On December 19, the following impressive master thesis proposal defense was presented by Eng. Mohamed Baziyad from Computer Engineering department. The title of the proposal is “Towards Optimal Utilization of Temporal Redundancy for Digital Video Steganography”.
The Master thesis will be supervised by Prof. Tamer Rabie and co-supervised by Pro. Ibrahim Kamel
The Abstract of the thesis is:
The massive improvements in the processing power of modern computers, and the advanced storage and networking capabilities in recent years have made processing video segments very popular in the digital world. A video signal is a collection of images (frames) presented successively at a constant rate. Therefore, a video signal can be expressed as a 3D signal where the rows and columns of pixels represent the first and the second dimension, while the third dimension is the time. The 3D nature of video signals has produced an additional source of data redundancy; that is, the temporal redundancy. Utilizing signal redundancy is the fundamental driving force for several digital video applications such as steganography and compression. In this thesis, three
novel techniques are proposed to optimally exploit the redundancy in a video segment; namely, the Flexible Pixogram, the Octree Segmentation, and the Temporal Region-Growing Technique. The Flexible Pixogram is a 1D vector that starts from a certain initial position and then grows in the direction of the motion vector associated with that initial position. On the other hand, the Octree segmentation is the 3D extension of the well-known quad-tree image-based segmentation technique. The Octree segments a video segment into variable-sized homogeneous cubes. Finally, the temporal region-growing technique is the 3D expansion of the 2D region growing technique where the video segment is divided into homogeneous free-shaped regions representing objects in the scene. It is expected that these segmentation techniques will maximize the correlation of data and thus stronger energy compaction is expected in the Discrete Cosine Transform (DCT) domain. In other words, most of the signal will be concentrated within few significant DCT coefficients leaving a large area with insignificant DCT coefficients which can be safely removed as in compression techniques or replaced by the secret data as in steganography. The performance of these proposed techniques for optimizing video redundancy will be demonstrated by developing novel steganography and compression schemes that can outperform competitive state-of-the-art techniques published recently in the literature.