Retinal segmentation is crucial for diagnosing ophthalmic diseases and quantifying retinal characteristics. The Canny operator, while widely regarded as an effective edge detector, is limited by its sensitivity to vascular shadows, vitreous artifacts, and noise, resulting in detection inaccuracies. A modified Canny detector is proposed for detecting retinal diseases, particularly Stargardt’s macular dystrophy. The modified algorithm improves the traditional Canny operator by refining preprocessing steps, incorporating Local Contrast Modification (LCM) and Contrast Limited Adaptive Histogram Equalization (CLAHE) for contrast enhancement, and a variational model for specular denoising. The proposed approach demonstrates significant improvements compared to state-of-the-art methods.