An improved contrast enhancement for dark images with non-uniform illumination based on edge preservation

Histogram Equalization is a typical method that enhances image contrast by means of a global transformation. However, it overemphasizes images without preserving edges, loses details near edges or amplifies noise. These are obvious in case of enhancing dark image with non-uniform illumination, which contains both bright and dark regions. The main cause is that Histogram Equalization performs a global transformation considering only pixel brightness level. To tackle this problem, we present a new method that enhances image contrast paying attention to edge information. We obtain a transformation function adjusting dynamic range adaptively according to edge intensity values, which are calculated for each brightness level. To avoid noise amplification, our method decomposes input image into a base and detailed layer and transforms them individually. The experimental results show that our method parades better performance than the state-of-the-art methods for dark images with non-uniform illumination.