Abstract

This paper puts forward a novel methodology for Heterogeneous Face Recognition (HFR), where we present a new-fangled image representation technique called the Local Radius of Gyration Face (LRGF), which has been theoretically proved to be invariant to changes in illumination, rotation and noise. Finally, a novel Local Triangular Coil Binary Pattern (LTCBP) is presented so as to apprehend the local variations of the LRGF attributes, and the method has been entitled as the Triangular Coil Pattern of Local Radius of Gyration Face (TCPLRGF). The proposed algorithm has been tested on a number of challenging databases to study the precision of the TCPLRGF method under varying condition of illumination, rotation, noise and also the recognition accuracy of sketch-photo and NIR-VIS image. The Rank-1 recognition accuracy of 98.27% on CMU-PIE Database, 98.09% on Extended Yale B Database, 96.35% on AR Face Database, 100% on CUHK Face Sketch (CUFS) Database, 89.01% on LFW Database and 98.74% on the CASIA-HFB NIR-VIS Database exhibits the supremacy of the proposed strategy in Heterogeneous Face Recognition (HFR) under various conditions, compared to other recent state-of-the-art methods. For reckoning the similarity measure between images, a hybridized approach amalgamating the Jaccard Similarity method and the standardized L₁ norm approach has been taken into account.

中文翻译：

---

旋转面局部半径的三角形线圈模式用于异类人脸识别

摘要

本文提出了一种新的异构面部识别（HFR）方法，在此我们提出了一种新型的图像表示技术，称为回转面部局部半径（LRGF），该方法在理论上已证明不会改变照明，旋转和噪音。最后，提出了一种新颖的局部三角形线圈二值模式（LTCBP），以了解LRGF属性的局部变化，该方法被称为旋转面局部半径的三角形线圈模式（TCPLRGF）。该算法已在许多具有挑战性的数据库上进行了测试，以研究TCPLRGF方法在光照，旋转，噪声变化条件下的精度，以及草图和NIR-VIS图像的识别精度。CMU-PIE数据库98的Rank-1识别精度为98.27％。扩展耶鲁B数据库占09％，AR人脸数据库占96.35％，中大人脸草图（CUFS）数据库占100％，LFW数据库占89.01％，CASIA-HFB NIR-VIS数据库占98.74％展示了所提出策略的优势与其他最新技术相比，在各种情况下的“异质人脸识别”（HFR）中的“人脸识别”功能。为了计算图像之间的相似性度量，将Jaccard相似性方法和标准L 已经考虑了_1种规范方法。