Design method of neural network for image definition discrimination

Abstract

The invention discloses a design method of a neural network for image definition discrimination, and the method comprises the steps: building a convolutional neural network which can discriminate thelocal definition of an image, enabling a to-be-detected image block to pass through a network with a specific structure, outputting an index of the definition degree of the to-be-detected image block,carrying out the classification, and forming a classified binary image. Because parameters of the convolutional neural network can be optimized based on a training sample and a training process, theproblems that feature indexes such as spatial frequency, SML, contrast and regional energy which are generally used for measuring the local definition of the image cannot well adapt to a complex imageenvironment, judgment errors are likely to happen in a gray level uniform region and the like are well solved, and therefore a better definition measurement mode can be obtained.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a design method of a neural network for image sharpness discrimination. Background technique [0002] In the evaluation of image quality, the sharpness of the image is an important index to measure the quality of the image, and how to evaluate the sharpness of the image objectively and effectively is still a hot research topic. The existing indicators that can be used to measure the local sharpness of an image usually include spatial frequency, SML, contrast, and regional energy. However, these a priori metrics are not well adapted to complex image environments, and are prone to make errors in the uniform gray area. [0003] In recent years, with the rapid development of big data applications and GPU-accelerated computing, the research on convolutional neural networks has gradually become extensive and in-depth, and many excellent convolutional neural networ...

AI Technical Summary

Problems solved by technology

[0004] At present, some scholars are also exploring new methods for image sharpness discrimination, but in the field of deep learning, there are relatively few complete networks for image sharpness discrimination that can achieve good results.

Images