การปรับปรุงคุณภาพสีของภาพถ่ายเรตินาด้วยแบบจำลองคู่สีตรงกันข้าม

Abstract

WHO reported the number of 65 million of Age-related Macular Degeneration (AMD) patients around the world, and expected the number would increase to 300 million patients by the year 2040. Currently, ophthalmologists rely on retinal fundus photographs to analyze AMD lesions. Nevertheless, sometimes the photographs have an unsatisfactory quality such as low contrast, under or over exposure which results in difficulties for the experts to analyze lesions. So, it is suggested to have the unsatisfactory photographs improved to enhance anatomical details appearance before use by the experts. This thesis proposed an effective retinal fundus image simulation modeling to enhance contrast and adjust the color balance. It is aimed to assist ophthalmologists in AMD lesion screening. The proposed method consists of a few steps to achieve the intent. Firstly, an input image is improved contrast with CLAHE technique by using CIE L^* a^* b^* color space. Then, the histogram of the output image from previous step is stretched and rescaled by a scaling histogram technique to adjust its overall brightness offset to meet the Hubbard’s retinal fundus image proper range standard. This thesis used images as experimental data from two datasets, DiaretDB0 and STARE datasets. The results indicate the proposed method yields a highly contrast and color-balance output which fits the Hubbard’s standard and easier to screen lesions.