Khalid Hassan, Mohamed Edris (2016) Robust image watermarking techniques using image features. PhD thesis, Universiti Malaysia Pahang (Contributors, UNSPECIFIED: UNSPECIFIED).
|
Pdf
Robust image watermarking techniques using image features.pdf Download (698kB) | Preview |
Abstract
This thesis addresses the issue of image watermarking robustness against attacks and especially geometrical attacks. The objectives of this research were, improve robustness watermarking technique based on local image features, and propose robust zero watermarking technique according to the global features of image, To obtain the local feature of image, the feature points extractor is very important. The improved robustness watermarking scheme adopted better feature points extractions named Brief and Efficient Scale Invariant Feature Transform, also adopted another feature points extractions named grouping Harris corner, so they can choose more robust feature points, then increase the robustness of watermark. This scheme used two watermarking techniques, namely, local circular region and block discrete cosine transform, to embed the watermark into two types of regions and extract it. To embed in local circular region, Brief and Efficient Scale Invariant Feature Transform extracts feature points, and then Local Circular Regions for embedding are found, finally, the watermark is embedded into Local Circular Regions by using Histogram. To embed into block discrete cosine transform, grouping Harris corner extracts feature points, and then the image is divided into 80 × 80 non-overlapping block to find candidate blocks, then each candidate block is divided into 8×8 non-overlapping sub-blocks and embeds the watermark in the DC components of each sub-block using its HVS-based embedding strength. For extracting watermark from Local Circular Regions, firstly, robust BE-SIFT feature points is extracted, then, Local Circular Regions are found, finally, the local histogram is computed to extract the watermark. For extracting watermark from Block DCT, first, grouping robust Harris corner feature points is extracted. Second, Delaunay tessellation and triangle matching are applied to restore the probe image. Third, the probe image is divided into 80×80 non-overlapping blocks. Fourth, each block is divided into 8×8 non-overlapping sub-blocks. Fifth, any sub-block is transformed into DCT sub-block. Sixth, the watermark is extracted from DC values. The experimental results showed that the improved scheme is robust against a wide variety of attacks. In particular, it is more robust against geometric attacks. The proposed method has 100% good performance for resisting attacks on Lena and Pepper images, and at least 84% good performance on Barbara and Plane images compared with Deng’s method, and it has 100% good performance compared with other methods. In addition, robust zero watermarking scheme based on global feature of image using complex Zernike moments is proposed, the contribution of this method is adopting complex Zernike moments, which can provide better robustness against geometric attacks and provide more information about image and more space for embedding and zero watermarking. Before calculating complex Zernike moments, standard translation and scaling of the tested image is performed, after getting the binarization of the argument value, the feature image is constructed and XOR operation with logo image is executed to generate verification image. Experimental results demonstrated that the proposed scheme has strong robustness to various attacks especially geometric attacks. The proposed scheme has at least 70% good performance of resisting attacks on tested images compared with the other methods.
| Item Type: | Thesis (PhD) |
|---|---|
| Additional Information: | Thesis (Doctor of Philosophy in Computer Science) -- Universiti Malaysia Pahang – 2016 |
| Uncontrolled Keywords: | watermarking; robust image |
| Subjects: | T Technology > T Technology (General) |
| Faculty/Division: | Faculty of Computer System And Software Engineering |
| Depositing User: | Ms. Nurezzatul Akmal Salleh |
| Date Deposited: | 12 Jan 2017 05:49 |
| Last Modified: | 17 Nov 2021 07:37 |
| URI: | http://umpir.ump.edu.my/id/eprint/15829 |
| Download Statistic: | View Download Statistics |
Actions (login required)
 |
View Item |