A Robust Biometric Image Texture Descripting Approach

Abstract

The huge demand in online transactions calls for a secure, safe and accurate authentication system .The biometric system such as face, iris, fingerprint, gait has already replaced the existing manual inspection process and surveillance systems in many disciplines. Amongst all these biometrics, face is more attractive as it provides information such as identity, expression, gender, ethnicity and age of an individual.Specially for surveillance purposes,the data acquisition for face is much more simpler and can be obtained without the subjects knowledge and cooperation (simply by installing camera in public areas) when compared to fingerprints and iris data accumulation.In this paper an edge texture feature using different weight mask coding is utilized for face recognition.Five different sign and difference operators named LSH LC, LSH GC,LDH LC, LDH GC and LSH LGV are developed and used to code the image texture feature.Each image is decomposed into four subset images which are used to generate the texture features. A decision fusion technique is then used for feature classification. The main advantage of this approach lies in the fact that they are computationally inexpensive when compared to most texture descriptors.The feature descriptor is applied for face biometric recognition to demonstrate the effectiveness of each approach in extracting textural features.It can also be tested with medical images or other pattern recognition applications.The dataset used for training and testing have considerable variances in lighting,viewpoint and other factors so that the potential of the feature extractor, when subjected to any kind of variations, can be judged.