Performance improvement of dehazing for fogged image using parameter optimization

Dehazed images can be turned into clear images by using haze-removal algorithms for image processing. This paper uses dark channel prior for the dehazing process. The dark channel prior algorithm smoothens the transmission through matting, which is time intensive. This paper adopts the Gaussian filter for smoothing the transmission, resulting in faster calculation time. However, the Gaussian filter causes a halo effect around the boundary of the object. To improve the performance of the dehazing algorithm, the parameters that affect the quality of the dehazed image are selected and optimized within an assigned range. As a performance index for the quality of the dehazed image, a blind contrast enhancement assessment method is applied to measure the edge level, contrast and saturation. Finally, the optimal parameters are selected by comparing several values from the performance index. © ICROS 2018.