自动化学报2016，Vol.42Issue(4)：617-630,14.DOI:10.16383/j.aas.2016.c150206
特征联合和旋转不变空间分割联合的局部图像描述符
Local Image Descriptor of Feature Combination and Rotation Invariant Space Division Combination
摘要
Abstract
This paper proposes a novel local image descriptor called FCSCD (feature combination and rotation invariant space division combination descriptor). A new local feature, WLBP (Weber local binary pattern), is proposed which combines Weber binary differential excitation and local binary pattern. A new rotation-invariant space division for feature pooling is also proposed which combines intensity order space division and annular space division. WLBP is computed in a rotation invariant local coordinate system. Intensity order and annular space division are inherently rotation invariant. So, FCSCD obtains rotation invariance without computing principle orientation of the image patch. Compared with other existing descriptors, this combination method makes FCSCD encode various types of information into a histogram, and so it is more distinctive and robust. Experimental results on image matching demonstrate the effectiveness and superiorities of the proposed descriptor compared to the state-of-the-art descriptors including SIFT, CS-LBP, OSID, LIOP, EOD, and MRRID.关键词
局部图像描述符/SIFT/图像匹配/旋转不变Key words
Local image descriptor/SIFT/image matching/rotation invariance引用本文复制引用
许允喜,陈方..特征联合和旋转不变空间分割联合的局部图像描述符[J].自动化学报,2016,42(4):617-630,14.基金项目
国家自然科学基金(61370173),湖州市重点科技创新团队(2012KC 04)资助Supported by National Natural Science Foundation of China (61370173) and Key Science and Technology Innovation Team of Huzhou City (2012KC04) （61370173）
