Fundus image enhancement method based on generative adversarial network

Abstract

The invention discloses a fundus image enhancement method based on a generative adversarial network. The fundus image enhancement method comprises the steps of selecting training data and test data, data preprocessing including cutting, zooming, rotating and normalizing processing of the image, constructing a convolution layer, a residual module group and a deconvolution layer as an image generator, inputting the preprocessed color fundus image, and outputting a corresponding high-quality fundus image, constructing a full convolutional neural network as a discriminator, inputting the generatedfundus image and a real fundus image, and outputting the probability that the generated image is judged as a real image, the task of the generator being to generate a real image as much as possible,the task of the discriminator being to discriminate authenticity from the generated image as much as possible, and the two tasks being alternately trained until a satisfactory generation result is achieved. The fundus image generated by using the generative adversarial network is clear in structure and fidelity in color, and a good fundus image quality enhancement effect can be achieved.

Description

technical field [0001] The invention relates to a fundus image enhancement method based on a generative confrontation network, which belongs to the field of medical image processing. Background technique [0002] Fundus image detection is one of the important means of eye examination. Obtaining fundus images with a fundus camera is the most effective and basic way to screen for common eye diseases. The quality of fundus images is the key to intelligent screening of eye diseases, and high-quality fundus images are the prerequisite for correct classification by intelligent systems. With the development of portable fundus cameras, we can easily acquire fundus images, but quite a few of these fundus images are of poor quality. Poor quality images can be caused by insufficient lighting, overexposure, occlusions, patient eye movement, etc. Existing studies have shown that there are a large number of poor-quality images in the actual data set, accounting for up to 60% of the tot...

AI Technical Summary

Problems solved by technology

The above methods can improve the local contrast of the image and enhance the image details in the darker areas with insufficient light, but there are serious color distortions

Images