Retinal Fundus Image Enhancement With Detail Highlighting and Brightness Equalizing Based on Image Decomposition

High-quality retinal fundus images are widely used by ophthalmologists for the detection and diagnosis of eye diseases, diabetes, and hypertension. However, in retinal fundus imaging, the reduction in image quality, characterized by poor local contrast and non-uniform brightness, is inevitable. Image enhancement becomes an essential and practical strategy to address these issues. In this paper, we propose a retinal fundus image enhancement method that emphasizes details and equalizes brightness, based on image decomposition. First, the original image is decomposed into three layers using an edge-preserving filter: a base layer, a detail layer, and a noise layer. Second, an adaptive local power-law approach is applied to the base layer for brightness equalization, while detail enhancement is achieved for the detail layer through saliency analysis and blue channel removal. Finally, the base and detail layers are combined, excluding the noise layer, to synthesize the final image. The proposed method is evaluated and compared with both classical and recent approaches using two widely adopted datasets. According to the experimental results, both subjective and objective assessments demonstrate that the proposed method effectively enhances retinal fundus images by highlighting details, equalizing brightness, and suppressing noise and artifacts, all without causing color distortion.