Date of Award

Spring 2005

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computing Sciences - (Ph.D.)

Department

Computer Science

First Advisor

Frank Y. Shih

Second Advisor

Stanley Chen

Third Advisor

Chengjun Liu

Fourth Advisor

James A. McHugh

Fifth Advisor

Lawrence O'Gorman

Sixth Advisor

Yun Q. Shi

Abstract

With the fast development of computer technology, research in the fields of multimedia security, image processing, and robot vision have recently become popular. Image watermarking, steganogrphic system, morphological processing and shortest path planning are important subjects among them. In this dissertation, the fundamental techniques are reviewed first followed by the presentation of novel algorithms and theorems for these three subjects.

The research on multimedia security consists of two parts, image watermarking and steganographic system. In image watermarking, several algorithms are developed to achieve different goals as shown below. In order to embed more watermarks and to minimize distortion of watermarked images, a novel watermarking technique using combinational spatial and frequency domains is presented. In order to correct rounding errors, a novel technique based on the genetic algorithm (GA) is developed. By separating medical images into Region of Interest (ROI) and non-ROI parts, higher compression rates can be achieved where the ROI is compressed by lossless compression and the non-ROI by lossy compression. The GA-based watermarking technique can also be considered as a fundamental platform for other fragile watermarking techniques. In order to simplify the selection and integrate different watermarking techniques, a novel adjusted-purpose digital watermarking is developed. In order to enlarge the capacity of robust watermarking, a novel robust high-capacity watermarking is developed. In steganographic system, a novel steganographic algorithm is developed by using GA to break the inspection of steganalytic system.

In morphological processing, the GA-based techniques are developed to decompose arbitrary shapes of big binary structuring elements and arbitrary values of big grayscale structuring elements into small ones. The decomposition is suited for a parallel-pipelined architecture. The techniques can speed up the morphological processing and allow full freedom for users to design any type and any size of binary and grayscale structuring elements.

In applications such as shortest path planning, a novel method is first presented to obtaining Euclidean distance transformation (EDT) in just two scans of image. The shortest path can be extracted based on distance maps by tracking minimum values. In order to record the motion path, a new chain-code representation is developed to allow forward and backward movements. By placing the smooth turning-angle constraint, it is possible to mimic realistic motions of cars. By using dynamically rotational morphology, it is not only guarantee collision-free in the shortest path, but also reduce time complexity dramatically. As soon as the distance map of a destination and collision-free codes have been established off-line, shortest paths of cars given any starting location toward the destination can be promptly obtained on-line.

Share

COinS