Quality Grading of Dehazed Images Through Visibility Index

Abstract

The foggy or hazy weather condition degrades the resolution and contrast of the acquired outdoor scene object. This leads to the poor visibility and necessitates the use of an image dehazing algorithm for the visibility enhancement of the captured images. This paper presents the proposed method to quantify the quality of the dehazed images in order to measure the effectiveness of the applied dehazing algorithm. In this paper, an algorithm based on Dark Channel Prior (DCP) and guided filter is implemented for image dehazing. The DCP is based on the characteristic of dark pixels in that the pixels hold low intensity value in atleast one color channel of the haze-free outdoor images. The transmission map so obtained from dark channel prior is refined by applying guided filter. The dominant advantage of using guided filter is its capability to keep sharp details while upholding the color quality. The proposed algorithm is tested over the number of hazy natural images. The image quality transition from an input hazy image to an output dehazed image is evaluated in terms of Visibility Index (VI). The dehazed image falls into either three quality grades; medium, slight and non-hazy. The experimental results so achieved are remarkable to show the quite difference between the quality of input and output images of the applied dehazing algorithm.