding to the plurality of image blocks, so as to obtain a high-quality reconstructed high-resolution images. Through the method and device, new high-frequency information can bebetter added in images, and high-resolution images can be rapidly reconstructed at the same time.

Description

technical field [0001] The invention relates to the fields of digital image processing, machine learning and artificial intelligence, in particular to a fast image super-resolution reconstruction method and device for rapidly enlarging low-resolution images to obtain high-quality and high-resolution images by using sample learning. Background technique [0002] As an important form of information for humans to perceive the world, images are rich and detailed in their content, which directly determines the level of detail that humans perceive. The higher the pixel density on the image unit scale, the clearer the image, the stronger the ability to express details, and the richer the information perceived by humans, which is a high-resolution image. The super-resolution reconstruction of images has been studied in many fields, such as remote sensing images, satellite imaging, medical images, and some high-definition display fields. [0003] A method to improve the resolution o...

AI Technical Summary

Problems solved by technology

These interpolation methods are intuitive and simple, but often lead to blurred details and cannot increase high-frequency information;

[0007] 2. The learning-based super-resolution method can obtain prior knowledge from a large number of training sample sets as the basis for super-resolution, and can generate new high-frequency details, but the speed is too slow

Images