Image super-resolution reconstruction method based on deep neural network

Abstract

The invention discloses an image super-resolution reconstruction method based on a deep neural network, and the method comprises the steps: inputting an LR original image into a feature extraction module, extracting a feature component from the LR original image after the LR original image passes through the feature extraction module, and transmitting the extracted feature component as an input to an iteration module, wherein the iteration module is composed of a residual network, a reconstruction network, a convolution network and a down-sampling network; and finally, performing weighted summation on the intermediate prediction HR images output by each iteration module to reconstruct a high-resolution image. A novel residual block structure is adopted, so that the convergence speed is effectively improved; an iteration module is adopted, weight summation can be directly transmitted to the iteration module in the early stage during back propagation, and convergence is guaranteed; and meanwhile, a data set is strengthened by adopting a geometric self-integration method. The result shows that compared with an existing network model, the method not only has better picture authenticity, but also has a better PNSR index and a higher model convergence training speed.

Description

technical field [0001] The invention relates to an image super-resolution reconstruction method based on a deep neural network, belonging to the technical field of image processing. Background technique [0002] An important source of information for humans is vision, so the importance of images is self-evident. In simple terms, image resolution represents the amount of information, and the amount of information can be understood as the number of pixels in a one-inch image. With the vigorous development of the era of big data, people's demand for information is getting higher and higher, and the accuracy of information is getting higher and higher. LR (Low Resolution, low resolution) images cannot meet people's specific needs, such as in the field of medical images and monitoring, how to improve image resolution has become an important issue. Due to the long improvement cycle of the manufacturing process of hardware devices and the high cost compared with software, more re...

AI Technical Summary

Problems solved by technology

They believe that most super-resolution algorithms use L2 as the loss function, so the PNSR index becomes better, but the reconstructed image is too smooth and loses the sensory authenticity of the image

Images