一种针对单幅室外图像的阴影检测算法

A Shadow Detection Algorithm for Single Outdoor Images

提出一种针对单幅室外场景图像的阴影检测算法。算法首先利用阴影的亮度和纹理特征,融合轮廓检测和区域分割两种方法分割图像;然后,通过训练SVM来检测单个区域信息和区域之间的匹配信息;最后,利用图割算法综合判断阴影区域。实验结果表明,该算法可以有效地分割单幅室外图像中的阴影区域。

Proposes a novel shadow detection approach for a single natural outdoor image. This algorithm firstly segments image by using the characteristic of shadow： brightness and texture. Then, the algorithm trains a SVM to detect single region information and region-pair information. At last, uses Graph Cut algorithm to classify shadow/non-shadow regions. Experiments show that this method significantly detects shadow regions in single outdoor images.