Image enhancement based on adaptive demarcation between underexposure and overexposure

Images taken under non-uniform illumination usually suffer from degenerated details because of underexposure and overexposure. In order to improve the visual quality of color images, underexposure needs to be brightened and overexposure should be dimmed accordingly. Hence, an important procedure is discriminating between underexposure and overexposure in color images. Traditional methods utilize a certain discriminating threshold throughout an image. However, illumination variation occurs easily in real life. To cope with this, we propose an adaptive discriminating principle according to local and global luminance. Then, a nonlinear modification is applied to image luminance to light up underexposure and dim overexposure regions. Further, based on the modified luminance and original chromatic information, a natural color image is constructed via an exponential technique. Finally, a local and image-dependent exponential technique is applied to RGB channels to improve image contrast. Experimental results shows that the proposed method produces clear and vivid details for both non-uniform illumination images and images with normal illumination.