Method and device for discriminating image sharpness

Abstract

The invention relates to a method and a device for discriminating image sharpness and belongs to the image processing field. The method comprises steps that an original image of which sharpness is to be discriminated is acquired, and the original image is zoomed to a designated pixel to acquire a target image; a first characteristic of the target image is acquired, and the first characteristic is identified through a gray matrix, an image contrast matrix and a gray gradient maximum accumulation sum matrix; the first characteristic is pre-processed to acquire a second characteristic; the second characteristic is inputted to an artificial neural network model acquired through pre-training, and whether the original image is intelligible is discriminated according to an output result of the artificial neural network model. According to the method, through the gray matrix, the image contrast matrix and the gray gradient maximum accumulation sum matrix, the first characteristic of the image is extracted, image characteristic variety is guaranteed, the image can be reflected from multiple aspects, a discrimination result based on the first characteristic is guaranteed to be relatively accurate, the second characteristic is inputted to the artificial neural network model acquired through pre-training to discriminate whether the original image is intelligible, and a discrimination mode is relatively simple.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method and device for judging image clarity. Background technique [0002] The discrimination of image sharpness is an important research content in the field of image processing, and it has a wide range of applications in the fields of image scaling, display and reconstruction. Therefore, how to judge the sharpness of an image has received extensive attention. [0003] When judging the sharpness of an image in the related art, an SVM (Support Vector Machine, support vector machine) model is pre-trained, and the SVM model determines the sharpness of the image according to the low-frequency noise of the image. On this basis, when judging the sharpness of the image, first extract the low-frequency noise features of the image, and input the low-frequency noise features into the SVM model, so as to determine whether the image is clear or not according to the output result...

AI Technical Summary

Problems solved by technology

[0005] Since only the low-frequency noise features of the image are extracted when judging the sharpness of the image, the extracted features are relatively simple. However, this single feature may not reflect the image features well, so the judgment result may be inaccurate.

Images