基于DCT和修正CSLBP的局部特征描述子  

Local feature descriptor based on DCT and modified CSLBP

作　　者：颜普[1] 朱明[1] 唐俊[1] 梁栋[1] 

机构地区：[1]安徽大学电子信息工程学院,安徽合肥230601

出　　处：《华中科技大学学报（自然科学版）》2014年第9期38-43,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61172127);高等学校博士学科点专项科研基金资助项目(20113401110006);安徽省自然科学基金资助项目(1408085MF121);安徽大学博士科研启动基金资助项目(02303319-33190182);安徽大学青年骨干教师培养项目(02303301-12333010284)

摘　　要：为了更好地兼顾特征描述子的鲁棒性和维数,提出了一种基于离散余弦变换(DCT)和修正中心对称局部二值模式(CSLBP)的局部特征描述子.综合利用了修正CSLBP算法的快速性、亮度不变性、旋转不变性和DCT算法的去相关性、能量集中性、可分离性.首先利用Hessian-Affine算子进行区域检测并进行规范化处理,然后在局部旋转不变坐标系下计算CSLBP描述子,最后应用DCT压缩该修正的CSLBP描述子,得到了一种64维的局部特征描述子.实验结果表明:提出的描述子在图像模糊变换、视角变换、仿射变换、线性亮度变化和JPEG压缩变换下具有良好的性能,特别对于纹理信息较强的图像序列集有更好的效果,而且该描述子生成和匹配速度都较快,适用于实时性要求较高的应用.In order to balance the robustness and the dimension of a feature descriptor,a local feature descriptor based on the discrete cosine transform（DCT）and modified center-symmetric local binary pattern（CSLBP）was proposed.The rapidity,rotation invariance and brightness invariance of modified CSLBP algorithm were combined with the de-correlation,energy compression and separability of DCT algorithm.Firstly,the affine invariant region was detected using by Hessian-Affine detector and then was normalized.Secondly,the CSLBP descriptor was obtained under the locally rotation invariant coordinate system.Finally,the modified CSLBP descriptor was compressed by DCT algorithm.On the basis of that,a 64-dimension local feature descriptor was built.The comprehensive experimental results demonstrate the good effectiveness of the proposed descriptor for the image with blurring changes,viewpoint changes,affine changes,linear illumination changes and JPEG compression changes.Particularly,the proposed descriptor has better results for the image sequences involved strong texture information and the speed of the descriptor generation and matching are faster,it is suitable for higher real-time requirements of applications.

关 键 词：局部特征 特征描述子 离散余弦变换 中心对称局部二值模式 图像匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]