A method for generating a convolution neural network model and an image enhancement method

Abstract

The invention discloses a method for generating a convolution neural network (CNN) model for image enhancement, which is executed in the computing device. The method comprises the following steps of:constructing a CNN model structure and setting a loss function expression; wherein the CNN model structure comprises a general feature module, a local feature module, a global feature module, a parameter table generation module, a boot diagram generation module, an enhancement parameter generation module and an image enhancement module; Acquiring a plurality of training image pairs, each trainingimage pair comprising an input image and a target image; training the CNN model by using a plurality of training image pairs as training samples, updating The model parameters of CNN model continuously, and calculating the loss function value of CNN model according to the loss function expression; When the loss function value converges, stopping the training process, and obtaining the model parameters of CNN model so as to generate the CNN model. The invention also discloses a corresponding image enhancement method and a computing device.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for generating a convolutional neural network model for enhancing image contrast and sharpness, an image enhancement method and a computing device. Background technique [0002] The development of mobile terminals and Internet technology facilitates users to obtain image information. For example, users can take pictures with mobile phones, or browse and download images from the Internet. However, most of the images obtained by users are of unsatisfactory quality. For example, the contrast of the image is not enough, and the overall image is white and dark; for example, the sharpness of the image is not enough, the quality of the image is blurred, and the edges are smooth. Therefore, it is necessary to improve the contrast and sharpness of the image to enhance the image information. [0003] The traditional method mostly uses curve adjustment to enhance the cont...

AI Technical Summary

Problems solved by technology

[0003] The traditional method mostly uses curve adjustment to enhance the contrast of the image. However, this method requires manual adjustment of parameters, which is inconvenient to operate and reduces the user experience. It is also highly targeted and cannot be flexibly applied to multiple image scenarios.

Existing algorithms mostly use filtering or spatial differentiation to enhance the sharpness of images. However, this method uses a uniform method to sharpen all images, which is not flexible enough, and it is easy to cause excessive sharpening and unnatural effects.

In addition, in the existing algorithms, image contrast enhancement and sharpness enhancement are two relatively independent issues, it is difficult to enhance the contrast while also enhancing the sharpness, and the implementation efficiency of the enhancement algorithm is low

Images