A Simplified and Fast Dehazing Processing Suitable for Embedded Systems

The main purpose of image dehazing is to remove undesirable degradation caused by fog and mist to improve the visual quality of an image and provide benefits to postprocessing. In particular, image dehazing systems applied to actual applications such as video surveillance and in-vehicle cameras are required to realize a balanced performance for video streams between dehazing quality and processing speed in low-cost portable embedded devices. To cope with this requirement, in this study, we propose a simple and fast dehazing algorithm which provides high-quality dehazed images and a practical processing speed. Concretely, the proposed method utilizes a simplified transmittance estimation technique with non-linear processing focusing on an optimization of its hardware implementation on a low-cost embedded target board with a graphics processing unit (GPU). Through a series of comparative experiments with typical simplified dehazing methods, we verify the effectiveness of the proposed method on dehazing quality and processing speed.