摘要
针对暗原色先验去雾算法对大面积天空区域处理不理想等问题,提出了一种基于天空区域分割的图像去雾算法.该算法首先使用最大类间方差法(Otsu)求得阈值将天空与非天空区域分割开,然后再对天空区域求大气光值A,作为整幅雾天图像的大气光值,进而精细化透射率,得到较好的复原效果.为了提高去雾效率,基于CUDA(Common Unified Device Architecture)并行编程算法,通过GPU并行处理和CPU任务调度的协同,在保证去雾效果的同时快速去雾.实验结果表明,该方法能很好的解决天空区域去雾出现的光晕和色彩失真现象.优化后的并行算法,针对高清图像基本上达到了实时去雾的效果.
Aiming at the problem that the dark-color prior defogging algorithm is not ideal for large-area sky region processing,an image dehazing algorithm based on sky region segmentation is proposed. The algorithm first used the Otsu method to obtain a threshold to separate the sky from the non-sky region;then,obtained the atmospheric light value A of the sky region as the atmospheric light value of the entire fog image,further refined the transmittance and got better recovery effect. In order to improve the efficiency of defogging,the algorithm based on the CUDA( Common Unified Device Architecture) parallel programming combined the cooperation of GPU parallel processing with CPU task scheduling,the dehazing effect can be quickly implemented while guarantee the defogging effect. The experimental results showthat this method can well solve the phenomenon of halo and color distortion in the sky area. The optimized parallel algorithm basically achieves the performance of real-time defogging in high definition images.
作者
曾浩洋
张红英
ZENG Hao-yang;ZHANG Hong-ying(Southwest University of Science and Technology,School of Information Engineering,Mianyang 621010,China;Robot Technology Used for Special Environment Key Laboratory of Sichuan Province,Mianyang 621010,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2019年第3期655-659,共5页
Journal of Chinese Computer Systems
基金
四川省科技厅科技支撑计划项目(2015GZ0212
2014SZ0223)资助
特殊环境机器人技术四川省重点实验室开放基金项目(13zxtk05)资助
