基于GPU的单幅图像去雾的实现及优化

Implementation and optimization of single image haze removal based on GPU

基于暗通道先验规律的去雾算法已取得了良好的去雾效果,但算法所需要的计算时间过长,无法达到实时去雾的要求。使用GPU初步并行实现了去雾算法,并确定了算法中需要优化的部分。在优化过程中,一方面将数据存储到高速内存中以实现对数据的快速读取,另一方面设计新的算法实现方式以减少算法的计算量,最终提高了加速比。优化后的加速算法处理768×1024的图像仅需21 ms,达到了实时去雾的要求。

The defogging algorithm which based on dark channel prior had achieved good results,but the time spent on computing was too long to meet the requirements of real-time defogging. With parallel GPU,this paper implemented the defogging algorithm and it determined the portion of algorithm which need to optimized. During the optimization process,on the one hand,it stored the data in the high-speed memory to achieve rapid data read. On the other hand,it designed a new algorithm implementation to reduce the amount of calculation,it improved the acceleration rate ultimately. The acceleration algorithm just needed 21 ms when dealing with images of 768 × 1024 after optimized,so it reached real-time defogging implementation.