摘要
提出一种融合多种特征的图像过曝光区域检测算法。利用像素的亮度特征和颜色特征,并新引入亮颜特征和边界邻域特征来构成特征向量,用L2正则化逻辑非线性回归方法对训练样本的特征向量进行训练,得到最优分类器模型。对实验图像进行过曝光区域检测,结果显示,相较于亮度阈值法和基于亮度和颜色特征的常规检测方法,引入新特征后的改进算法检测出的过曝光范围区域连通性更好。
An over-exposed region detection learning algorithm which uses multiple image fea- tures is proposed in this paper. In this algorithm pixel's brightness and color features, as well as novel features of light-chrominance and boundary neighborhood are used to construct the feature vector. The L2 regularized logistic regression method is used to obtain the optimal classifier mod- el. Experimental results show that compared to the direct intensity threshold method and other method based on brightness and color features, the detected over-exposed regions by the proposed algorithm are better in terms of regions connectivity.
出处
《西安邮电大学学报》
2015年第6期5-9,共5页
Journal of Xi’an University of Posts and Telecommunications
基金
国家自然科学基金资助项目(61202183)
陕西省国际科技合作计划资助项目(2013KW04-05)
陕西省国际科技合作与交流计划资助项目(2015KW-014)
关键词
过曝光检测
L2正则化逻辑非线性回归
连通性
特征提取
over-exposed region detection, L2 regularized logistic regression, connectivity, lea- ture extraction