Abstract:

Images taken in rainy days can be easily disturbed by rain streaks, which affects the feature extraction of the images and then leads to the degradation of the processing performance of the visual system. This paper proposes a single image rain removal algorithm based on similar patch matching and minimum value filtering. The method consists of two stages：detection and repairing. In the detection stage, rain pixels are marked by leveraging the brightness prior and direction information of rain. In the repairing stage, after similar patches are matched in both coarse-grained and fine-grained level, the pixels marked as rain pixels are replaced by the minimum value of the corresponding pixel in the similar patch found by the minimum value filter, while the ones marked as non-rain pixels remain unchanged. The algorithm is tested on both synthetic and real images. The experimental results show that the algorithm can accurately locate the rain pixel and match the similar patches, and thus eventually repairs the rain pixels with high quality. Compared with other methods, it has certain advantages in objective measurement and subjective vision.

Key words: image de-raining, rain detection, similar patch matching, minimum value filter

摘要：

雨天环境拍摄的图像容易受到雨线的干扰，影响图像的特征提取，继而导致视觉系统的处理性能下降。提出了一种基于相似块匹配与最小值滤波的单幅图像去雨算法，该算法包含检测和修补两个阶段。检测阶段利用雨的亮度先验和方向信息对图像中的雨像素点进行标记。在修补阶段，先采用粗粒度和细粒度层级结合的方式匹配相似块，然后将雨像素点通过最小值滤波用相似块中对应像素最小值替换，而非雨像素点则保持原有值不变。算法在合成雨图与真实雨图上进行了验证，实验结果表明该算法能实现雨像素精确定位，精准匹配相似块，最终高质量修补雨像素，相比其他方法在客观度量值和主观视觉上均有一定的优势。

关键词: 图像去雨, 雨线检测, 相似块匹配, 最小值滤波