Image Integrity, Copyright and Palm-Vein Biometric-Based Personal Identity Verification Using Wave Atom Transform


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date3 Sep 2019


There is an increasing need of authorized access to sensitive resources and rightful utilization of private data. The use of human biometrics has been accepted worldwide as a successful mode for personal identity verification for authorized access to both digital services and physical localities. Similarly, verification of data integrity has been traditionally provided by cryptographic hash functions, such as SHA-1 and MD5. However, with the recent explosion of connectivity-platforms and information, a huge amount of data is generated in multimedia form, which include images, videos, audio, documents and computer graphics etc. As a major subset of multimedia content, verification of image integrity and its copyright, for authorized utilization, is mainly provided by perceptual image hashing and digital watermarking techniques, respectively. This research focuses on reliable biometric verification and trustworthiness of printed domain images. The objective of reliable biometrics is to ensure privacy of unique personal features in the biometric context. The objective for printed image verification is to achieve print-scan robustness to provide content and copyright verification under printing and scanning processes.

This thesis investigates capabilities of wave atom transform (WAT) for better image characteristics representation in two application areas: (i) privacy-preserving biometric-based personal identity verification, and (ii) printed media verification. The key challenge in both the applications is to extract desired features for the required and efficient representation of the captured signal. WAT offers sparser expansion and better capability to extract texture features in an image. It can extract features in several scale bands which illustrates a better classification of wave atom coefficients.

The thesis proposes generation of a lightweight and privacy-preserving template for palm-vein-based human recognition. Palm-vein modality is difficult to capture and duplicate without the will of its possessor. Palm-vein traits are extracted in the WAT domain. A binary feature vector known as palm-vein code is generated using original biometric features from each sample palm-vein image. Palm-vein code safeguards privacy of the original features by exhibiting properties of non-invertibility and cancelability. The proposed scheme keeps a decent recognition performance along with less computational and storage requirements.

A hybrid construction of WAT-based image hash is proposed which provides a better tradeoff between print-scan noise resistance and discrimination capabilities. Experimental results demonstrate that the proposed hashing scheme can provide integrity verification of printed media. The proposed print-scan resistant hybrid image hashing is further employed, along with the public-key cryptographically, for paper document authentication. The thesis also investigates feasibility of WAT-based digital watermarking to support copyright verification of printed images under both print-scan and print-cam processes.