Estimating Depth and Global Atmospheric Light for Image Dehazing Using Type-2 Fuzzy Approach

This article proposes a novel single image dehazing method using a Type-2 membership function based similarity function matrix. The proposed method estimates the depth map and global atmospheric light of the observed hazy image. The estimated depth map is further subjected to produce true scene transmission. Finally, the observed hazy image is dehazed by the atmospheric scattering model using scene transmission and global atmospheric light. The qualitative and quantitative comparisons of the proposed method have been presented with benchmarked state-of-the-art methods. The experiments have been extensively performed on benchmarked natural hazy images, MiddleBury Stereo dataset, REalistic Single Image DEhazing (RESIDE) dataset, RESIDE-$\beta$ dataset, and Stanford ImageNet dataset. The performance metrics used for comparison are peak signal to noise ratio and structural similarity index as quantitative measures; and lightness order error and naturalness image quality evaluator as qualitative measures. Moreover, the detection results using YOLOv2 on RESIDE-$\beta$ dataset have also been compared in terms of F1-score and area under curve measures. The qualitative and quantitative comparisons show that the proposed method outperforms others and dehazed images are restored effectively maintaining their naturalness.