Characterization of optic disc in human retinal images using Tsallis entropy based method

The detection of the optic disc is an important step in the automated analysis of retinal images. A method to identify the optic disc in human retinal images using, non-extensive entropy, also called Tsallis entropy, is presented in this paper. The segmentation process is carried out on the retinal images using conventional Tsallis entropy, with and without morphological preprocessing. An optimum threshold value is found to separate optic disc from the blood vessels and other structures. The performance of the conventional Tsallis algorithm with and without preprocessing is compared and analyzed based on the different entropic index, visual quality, pixel distribution and the geometrical size of the optic disc. The conventional Tsallis entropy with morphological preprocessing is able to segment the optic disc. The detection efficiency is found to be high for the entropic index equal to 4 and also it is found that optic disc index is a useful parameter to differentiate normal from abnormal retinal images. The results are useful for detection of various retinal related diseases.