摘要
为了准确检测户外图像中的影子,本文提出一种基于区域分割结合梯度方向特征的影子检测方法。首先利用均值漂移分割原始图像,其次利用影子区域分类器并结合亮度信息初步检测出图像中的影子区域,然后利用相同区域对分类器检测出材质、亮度均相同的相邻区域对,最后将与初步检测到的影子区域相邻且具有相同材质、亮度的区域均标记为影子区域,最终得到图像中所有的影子区域。实验结果表明,在现有颜色、纹理、亮度特征的基础上,结合分割区域的梯度方向特征进行影子检测,可有效提高影子检测的准确性。
In order to detect shadow in an outdoor image accurately,a shadow detection approach based on region segmentation combining with gradient direction feature is proposed in this paper.Firstly,the image is segmented by using mean shift method.Secondly,the shadow regions are preliminarily detected according to shadow region classifier and region illumination.Thirdly,the adjacent region pairs with same material and illumination are detected by using the same region pair classifier.Lastly,the adjacent regions that have the same material and illumination with shadow region are marked as shadows.The experiment results show that the accuracy of shadow detection can be effectively improved based on the existing color,texture,illumination features and combining with the gradient direction features of segmented regions.
出处
《燕山大学学报》
CAS
2013年第2期137-142,共6页
Journal of Yanshan University
基金
河北省自然科学基金资助项目(F2010001276)
关键词
户外图像
影子检测
区域分割
分类器
梯度方向
outdoor images
shadow detection
region segmentation
classifier
gradient direction
