A Study of Low Distortion Sampling Techniques and Distortion Free Data Hiding Algorithms
| Autor: | Nien-Ching Huang, 黃年慶 |
|---|---|
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 97 Photorealistic Rendering (PR) is an active research topic in the computer graphics community. It aims to produce an image of the highest quality that is indistinguishable to a real scene. By utilizing a physical based equation, the rendering equation, we can describe a virtual scene with physical character, and then we can render the realistic image by computer operation. In order to calculate the rendering equation, the Monte Carlo method is most often used. The sampling technique is the actual operating technique of the Monte Carlo method. In the sampling technique, the variance represents the potency of the whole operation. Usually, high variance means low efficiency. Compared to the basic sampling technique and the uniform sampling technique, the stratified sampling technique is an effective sampling technique with lower variance. This low distortion sampling technique is an advanced technique based on the stratified sampling technique in order to enhance the potency further. In contrast, in the PR topic, the basic technique samples random samples on the light source. For a different light source, the corresponding sampling technique is different. There are corresponding sampling techniques for familiar geometries, such as rectangle, triangle, disc, cylinder cone, ellipse, and torus. In the first part of this paper, we propose a low distortion sampling technique for an ellipse. By the low distortion mapping relation, we calculate the corresponding sampling function. Our technique provides ellipse sampling with efficiency, stratified sampling, and low distortion. On the other hand, in current steganographic techniques, the secret message is hidden in the value domain of the cover medium. When we hide the message into the cover medium and get the stego medium, there are always distortions between the cover and the stego medium. In existing researches, the goals usually are to reduce distortion or to enhance the capacity under the given distortions. Unlike traditional steganography, distortion free steganography hides the message in the structure domain of the medium. In that way, when we use the standard viewer to open those two files, we can deduce that there is no distortion between the cover and the stego medium,. Since the message is hidden in the structure domain of the media, for different media the corresponding distortion free steganographic techniques are different. The most common distortion for steganography is the permutation scheme, which works on those data structures with the condition that the order of the element is independent to the final result, such as the 3D polygonal mesh file. In the second part of this paper, we propose two distortion free steganographic techniques. The first technique belongs to the permutation scheme. Our technique is an advance technique of Bogomjakov et al’s algorithm. By changing the encoding part, our technique achieves higher capacity in the same condition. The second distortion free steganographic technique is based on the Radiance RGBE file, which is a commonly used high dynamic range image. As we say, the message is hidden in the structure domain. Our technique is based on the adaptive run-length encoding, which is used in the Radiance RGBE file. By changing the encoding process of the ARLE, we can achieve the goal of distortion free steganography. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|