Quality Enhancement of Foggy Images Comprising of Large Sky Region on SAMEER TU Dataset

Images acquired by a visual system are critically degraded under the hazy and foggy weather, which will exert bad influence on several computer vision applications like remote sensing, intelligent vehicles, and visual surveillance detection, tracking, and recognition of objects. Thus, reconstructing the true scene from such a foggy image is very important. In this paper, we have adopted a segmentation method on fog degraded images containing large sky region as restoring the foggy images comprising of sky region is a very difficult task. Here, we presented different issues related to poor visibility in foggy weather condition. Then, we investigated existing image defogging algorithms based on atmospheric scattering model, fusion based strategy, dark channel prior and filtering approach with relative advantages and shortcomings. Most of the existing techniques fail to enhance image from fog degraded input image comprising the sky areas. This challenge is solved in our approach by adopting an FCM clustering based segmentation method. The experimental results demonstrate that by using this segmentation method, the foggy image with the sky region shows better results than other existing defogging algorithms that make it suitable to use in many computer vision based real-time applications.