基于改进均值标准差曲线描述子的反射对称轴检测

Reflection Symmetry Axis Detection Based on Improved Mean-Standard Deviation Curve Descriptor

针对反射对称轴难以检测的问题,该文提出一种基于改进均值标准差曲线描述子(Mean-Standard deviation Curve Descriptor,MSCD)的反射对称轴检测算法.该算法首先对MSCD曲线描述子进行改进,使其具有镜像反射不变性并实现对称曲线对检测;然后采用距离约束并使用Hough变换获取图像的局部对称轴;最后通过局部对称轴合并得到最终对称轴.实验结果表明,该算法可实现图像的单对称轴和多对称轴检测,在亮度变化、对比度变化、噪声污染、模糊以及形变情况下,均能够准确定位图像对称轴并具有较强鲁棒性.

Due to the difficulty in detecting reflection symmetry axis,a detection algorithm based on improved meanstandard deviation curve descriptor（ MSCD） is proposed. Firstly,the proposed algorithm improves the MSCD so that its mirror reflection invariance can guarantee the accurate detection of the symmetrical curve pairs. Then the distance constraint and Hough transform are used to gain local symmetry axis of the image. Finally,final symmetry axis can be acquired by merging the local ones. Experimental results showthat the proposed algorithm can realize the symmetric axis detection for image with both single symmetric axis and multi ones. Besides,it can locate the image symmetry axis precisely and robustly under brightness changes,contrast changes,noise pollution,fuzzy and deformations.