A comparison between the performance of DWT and LWT in image watermarking

Dujan B. Taha,Taha Basheer Taha,Najla Badie Al Dabagh

doi:10.11591/eei.v9i3.1754

Dujan B. Taha, Taha Basheer Taha + Show 1 more

Open Access

https://doi.org/10.11591/eei.v9i3.1754

Copy DOI

Abstract

Image watermarking algorithms can be implemented using time domain or frequency domain-based algorithms. Frequency domain watermarking produces watermarks with higher robustness; hence, many attempts have been proposed in literature using different transformations as DCT, SVD, and DWT. DWT was widely used for many reasons as its spatio-temporal feature, in which the alteration of certain portion will affect only the affected portion. In this paper an experimental comparison between the traditional DWT and the second generation of wavelet which is LWT is initiated. The experimental tests evaluated the performance of both transforms in terms of image quality and watermark robustness.

Highlights

Digital image watermarking is the process of inserting some data as ownership information into images for different purposes as copyright protection, prevent unauthorized duplication and authentication [1]
lifting wavelet transform (LWT) on other hand, had been used in recent watermarking attempts. It is combined with SVD in [14, 15], and it is used with support vector machine (SVM) in [16]
For SSIM [24], which consider the human vison system sensitivity and has better estimation than peak signal to noise ratio (PSNR) [25, 26], the results was equal to 1 for all images except I10 in discrete wavelet transform (DWT) it was 0.99, due to the embedding in middle frequency area with low intensity

Summary

INTRODUCTION

Digital image watermarking is the process of inserting some data as ownership information into images for different purposes as copyright protection, prevent unauthorized duplication and authentication [1]. A watermark that survive different attacks is called a robust watermark Another requirement of the watermark is the invisibility, which is keeping the perceptual quality of the original images high. It is combined with SVD in [14, 15], and it is used with support vector machine (SVM) in [16] Both transforms are commonly used in literature, and both have the same objective which is transform the image into frequency domain coefficients but with different methods and calculations. A comprehensive experiments and evaluation had been initiated to compare the performance for both transforms by embedding the same watermark with the same intensity into similar bands and coefficients in DWT and LWT.

THEORITICAL BACKGROUND

RESULTS

CONCLUSION