Feature enhancement-based residual neural network and image deblocking effect method

Abstract

The invention relates to the field of digital image post-processing, in particular to a feature enhancement-based residual neural network and an image deblocking effect method. The residual neural network based on feature enhancement introduces a local residual unit, a global feature enhancement unit and a local feature enhancement unit. The three basic units promote each other, so that the learning ability and the modeling ability of the target neural network are greatly enhanced. Accurate mapping from a low-quality image with a blocking effect to a high-quality image can be established for an image blocking removal effect problem, and finally, a JPEG compressed image with given quality can be processed through the established effective mapping to obtain a high-quality image. According tothe image deblocking effect method, the peak signal-to-noise ratio (PSNR) and the structural similarity (SSIM) of the image can be remarkably improved, the efficiency, quality and robustness of imagedeblocking are greatly improved, and the method has profound significance in the field of image post-processing.

Description

technical field [0001] The invention relates to the field of post-processing of digital images, in particular to a residual neural network based on feature enhancement and an image deblocking method based on the residual neural network. Background technique [0002] Multimedia information mainly has three forms of expression, namely text, sound and image; among them, image is the most common information storage method, its expression form is vivid and intuitive, and can provide more information than other forms of data; however, image is the three In the form of information, the data volume is the largest. If it is not compressed, it will cause huge pressure on data transmission and storage. The main task of image compression is to remove various redundant and irrelevant information and retain useful information; convert a large data file into a smaller file to represent the image with as few bits as possible; while maintaining the decoding image quality so that it meets th...

AI Technical Summary

Problems solved by technology

In the document "Compression Artifacts Removal Using Convolutional Neural Networks", CNN was first applied to the problem of image deblocking effect. Compared with the traditional algorithm, it achieved a good performance gain. However, due to the simple network structure, the feature information of the image could not be fully extracted.

In the document "An efficient deep convolutio nal neural networks model for compressed image deblocking", a deep CNN was constructed, which fully reflects the excellent modeling ability and learning ability of the neural network, but blindly stacking the network depth will cause new problems; Poor technology can train a very deep network structure, but as the network depth increases, the number of features extracted by the network increases, and the use of cascade operations produces feature redundancy, which hinders the improvement of network performance.

Images