Selection of color channels is the primary task for preprocessing color fundus images in automatic detection of various eye diseases. Preprocessing of fundus images helps in image enhancement and improving the quality of the image before computational or post-processing. In this paper, the mean square error and the peak signal-to-noise ratio are considered as a parametric measure for various color models. Color space conversion which helps in the automatic detection of various eye diseases is discussed in detail. The sample images from the Kaggle fundus database are tested using color space models. PSNR and MSE are calculated between the original image and color-converted images. The results are compared to get the best color space model for detection of cotton wool spot and conjunctiva abnormalities.