Sky Detection in Hazy Image

Abstract

Sky detection plays an essential role in various computer vision applications. Most existing sky detection approaches, being trained on ideal dataset, may lose efficacy when facing unfavorable conditions like the effects of weather and lighting conditions. In this paper, a novel algorithm for sky detection in hazy images is proposed from the perspective of probing the density of haze. We address the problem by an image segmentation and a region-level classification. To characterize the sky of hazy scenes, we unprecedentedly introduce several haze-relevant features that reflect the perceptual hazy density and the scene depth. Based on these features, the sky is separated by two imbalance SVM classifiers and a similarity measurement. Moreover, a sky dataset (named HazySky) with 500 annotated hazy images is built for model training and performance evaluation. To evaluate the performance of our method, we conducted extensive experiments both on our HazySky dataset and the SkyFinder dataset. The results demonstrate that our method performs better on the detection accuracy than previous methods, not only under hazy scenes, but also under other weather conditions.

Single Image Dehazing Using Adaptive Sky Segmentation

“…In the previous researches [17–19], important characteristics fall into three categories: color feature, gradient feature and position feature. For accuracy and efficiency, we choose two color features that are dark channel and color saturation, and two gradient features that are contrast energy and color gradient.…”

Sky and Ground Segmentation in the Navigation Visions of the Planetary Rovers

“…It is noteworthy that this research names the non-sky pixels as the ground pixels for better discussion. The Skyfinder dataset is a widely used dataset in sky and ground segmentation [5,26,31,34,35].…”

“…Figure A2 displays the examples of eliminated images. The same action for eliminating the "bad" images was also conducted in the related studies [5,26,31,34,35], which is a routine for using the Skyfinder dataset. Besides, the upset images (Figure A2) do not stay in the scope of this research.…”

“…De Mattos et al [25] used a whiteness-based SVM. Song et al [26] used two imbalance SVM classifiers and similarity measurement. Ye et al [6] and Dev et al [27] focused on The whole-sky imager and proposed a superpixel-based SVM method.…”

“…Ahmad et [41] proposed a fusion of edge-less and edge-based approach to detect the horizon. Shan et al [42] proposed a superpixel-based approach, which works for robotic navigatio McGee2005, Liu2017, Mattos2018, Song2018, Ye2019, Dev2017, Beuren2020, Tighe2013, Mihail2016, Tsai2016, Liu2016, Vargas2019, Dev2019, Fu2019, Place2019, Nice2020, and Ho ży ń2020 refer to [2,10,12,[23][24][25][26][27][28][31][32][33][34][35][36]39], respectively. Some related studies focus on the sky and ground segmentation in the planetary rover scenario.…”