摘要
为了降低雾天对成像的影响,获得实时的去雾效果,对彩色图像处理流程进行改进,提出一种新颖快速的基于Bayer图像和暗原色先验模型的单幅图像去雾方法。针对Bayer图像像素排列特点,对原有暗原色先验去雾算法进行了优化和改进。运用四叉树细分算法估算大气光,根据Bayer图像特点修正了引导滤波器(Guided Filter),并利用修正的滤波器优化大气透射图,恢复出无雾Bayer图像,采用去马赛克及系列后处理算法获得清晰的显示图像。实验结果表明,新方法在一定程度上改善了原算法去雾效果,并显著提高了运算速度。
In order to reduce the bad effect in the outdoor visibility system in foggy days and achieve the real-time dehazing effect, a fast and effective method, which is based on Bayer pattern image and dark channel prior, was developed to remove the haze from single image. The color image processing pipeline was improved through this method. Bayer pattern image is the raw data captured by digital camera and costs low memory. According to the characteristics of Bayer pattern image, the original dark channel prior algorithm was optimized and improved. The atmospheric light was estimated based on the quad-tree subdivision. In the optimization process of the transmission map, the coarser estimation was refined using the Guided Filter, which was modified with the property of Bayer pattern image. Finally, the RGB image, which is displayed by the device, was recovered from the haze removal Bayer image using the demosaic algorithm. The experimental results show our new single image haze removal method improves the image quality by reducing the haze effect, and enhances the calculation speed.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2013年第6期109-115,共7页
Journal of National University of Defense Technology
基金
国家自然科学基金资助项目(61175006
61175015
61275016
61271438)
