An image super-resolution reconstruction method focusing on foreground information

Abstract

The invention discloses an image super-resolution reconstruction method focusing on foreground information. The foreground information and high-frequency features of the image are extracted through the PAM module, and the channel domain attention and spatial domain attention weight coefficients are extracted by using the gating network. The joint loss dynamically modifies the weights of both during backpropagation. PAMNet is further proposed, multiple PAM modules are connected in series in PAMNet, and skip connections are introduced at the same time, making full use of the shallow features of the image, training through the designed network, and completing the reconstruction of the super-resolution image. The present invention can not only focus on image foreground information and identification feature extraction, but also retain the color and texture features of the image, and improve the utilization rate of the shallow layer; the present invention can reduce the amount of parameters and have better objective scoring; the present invention is complex in performance and model A good balance has been achieved between degrees, and the PAM module is versatile and can be embedded in many types of network structures.

Description

technical field [0001] The present invention relates to the technical field of image processing and recognition methods, in particular to an image super-resolution reconstruction method focusing on foreground information. Background technique [0002] Image super-resolution reconstruction (Super-Resolution, SR) is an image processing technology that uses low-resolution image restoration to obtain high-resolution images. SR technology aims to improve image resolution through signal processing and software methods without changing the limitations of physical imaging equipment. SR not only has important academic research value, but also has practical applications in many fields, such as medical imaging, video security monitoring, Remote sensing image processing, etc. In addition to improving image quality, SR technology can also improve many computer vision tasks. Improving image resolution to obtain high-quality images has become an urgent problem in the research community. ...

AI Technical Summary

Problems solved by technology

[0006] In view of the above-mentioned super-resolution tasks in image processing still have the problems of not fully utilizing the shallow features of the image, the foreground of the image is not prominent, and the lack of visual emphasis, the present invention proposes an image super-resolution reconstruction method that focuses on foreground information, focusing on image foreground information and While reducing the amount of training parameters for detailed features, the utilization rate of image shallow features and model performance are improved, and the visual quality of image super-resolution is greatly improved.

Images