Lightness-aware contrast enhancement for images with different illumination conditions

It has become more convenient to take photographs in our daily life. However, without sufficient skills, we often produce poor photographs with low contrast and unclear details under various imperfect illumination conditions. Although plenty of image enhancing models have been developed, most of them impose a uniform enhancing strength to the whole image region, and thus tend to generate over-enhancement effects for regions with originally-satisfying illumination. To address this issue, we propose a novel contrast enhancing model, which is a simple linear fusion process based on an original image and its initial enhancement. As the key of our model, we construct a lightness map that estimates the scene lightness, which is aware of the image structure at pixel-wise level. In the fusion process, this map dynamically weighs between the initially enhanced image and the original image, and thus ensures a seamless fusion result. In our experiments, we validate our model on images with various illumination conditions, such as strong back light, imbalanced light, and low light. The results empirically show that our model performs well on simultaneously improving image contrast and keeping its naturalness.