A FPGA-based System Implementation of Single Image Haze Removal using Improved Dark Channel Prior
Nguyen Hoang Hai Pham, Chi‐Chia Sun · 2023
Removing haze from images remains an ill-posed challenge in image processing. The dark channel prior-based dehazing algorithms demonstrates favorable outcomes for a wide range of scenes, despite the inherent impact of haze on image quality. However, the DCP algorithm may struggle to accurately remove haze and produce artifacts in scenes with extremely bright regions, leading to potential shortcomings in haze removal effectiveness. In this work, we combine a bright channel prior with an adaptive blending to enhance the transmission map refinement. The proposed method is evaluated using a widely accepted single image dehazing benchmark, then deployed on Field-Programmable Gate Array (FPGA) platform to produce haze-free output with better details in real-time.