Atmospheric Light Correction and Transmission Optimization Based Robust Image Dehazing

To address the dusky sky and the insufficient contrast enhancement of the existed single image dehazing methods, we propose a high visibility image dehazing algorithm with adaptive atmospheric light correction and robust transmission optimization strategies. The input image is preprocessed with white balance and gamma correction in this paper to increase brightness, enhance contrast and avoid the color cast problem. In order to prevent the atmospheric light from over-estimation, an adaptive atmospheric light correction strategy based on sky detection is put forward, which is beneficial to brighter restoration of the sky regions. Finally, we recognize those pixels with unreliable transmission by the detection of halo effect and the inference of transmission context consistency. And, we design a robust transmission optimization model to correct the unreliable transmissions, under the constraint of the reliable transmission maintenance term, the unreliable transmission interpolation term and the transmission correlation term among the similar pixels. Experimental results show that our transmission consists with the depth variation better, and the haze-free images possess high definition, contrast and color saturation, as well as natural restoration of sky regions.