earticle

영어

Poor weather condition such as fog, rain or haze causes problem in getting information from image. As computer vision system extracts features from the scene, it is important that reliable information could be obtained by removing haze. This paper proposes a fast fog removal method which is implementable in hardware for single image. Using pixel-based median channel of haze image, we can estimate atmospheric light. As a result, high-quality image can be recovered with lower computation complexity compared to patch-based dark channel prior. FPGA implementation shows that real-time dehazing is achievable with median channel prior.