|Title||Issues in Information Hiding Transform Techniques|
|Year of Publication||2002|
|Series Title||NRL Memorandum Report|
|Full Text|| |
In this paper, we give a brief review of prevailing transform embedding techniques for information hiding, and we discuss the robustness and detectability of two specific methods. Spatial embedding inserts messages into image pixels. Transform embedding embeds a message by modifying selected frequency coefficients of the cover image. Ideally, the message embedded with transform techniques is more robust to different image processing attacks and more transparent to human visual systems than that of spatial domain techniques. Transform embedding techniques that we are particularly interested in are the Discrete Fourier Transform (DFT), the Discrete Cosine Transform (DCT) and the Discrete Wavelet Transform (DWT). Detection of the hidden message and robust embedding are two of the principle objectives of our information hiding research. We show that a watermarked image that is perceptually invisible in the spatial domain may fail our detectability test. Our approach to detectability is based on DFT domain analysis. Embedding a message under the least significant bits of pixels (LSB embedding) is commonly used. We investigate ways to improve the robustness of LSB embedding and propose a method for protecting embedded data against two specific forms of filtering.