Fundus image quality evaluation method based on dark channel and bright channel

Abstract

The invention discloses a fundus image quality evaluation method based on a dark channel and a bright channel. The fundus image quality evaluation method comprises the steps of collecting and markinghistorical data of a fundus image; constructing a preliminary fundus image quality evaluation model; training the preliminary fundus image quality evaluation model by adopting the marked fundus imagehistorical data to obtain a fundus image quality evaluation model; and adopting the fundus image quality evaluation model to identify and evaluate the target fundus image to obtain a final fundus image quality real-time evaluation result. The method not only realizes evaluation of image quality, but also is high in reliability, good in practicability, high in accuracy and easy to implement.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a fundus image quality evaluation method based on a dark channel and a bright channel. Background technique [0002] With the development of economy and technology and the improvement of people's living standards, people pay more and more attention to health. [0003] Retinal images can better reflect the state of the fundus of the eye, and are an important basis for people to understand the state of their own fundus. At present, as people pay more and more attention to health, the demand and application of retinal images in medicine are becoming more and more extensive. After obtaining retinal images, medical technicians first need to evaluate the image quality of the obtained retinal images: only high-quality retinal images can be delivered to clinical doctors, and then help clinicians obtain more useful information. [0004] The current image quality assessment me...

AI Technical Summary

Problems solved by technology

On the one hand, this type of method requires the complexity of a single network, and introduces multiple frameworks in parallel to learn the prior knowledge of image quality assessment in order to obtain more comprehensive quality assessment basis information to make the learned model more robust; on the other hand, its training Due to the complex framework and the rapid increase of parameters that need to be optimized, the learning of auxiliary tasks requires additional annotation data and the entire framework cannot be trained end-to-end

[0006] Therefore, whether it is manual evaluation or intelligent algorithm evaluation, the current image quality evaluation methods for retinal images are not reliable and have poor practicability

Images