OpenCL并行加速优化暗通道去雾算法

OpenCL parallel acceleration optimization dark channel defogging algorithm

暗通道先验去雾算法是目前细节还原效果最好的去雾方法之一,但由于其运算量大,难以对视频图像进行实时去雾。针对该问题,文中提出一种OpenCL并行加速优化方案,用以减小计算量,增强计算过程的可并行性。首先,对原始有雾图像进行下采样处理后计算小图暗通道,再计算透射率并用插值法恢复原图透射率,减小获取暗通道过程的计算量;其次,采用分块算法将大气光值估算过程划分为适合OpenCL并行处理的子块,增强大气光值估算过程的并行性。实验结果表明,该方案可以保证良好的图像去雾效果,并将720P图像的去雾耗时减小到20 ms,可实现对帧率为50 f/s视频的实时去雾。

The dark channel prior defogging algorithm is one of the best defogging methods for the detail restoration at present,but it is difficult to defog video images in real time due to the large amount of computation.To solve this problem,an OpenCL parallel acceleration optimization scheme is proposed to reduce the amount of computation and enhance the parallelism of the calculation process.After down sampling the original foggy image,the small image dark channel is calculated.The transmittance is calculated and the original transmittance is restored by means of the interpolation method to reduce the amount of calculation in the process of obtaining dark channels.The atmospheric light value estimation process is divided into sub blocks suitable for OpenCL parallel processing by means of the block algorithm to enhance the parallelism of atmospheric light value estimation process.The experimental results show that the scheme can ensure good image defogging effect,reduce the defogging time of 720P image to 20 ms,and realize real⁃time defogging of video with frame rate of 50 f/s.