Abstract
Sun Glare widely exists in the images captured by unmanned ground and aerialvehicles performing in outdoor environments. The existence of such artifacts inimages will result in wrong feature extraction and failure of autonomoussystems. Humans will try to adapt their view once they observe a glare(especially when driving), and this behavior is an essential requirement forthe next generation of autonomous vehicles. The source of glare is not limitedto the sun, and glare can be seen in the images captured during the nighttimeand in indoor environments, which is due to the presence of different lightsources; reflective surfaces also influence the generation of such artifacts.The glare's visual characteristics are different on images captured by variouscameras and depend on several factors such as the camera's shutter speed andexposure level. Hence, it is challenging to introduce a general - robust andaccurate - algorithm for glare detection that can perform well in variouscaptured images. This research aims to introduce the first dataset for glaredetection, which includes images captured by different cameras. Besides, theeffect of multiple image representations and their combination in glaredetection is examined using the proposed deep network architecture. Thereleased dataset is available at https://github.com/maesfahani/glaredetection