改进Canny算法的红外人脸图像边缘轮廓提取  

Edge contour extraction of infrared face image based on improved Canny algorithm

作　　者：王鑫刚 田军委[1] 于亚琳 王沁[1] 张杰 WANG Xin'gang;TIAN Junwei;YU Yalin;WANG Qin;ZHANG Jie(School of Mechatronic Engineering,Xi'an Technological University,Xi'an 710021,China;School of Opto-electronic Engineering,Xi'an Technological University,Xi'an 710021,China;Inner Mongolia North Heavy Industries Group Co.,Ltd.,Baotou 014030,China)

机构地区：[1]西安工业大学机电工程学院,陕西西安710021 [2]西安工业大学光电工程学院,陕西西安710021 [3]内蒙古北方重工业集团有限公司,内蒙古包头014030

出　　处：《应用光学》2023年第1期61-70,共10页Journal of Applied Optics

基　　金：陕西省科技厅重点研发计划(2022GY-068,2022GY-316)。

摘　　要：红外人脸图像的边缘轮廓特征对于红外人脸检测、识别等相关应用具有重要价值。针对红外人脸图像边缘轮廓提取时存在伪边缘的问题,提出了一种改进Canny算法的红外人脸图像边缘轮廓提取方法。首先通过对引导滤波算法引入“动态阈值约束因子”替换原始算法中的高斯滤波,解决了原始算法滤波处理不均匀和造成红外人脸图像弱边缘特征丢失的弊端;接着对原始算法的非极大值抑制进行了改进,在原始计算梯度方向的基础上又增加了4个梯度方向,使得非极大值抑制的插值较原始算法更加精细;最后改进OTSU(大津)算法,构造灰度-梯度映射函数确定最佳阈值,解决了原始算法人为经验确定阈值的局限性。实验结果表明:提出的改进Canny算法的红外人脸轮廓提取方法滤波后的图像,相较于原始Canny算法滤波处理,信噪比性能提升了34.40%,结构相似度性能提升了21.66%;最终的红外人脸边缘轮廓提取实验的优质系数值高于对比实验的其他方法,证明改进后的算法对于红外人脸图像边缘轮廓提取具有优越性。The edge contour features of infrared face images are of great value for applications related to infrared face detection and recognition. Aiming at the problem of false edges in the edge contour extraction of infrared face images, an edge contour extraction of infrared face image based on improved Canny algorithm was proposed. Firstly, by introducing dynamic threshold constraint factor to the guided filtering algorithm to replace the Gaussian filtering in the original algorithm, the disadvantages of uneven filtering processing and the loss of weak edge features in the infrared face image were solved. Then, the non-maximum suppression was improved, and four gradient directions were added on the basis of the original gradient direction, which made the interpolation of non-maximum suppression more precise than the original algorithm. Finally, the OTSU algorithm was improved by constructing a gray-gradient mapping function to determine the optimal threshold value, which solved the limitation of the original algorithm to determine the threshold value by human experience. The experimental results show that, compared with the original Canny algorithm filtering processing, the performance of signal-to-noise ratio of the filtered image from edge contour extraction of infrared face images based on improved Canny algorithm is improved by 34.40%, and the performance of structural similarity is improved by 21.66%. Finally, the quality coefficient value of the experiment of infrared face edge contour extraction is higher than that of other methods in the comparison experiment, which proves that the improved algorithm has superiority for the edge contour extraction of infrared face image.

关 键 词：红外人脸图像 边缘轮廓提取 CANNY算法 引导滤波 OTSU算法 

分 类 号：TN29[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]