节点文献

用于虹膜识别的旋转角度估计算法研究

Rotate Angle Estimating for Iris Recognition

【作者】 王雪

【导师】 卢朝阳;

【作者基本信息】 西安电子科技大学 , 通信与信息系统, 2008, 硕士

【摘要】 虹膜图像采集时,由于人脸的偏转或是眼珠的转动,造成采集到的虹膜图像与原始虹膜数据库中的图像有旋转夹角产生。本文对如何估计两幅虹膜图像间的旋转夹角这一课题进行了初步的研究,提出了两种基于图像矩的旋转角度估计算法。根据几何矩的物理特性知某区域的“取向”可由此区域主轴与水平轴的夹角表示,主轴则是关于此区域转动惯量最小的一条直线。因此某一区域关于水平轴的旋转角度可以理解为这一区域的“取向”。基于几何矩描述的旋转角度估计算法将分割后的眼睛区域作为目标区域,眼睛区域关于水平轴的旋转角度则可以通过计算该区域最小转动惯量得到。由于Zernike矩具有旋转不变性,旋转后的虹膜区域与未旋转的虹膜区域的Zernike矩模值是相同的,只存在相位角的偏差,因此则可以通过计算两幅图像Zernike矩值对应的相位角之差估计其相对旋转角度。基于Zernike矩的角度估计算法将定位后虹膜区域的灰度图像作为目标区域,并将其映射到单位圆上,然后计算其Zernike矩值所对应的相位角,通过与模板图像Zernike矩所对应相位角之间的数学运算得到两幅图像之间的相对虹膜旋转角度。实验结果表明,两种算法均能较准确估计出虹膜的旋转角度,对系统性能的提升有一定的帮助。

【Abstract】 On the collection of iris image, there is a rotate angle between the collected iris image and the original iris image in the database because of the deflexion of the face or the turn of the eyeball. In this paper, the estimation of this rotate angle is researched, and two novel algorithms are proposed to calculate this angle based on the image moment.According to the physical characteristic of the geometric moments, it is known that the orientation of a region can be represented by the angle between the horizontal axis and the principal axis of this region which is the line with the smallest turning inertia of the region. So the angle of a region rotated from the horizontal axis can be regarded as the orientation of this region. Making the segmented iris image as the target region, the geometric moments based angle estimating method can get the rotate angle of the iris by calculating the smallest turning inertia of the target region.Because Zernike moment has rotation invariance, the primal iris image and the rotated iris image have the same Zernike moment modules except for the different phase angle. The relative rotate angle can be got by calculating the difference between the Zernike moments phase angles of different images. The Zernike moment based angle estimating algorithm regards the located iris gray image as the target region and maps it on the unit disk. Then the corresponding phase angel of the Zernike moment is computed and the relative rotate angle can be obtained according to the Zernike moment phase angles of two iris images.The experimental results shown that both these two proposed algorithms can basically estimate the rotated angle and improve the performance of the iris recognition system.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络