硬件优化的高清视频实时去雾算法

Hardware optimized real-time dehazing algorithm for high-definition video

去雾技术已经在单幅图像上取得了较大的进展,但是由于时间复杂度较高,无法满足高清视频去雾的实时性要求。针对该问题,提出了基于硬件架构优化的暗通道先验去雾算法,通过硬件导向的双尺度联合滤波法和阈值比较法来简化透射率和大气光值的计算复杂度,同时利用帧间依赖性约束来抑制视频去雾中的闪烁噪声。实验结果表明,所提算法的去雾速度达到了148.2 Mpixel/s,相比软件的方式提高了约55倍,实际对1920×1080分辨率全高清视频的去雾速度达到69 fps,满足实时性要求且去雾质量高。

Dehazing technology has made great progress on a single image,but due to the high time complexity,it cannot meet the real-time requirements of HD video dehazing.Aiming at this problem,this paper proposed a dark channel prior dehazing algorithm based on hardware architecture optimization.It used the hardware-oriented dual-scale joint filtering method and threshold comparison method to simplify the computational complexity of transmittance and atmospheric light value.At the same time,it used the interframe dependency constraint to suppress the flicker noise in the video defogging.The experimental results show that the dehazing speed of the proposed algorithm reaches 148.2 Mpixel/s,which is about 55 times higher than that implemented by software programming,and the dehazing speed of full HD video at 1920×1080 resolution reaches 69 fps,meeting the real-time requirements and high dehazing quality.