Performance evaluation of an incorporated DCT block-based watermarking algorithm with human visual system model

Abstract

In this paper we propose a new approach to predict a fixed fidelity level of a watermarked image while increasing the detection rate. The new approach incorporates a DCT block-based algorithm with human visual system (HVS). The performance of the new approach has been defined in terms of its effectiveness (detection rate) and fidelity. We show experimentally how our novel approach can be used to control distribution of a watermark to get a higher detection rate as well as better fidelity.

Introduction

There are numerous digital watermarking algorithms in the literature. Digital watermarking is a technique, which embeds data (mark) in digital and analog content in order to identify, track, manage and enhance the usefulness of such content (Watson, 1993a, Piva et al., 1997). Digital watermark is an invisible structure embedded into the host media. To be effective, a watermark must be imperceptible within its host (Watson, 1993a), discrete to prevent unauthorized removal, easily extracted by the owner, and robust to distortions. According to the literature in this field, many watermarking techniques for images and video were proposed, mainly focusing on the invisibility of the watermark and its robustness against various signal manipulations and hostile attacks (Piva et al., 1997, Eggers et al., 2000, Osberger et al., 1998, Kim et al., 1999, Degullaume et al., 2002, Hartung and Kutter, 1999).

Nowadays, there is a trend towards approaches that make use of information about the HVS to produce a more robust watermark. Such techniques use explicit information about the HVS to exploit the limited dynamic range of the human eye. In order to design an optimal digital watermarking algorithm or to optimize an existing digital watermarking algorithm it is important to take the HVS into account. Therefore, our novel model takes advantages of the HVS and incorporates it with DCT-block based algorithm. To recognize the effects of such a system on a digital watermarking algorithm, a brief introduction to the HVS is given in Section 1.1.

The retina of our eye splits a visual signal into different components and each component excites the visual cortex via separate channels (Watson, 1993b, Wolfgang et al., 1999a, Osberger et al., 1998, Kim et al., 1999). Each component has the following characteristics: the spatial location in the image, the frequency of the image, and the orientation of the signal (horizontal, vertical, or diagonal).

Based on the knowledge of the structure of the human eye and human visual system, a hypothetical cortex transform (Wolfgang et al., 1999a) has been devised that models the known properties. When two signals have similar component characteristics, they excite the same channel in the cortex but are subject to the masking effect. Masking occurs when the detection threshold is increased because of the presence of another stronger signal with similar characteristics. However, human perception mechanisms are not uniform. For example, the response of the HVS varies with the spatial frequency, brightness, and color of its input. This suggests that all components of a watermark may not be equally perceptible. Perceptual variation can be measured, and models constructed to account for them. A perceptual model generally attempts to account for three basic types of phenomena: sensitivity, masking, and pooling (Miller et al., 2001). Detailed description of these phenomena can be found in (Podilchuck and Zeng, 1997, Bartolini et al., 1998, Miller et al., 2001, Wolfgang et al., 1999a, Wolfgang et al., 1999b).

The paper is organized as follows: Section 2 presents the concept of incorporating the HVS model with the DCT block-based algorithm and Section 3 describes the testing procedure. Section 4 presents the experimental results and finally, Section 5 discusses conclusions and proposes future research problems worth investigating.