A Real-time Framework for HD Video Defogging using modified Dark Channel Prior
Xinchun Wu, Xiangyu Chen, Xiao Wang, Xiaojun Zhang, Shuxuan Yuan, Biao Sun, Xiaobing Huang · Research Square · 2023
Abstract Foggy weather reduces the quality of video capture and seriously affects the normal work of video surveillance, remote sensing monitoring, and intelligent driving. Many methods have been proposed to remove video haze. However, under the premise of ensuring real-time performance, their defogging effect needs to be further improved. This paper improves the dark channel prior (DCP) dehazing algorithm, and designs a defogging framework that takes into account good dehazing effect and real-time processing. First, an adaptive threshold segmentation algorithm is proposed, which can well solve the serious color cast problem in brighter areas in DCP. Second, an algorithm for preserving image details using gradients is proposed, which achieves agood balance between detail preservation and computational efficiency. Then, each frame of video is evenly divided into a plurality of sub-areas, and the sub-ares are sequentially processed in a pipeline manner, which improves calculation efficiency. Finally, a high-definition real-time video defogging framework with a resolution of 1920×1080 and 60 frames/s is realized on the ZYNQ7035.