基于SCM的虹膜干扰区域去除算法

为了提高虹膜识别率，提出了一种新的睫毛及眼睑区域定位算法。为了提取呈不同角度分布的睫毛，结合SCM模型和动态交叉熵准则对归一化虹膜图像进行多次迭代，根据图像的灰度分布特性，选取合适迭代结果，对多幅迭代图像进行边缘像素点跟踪融合来获得理想的干扰区域轮廓定位。采用高斯金字塔尺度变换与霍夫变换相结合的方法对眼睑的类椭圆区域进行拟合，进而获得连续的眼睑边缘，实现对归一化虹膜图像中干扰区域的准确定位。实验结果验证了方法的有效性。

Iris noise removal algorithm based on SCM

This paper presents a novel method of eye lashed and eyelid regional location in iris images. In order to extract the distribution of different angles eyelashes, it puts the normalized iris image to carry on many iterations using SCM. According to the image gray distribution feature, it uses cross entropy criterion and minimum cross-entropy criterion to select the better iteration results. In order to get a good noise profile targeting region, it uses tracking and fusion methods to the edge pixels in multiple iterations images. It adopts the method of combining the Gaussian pyramid scale transform and Hough transform, fitting the elliptic region of the eyelids, to obtain a continuous edge of the eyelid, to achieve the exact location of the noise region in the normalized iris images. Experimental results confirm the validity of this method.