中文摘要：

针对目前主流去雾算法结果尚未达到人眼视觉愉悦度且消耗计算资源大等不足,以人眼视觉特性规律为指导,运用亚当斯区域曝光理论与非下采样Contourlet多尺度分析工具,在HSV彩色空间中提出一种雾霾背景下的彩色图像质量提升算法。该算法采用最大视觉显著度作为中度灰基准,模拟人眼控制瞳孔曝光程度,适度拉伸亮度分量的动态范围;鉴于雾霾遮掩图像大量细节特征,仿照人眼多频信道分解特性,将处理后的亮度分量纳入改进的非下采样Contourlet滤波器组以凸显图像边缘信息;依据颜色恒常性原理,色调分量保持不变,通过统计彩色图像各分量间的先验信息,融合亮度与色调分量,构造线性变换以校正饱和度分量,使各颜色分量间维持良好的相关性。实验结果表明:本文算法能有效提升雾霾图像的视觉效果,具有较好的实用性和较少的计算资源消耗。

英文摘要：

Since the results of mainstream defogging algorithms are far to reach human visual enjoyment and exhaust plenty of computing resource, a lifting color image quality algorithm based on the human visual mechanism was put forward in HSV color space through the Adams zone system and the non-subsampled contourlet transform under hazy weather. According to the algorithm, by imitating the process of papilla exposure, the maximum visual saliency was considered as medium grey standard to pull up dynamic ranges of the V component properly. Large amounts of detail features covered under fog and haze were given, the V component was then brought into non-subsampled contourlet filters to outstand image edge information according to the multi-frequency channel decomposition of human eyes. The H component remained unchanged because of the color constancy theory, but the S component merging together with the other adjusted-well components was revised by linear transform which relied on the prior statistics of color image components to keep them in good correlation. The simulation reveals that the novel algorithm not only improves visual effects of hazy images, but also provides higher practicability and less computing consumption.