Research On Edge Feature Enhancement And Coloring Algorithm Of Fundus Image

Fundus images are important carriers that directly reflect eye related diseases,and are the most fundamental and effective basis for doctors to diagnose eye diseases.With the vigorous development of computer technology,fundus imaging devices are no longer limited to traditional imaging methods,and confocal laser technology is widely used in ophthalmic imaging devices.However,confocal laser fundus imaging devices need to be combined with digital image processing related technologies to better and more accurately display the social value of disease prevention and diagnosis in the medical field.Fundus image enhancement is a prerequisite for fundus image analysis.Image coloring makes color imaging in confocal laser technology possible,making fundus images closer to the true color of the human eye.Based on this,this article has conducted the following research on fundus image processing:1.For the first time,this paper takes the confocal fundus image as the research object,studies and analyzes the characteristics and advantages of the confocal fundus image,establishes the enhancement and coloring network for the confocal fundus image without the available data set of the confocal fundus image,overcomes the defects and deficiencies of the confocal fundus image in medical diagnosis,and optimizes and improves the processing methods of the traditional fundus image to some extent,It is transformed into a new method more suitable for confocal fundus images.2.Aiming at the problems of poor edge sensitivity,information occlusion and indistinct blood vessel differentiation in the confocal fundus image,a blood vessel edge feature enhancement algorithm based on optimized Laplace filter is proposed.On the basis of Laplace operator edge enhancement,combined with gamma change,the Laplace edge feature enhancement model is constructed by introducing similarity measurement theory.When processing the confocal fundus image,the detailed division of image information is considered,the background information and edge information are differentiated,and the adaptive enhancement processing of different regions is carried out,which realizes the enhancement of the edge feature of the confocal fundus image.The effectiveness of the proposed method has been compared and analyzed through multiple experiments.3.A fundus image coloring algorithm based on the degradation model is proposed.The colorization of the confocal fundus image is realized by using the grayscale image coloring theory.On the basis of the color migration method based on the reference image,the color histogram of the reference image is used as a priori,and the semantic matching relationship between the reference image and the confocal fundus image is used,based on the color migration theory,the confocal fundus image is colored by using the convolution neural network,Aiming at the problems of image quality degradation and resolution reduction,GAN network was introduced to simulate the degradation model of the confocal fundus image in the process of coloring,and the resolution of the confocal fundus image after coloring was improved by integrating the scale factor,fuzzy kernel,noise and other factors of image degradation,The overall network architecture proposed in this paper can well complete the coloring task of confocal fundus images.(4)A user oriented confocal fundus image simulation system has been designed and implemented,integrating the algorithms proposed in this article.Users can implement enhancement and coloring algorithms through simple operations to obtain real-time output confocal fundus images.