基于多特征融合的运动阴影去除算法  

Moving Shadow Removal Algorithm Based on Multi-feature Fusion

作　　者：赵林林[1] 陈一民[1] 邹一波[1] 

机构地区：[1]上海大学计算机工程与科学学院,上海200444

出　　处：《微型电脑应用》2017年第3期71-74,80,共5页Microcomputer Applications

摘　　要：在运动目标检测中,由于运动阴影与前景物体具有相似的运动特征,传统的方法很难区分运动对象及其阴影。为了解决这一问题,提出了一种基于多特征融合的阴影去除算法。在HSV颜色空间下利用阴影所具有的色度不变性特征对前景图像进行预处理并得到阴影的候选区域;利用小波变换分析候选区域和与之对应背景区域纹理特征的相似程度,去除其中的相似成分得到最终的检测结果。实验结果表明,提出的方法在不同场景下能够有效地去除运动阴影区域。In the field of motion detection, the traditional method is difficult to distinguish the moving object and its shadow due to the similar motion features between the moving shadow and the foreground object. In order to solve this problem, a novel shadow removal algorithm based on multi-feature fusion is proposed in this paper. Firstly, in the HSV color space, we use chromaticity invariance feature of shadow to obtain the candidate region of shadow. Then, we use the wavelet transform to ana- lyze the texture similarity between the candidate region and background region, and remove the similar wavelet components to obtain the final detection result. The experimental results show that our method can effectively remove moving shadow in dif- ferent scenes.

关 键 词：运动阴影去除 色度不变性 纹理相似性 小波变换 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]