High-resolution image prediction method based on loss function constructed considering image texture information

Abstract

The invention discloses a high-resolution image prediction method based on a loss function constructed considering image texture information. According to the method, first, a connection weight and offset of an SRCNN (convolutional neural network) are randomly initialized, and network parameters are set; after training data is preprocessed, a high-resolution image pair training set and a low-resolution image pair training set are obtained; next, low-resolution images are input into a network framework, and high-resolution images output by the network are obtained; then, the loss function considering image texture information is adopted to perform error calculation, if the number of iterations is not reached, weight correction is performed, and finally a trained network is obtained; and ata test stage, the low-resolution images are input into the trained network to obtain predicted high-resolution images. Through the constructed loss function, pixel loss can be measured, image textureinformation loss also can be measured, the defect of an SRCNN super-resolution algorithm is overcome, and further improvement to the performance of the SRCNN algorithm is effectively realized.

Description

technical field [0001] The invention relates to an image prediction method, in particular to an image prediction method based on deep learning, and belongs to the technical field of artificial intelligence. Background technique [0002] Deep Learning, as the research focus of computer vision and pattern recognition in recent years, has received more and more research attention from scholars, and image super-resolution technology based on deep learning theory has developed rapidly. At present, among many research results, there are endless researches on improving the performance of deep learning super-resolution algorithms from the perspective of reasonably constructing loss functions, and have produced many beneficial effects on the quality of image super-resolution. for example: [0003] [1] Ledig C, Wang Z, Shi W, et al.Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network[C] / / ComputerVisionandPattern Recognition.IEEE,2017. [0004] [2] Yu ...

AI Technical Summary

Problems solved by technology

[0009] However, the detailed features of the super-resolution image obtained by minimizing the above loss function are still not rich and obvious, and many important details of the image appear too smooth

Analyzing the reason, it can be found that the MSE loss function can only simply measure the difference in pixel values ​​between the high-resolution image output by the network and the real high-resolution image, but cannot measure the change of image information from the perspective of regional structure such as image texture. with the difference

Therefore, it is obviously not enough to measure the difference between super-resolution images and real high-resolution images only from the perspective of low-level features such as pixel values.

Images