Design a Hardware Applying Fog Removal Algorithm Using Median Dark Channel Prior for Autonomous Driving Car

Autonomous cars require camera images to obtain accurate information about the driving environment. However, there are times when clear images cannot be obtained due to fog, rain, or nighttime, which may adversely affect the quality of autonomous driving. In this paper, we propose an algorithm and hardware architecture to remove fog or rain from images in real time. A commonly used method to remove fog is the Dark Channel Prior method, which has the disadvantage of causing blocking phenomenon. Taking this into consideration, we modified it to the Median Dark Channel method. To design this algorithm in Verilog HDL, we could reduce hardware resource by sharing comparators and achieve real-time processing by pipelining and line buffer to increase memory usability.