A Light Intensity Reverse Algorithm for Improving Dark Channel Prior Dehazing

Image haze-removal algorithm based on the dark channel prior is effective and simple. But it can also encounter problems such as color distortion, noise, detail loss on processing images with large bright patches filled with sky, water, snow and almost white object, because the dark channel prior assumption is invalid within these patches. In this paper, the light intensity reverse (Lgt-Int-Rev) algorithm is proposed which simply reduces the light intensity of high luminance regions of an image and successfully solved the above-mentioned problems with images in snow scene, and images including large sky regions. This algorithm needs not to change He's original algorithm, so all the advantages such as edge-preserving smoothing and gradient preserving of the original algorithm are completely reserved. A large number of experiments have been conducted and results demonstrate the effectiveness of the Lgt- Int- Rev algorithm.