Contrast and color balance enhancement for non-uniform illumination retinal images

Abstract

Color retinal images play an important role in supporting a medical diagnosis. However, some retinal images are unsuitable for diagnosis due to the non-uniform illumination. In order to solve this problem, we propose a method for improving non-uniform illumination that can enhance the image quality of a color fundus photograph suitable for reliable visual diagnosis. Firstly, a hidden anatomical structure in dark regions of the retinal images is revealed by improving the image luminosity with gamma correction. Secondly, multi-scale tone manipulation is then used to adjust the image contrast in the lightness channel of L*a*b* color space. Finally, color balance is adjusted by specifying the image brightness based on Hubbard’s specification. The performance of the applied method has been evaluated against the data from the DIARETDB1 dataset. The results obtained show that the proposed algorithm performs well for correcting the non-uniform illumination of color retinal images.

Keywords

• human visual perception
• image contrast and color balance enhancement
• non-uniform illumination
• retinal image