| Data representation (compression) and content description are two basic operations in any imagery database. In this work, we propose a wavelet system that can integrates these two operations naturally in an unified framework. Images are compressed by the state-of-the-art coding techniques and indexed in the aspects of color, texture, and object shapes simultaneously.; For data compression, we developed two wavelet schemes. The first one is an embedded multiwavelet coder with GHM filter and the second one is an embedded wavelet coder with multistage vector quantization. Both coding schemes fully exploit the correlation between subbands, achieve natural progressive quantization and superior rate-distortion performance, and are computational efficient.; As with the retrieval of wavelet image, we proposed a system that is closely based on data compression. We investigated a joint set of features—color, texture, and shape—for complete content description, and the descriptors are by-products of wavelet compression. The feature sizes are carefully controlled to reduce the matching complexity and to speed up the retrieval process. Finally, with the scalability property of wavelet coding, the retrieved items can be progressively transmitted to the user for efficient confirmation.; To summarize, our work demonstrates that wavelets are suitable for both content description and compression. And the compression and description can be integrated naturally to simplify the management of a database. |
