A fast sand-dust video quality improvement method using simple color balance and dynamic guided filtering

Sand-dust weather seriously reduces the acquisition effect of computer vision equipment. To solve this problem, this paper proposes a fast sand-dust video quality improvement method using simple color balance and dynamic guided filtering. Our method extracts all frames of the video and then processes each frame in two steps by using the parallel computing method. The first step is to quickly correct the color deviation of the frame by a simple color balance method while eliminating the influence of nonuniform illumination. The second step is to use guided filtering with dynamic adjustment of the penalty coefficient to eliminate the interference of noise to the frame, enhance the contrast and detailed information of the frame, and finally reassemble the processed frames into the video with improved quality. Through qualitative and quantitative comprehensive experiments on sand-dust videos, the experimental results are compared with the existing methods, which prove that our method has advantages in improving the quality of sand-dust videos. The contribution of the proposed method can be summarized as follows: 1) A color balance method combined with screen Poisson equation is proposed. Due to light scattering by sand dust, the illumination of video frames are uneven. Our color balance method can effectively solve the problem that the difference between the target and background is small and difficult to identify when there is insufficient illumination. 2) A strategy of dynamically adjusting the penalty coefficient of guided filtering is proposed. The experimental results show that our method can effectively solve the problem of edge blur in some frames when guided filtering processes sand-dust videos. 3) A method of using multicore parallel processing of video frames is proposed to improve the quality of sand-dust video quickly.