Objectives

The objective of the digital watermarking group is to be at the leading edge in research, development and applications of digital watermarking technologies for digital multimedia data, with emphasis on copyright protection and authentication of image, audio, video, document and vector graphic data.

The main activities of the group focus on the research, development and analysis of novel digital watermarking techniques for multimedia data and documents. For grey scale and colour image watermarking, adaptive-based watermark embedding and retrieval techniques are investigated by the group to provide robust, fragile and semi-fragile watermarking against various malicious attacks and system processing. Malicious attacks include geometrical distortions such as rotation, scaling and translation. However, some system processing operations such as digital-to-analogue-to-digital (DAD) conversion, present a challenging problem that is inherent in many traditional printing processes.

Theoretical modelling into the print-scan processes is currently being studied with the aim to enhance the watermark embedding and retrieval techniques. A related research tropic of interest is image restoration of tampered watermarked images commonly found in image authentication. Irregular sampling and reconstruction techniques are investigated for incorporation in the restoration of tampered watermarked images.

Another research area for image watermarking is the investigation of using non-conventional orthogonal transforms such as Slant, Integer Consine Transform, Pinned Sine Transform and independent component analysis for digital watermarking. Moreover, some of these transforms have been found to achieve significantly better performance than conventional transforms used in copyright protection and authentication.

Our group is also very active in binary image or document watermarking research, particularly in the analysis of data hiding and detection techniques in the spatial domain. Current research areas include the data hiding capacity in text-based images and the development of image quality metrics based on novel perceptual models for evaluating binary images.

Our research interests in video watermarking are focused on the design and development of robust digital watermarking techniques for compressed video formats such as MPEG, as well as watermark attacks and their counter-measures.

In particular, we have developed and patented a fast MPEG watermarking scheme to embed secure watermarks efficiently and invisibly into MPEG bit stream without the need for MPEG decompression. Novelties included in this technology include compressed domain block classification, enhanced bit rate control, and enhanced blind detection for drift-compensated video. We have also investigated the effect of linear and non-linear collusion attacks on digital fingerprints (personalised watermarks) and devised new techniques to counter them.

Over the past year, due to the strong interest in information security, members of the watermarking research group have secured substantial research funding of more than $800k for binary and document image watermarking from LIT and NTU. Through external requests, the group have also conducted a number of well-received seminars, workshops and courses in digital watermarking and steganography to the local industry and government agencies.

As part of its mission, the research group will continue to actively strive for innovation and advance the above mentioned research areas through close collaboration with researchers from both local and international tertiary institutions, as well as the industrial sector.