Joint Median Channel Prior and Edge Enhancement for Single Image Dehazing

Adopting reasonable approaches to process local patches for varying haze conditions is crucial for optimizing the performance in single-image dehazing. Motivated by the success of patch-size techniques in image dehazing, we proposed a novel median channel that respectively implements the median filter on the minimal and maximal color channels depending on the luminance differences in a local patch. In addition, we designed an algorithm to adaptively select the patch size when implementing the median filter determined by the size of the image. The proposed prior can effectively estimate the scene depth and avoid underestimation and local halos. Subsequently, the atmospheric light and transmission can be calculated to restore the degraded images. Furthermore, we developed an edge extraction method and implemented it to enhance local visibility and contrast. The proposed framework was tested by the real-world images and the experimental results presented that our method can perform well in both subjective visual and quantitative evaluations.