Image super-resolution reconstruction method based on convolutional neural network

Abstract

The invention relates to an image super-resolution reconstruction method based on a convolutional neural network, and the method comprises the steps: training an SRCNN convolutional neural network model through a data set, and obtaining the shallow texture feature information; establishing an eight-layer end-to-end neural network model based on feature transfer, and migrating shallow texture feature information to the first four layers of the neural network model to obtain model parameters of the first four layers; obtaining model parameters of four rear layers of the neural network model, andenhancing learnt characteristics; inputting image data to be reconstructed, and preprocessing the image data; obtaining a high-resolution image of the Y channel; and fusing the high-resolution imageof the Y channel, the image of the Cb channel and the image of the Cr channel to obtain a reconstructed image. According to the convolutional neural network model provided by the invention, a better super-resolution result is obtained, the subjective vision and objective evaluation indexes are obviously improved, the image definition and the edge sharpness are obviously improved, the convergence speed is higher, and the method has higher advantages in the aspect of fineness.

Description

technical field [0001] The invention relates to the fields of computer vision and digital image processing, in particular to an image super-resolution reconstruction method based on a convolutional neural network. Background technique [0002] Single image super-resolution (Super Resolution, SR) reconstruction is a classic problem in the field of computer vision, and its purpose is to obtain a high-resolution image from a low-resolution image. Through the method of signal processing and image processing, the problem of reconstructing a high-resolution (High Resolution, HR) output image with the highest quality. There are multiple solutions for any given low-resolution pixel image. This is a typical ill-conditioned inverse problem with serious ill-posedness, and prior information is needed to solve this problem. [0003] Convolutional neural network (CNNs) can obtain high-frequency features of images, enhance detailed information, and is particularly good at obtaining correl...

AI Technical Summary

Problems solved by technology

However, the three-layer structure of SRCNN cannot meet the requirements of higher fineness. According to the characteristics of convolutional neural network, the first few layers can only obtain the shallow texture information of the image, and only the deeper convolutional network can be used to obtain more detailed features. to reconstruct high-quality images

Images