Color Image Watermarking Using QUATERNION FOURIER TRANSFORM using MATLAB

Introduction

MATLABSolutions demonstrate In this particular task,Color images have not received as much attention from the watermarking community as grey-level images. One com-monly used technique to address color images is to markonly the luminance component. If on one hand this solu-tion is practical due to implementation issues, on the otherhand the luminance component represents the most percep-tive and sensible information.

Consequently for watermark-ing and especially data-hiding methods, it is important notto discard chrominance information because such a channelrepresents an interesting amount of data that can be used forinformation embedding.This paper is organised as follows: previous work on colorimage watermarking is first recalled. In the next sectioncolor images and color images transform are revisited usingthe hypercomplex representationand the Quaternion FourierTransforms (QFT). In the next section a QFT is constructed considering human perception issues. A watermarking sche-me using this transform is afterwards presented and evalu-ated. Finally concluding remarks outline the different issuesof this work.

PREVIOUS WORK

K The aim of this section is to recall briefly important works on color image watermarking and data-hiding techniques.

.
Kutter et. al. proposed a straightforward and practical solution: the blue channel is used as the embedding channel. The authors claim that blue represents the less sensitive channel inside the RGB space. The embedding process uses spread spectrum techniques in the spatial domain.
Piva et. al. proposed to take into account the three RVB components to increase the embedding space. In this work, the power of the mark is based on the human eye sensibility to the light spectrum. The magnitude of the mark is consequently 10 times more significant for the blue component than for the green component and 5 times more significant for the blue component than for the red one. Inthe proposed scheme, the embedding process is an additive(spread-spectrum) embedding scheme in the DCT domain.