| Advances in computer networks and multimedia systems have led to a rapid growth in large digital image databases. Sophisticated image management techniques are needed to provide efficient storage and retrieval of such image collections. Furthermore, efficient access and reproduction afforded by digital media leads to problems in enforcing intellectual property rights.; This thesis addresses two key issues in multimedia databases. The first issue is efficient management of large image libraries. We introduce a system which efficiently stores a large collection of images while simultaneously providing content-based access to the images. The new system is designed to minimize a weighted sum of the expected compressed image sizes and the expected query response time. The image management system supports several advanced database features, including compressed data modification and progressive refinement retrieval based on low complexity hit pattern matching. In experimental results based on a 1.4 Mbyte image database, the number of bits read during a typical retrieval was less than 1000 bytes (i.e., less than 0.08% of the database).; The second issue we address is watermarking digital multimedia, e.g., images, audio, and video. Watermarking is the process of encoding hidden copyright information in data by making small modifications to the data samples. To be practical, the hidden information must be perceptually invisible yet robust to common signal processing operations. We introduce a technique to watermark multimedia data using the masking properties of the human auditory and visual systems. The watermark is constructed by breaking each audio, image, or video signal into smaller segments and adding a perceptually shaped pseudo-random sequence. The noise-like watermark is statistically undetectable to prevent unauthorized removal. Furthermore, the author representation we introduce resolves ownership deadlock. The watermarking technique is shown to be extremely robust to a wide range of data distortions, including low bit rate coding, colored noise, multiple watermarks, cropping, resizing, and resampling. |
