Research On Retinal Image Preprocessing And Blood Vessel Segmentation Algorithm

Fundus image blood vessels are the only microscopic human blood vessels that can be detected through non-destructive testing.Their morphological structure can provide a diagnostic basis for many related diseases,such as cataracts,diabetes,and arteriosclerosis.Vascular segmentation of fundus images is often an important part of the corresponding disease in the aided diagnosis system,and it is of great significance for the realization of related disease screening and automatic diagnosis.Due to the individual differences in fundus images,the effects of the imaging system,disturbances in the lesion area,and other factors,the blood vessel segmentation method still has many improvements.The existing fundus image segmentation method can be divided into two categories: tracking method and pixel point processing method.The advantage of the tracking method is to ensure the continuity of the blood vessel.The drawback is that the processing result in the next step is inseparable from the previous step.If the tracking deviation occurs,the algorithm may fail.In contrast,the pixel method is better than the tracking algorithm.However,the pixel method traverses all pixels,leading to a large amount of computation.To solve the above problems,in order to reduce the amount of calculation and ensure the accuracy of blood vessel segmentation,a blood vessel segmentation method for the fundus image is proposed by the firefly algorithm to optimize the Renyi entropy.According to the features of the fundus blood vessel,the G channel image is selected first,and the Gauss homomorphic filter is used as the frequency domain enhancement processing.Then the histogram equalization of low and high cap transformation is carried out respectively,and the result of the two is calculated as an arithmetic of subtract;Finally,the firefly algorithm is introduced to convert the problem of solving the optimum value of two-dimensional Renyi entropy into searching for the maximum brightness of fireflies.The image segmentation isperformed according to the threshold of entropy function corresponding to the location of the maximum brightness of fireflies.Experimental results show that the true positive rate and area under the ROC curve have increased,can accurately segregate the retinal blood vessels.