Biometric watermark authentication with multiple verification rule

Abstract

The paper describes a biometric watermarking procedure for accurate facial signature authentication. The multiple decision rules of watermark and facial signatures are formulated to map a pattern in an overlapping region between classes to a separable disjoint one. The decision rule of the watermark signature, which is uniquely assigned to a face image, reduces uncertainty in assuming the missing patterns of unknown intruder classes. The present algorithm incorporating multiple signatures improves a recognition error rate from 2.29% to 0.07%. The watermarking approach has attained robustness to various attacks and transformation by focusing on salient facial features.