Research Issues

• Which type of wavelet decomposition is appropriate for different image databases? Of particular interest is the possibility of using recursive wavelet decomposition developed recently at UCSB [14]. These wavelets are computationally very efficient and can be implemented in real time.
• The use of more conventional subband decomposition [68] in developing subimages suitable for image storage and browsing. The comparison of their performance with regard to wavelet decomposition.
• Image compression based on wavelet decomposition will be important in storing large amounts of data efficiently. Wavelet decomposition permits easy integration of image compression techniques, and research issues involve whether lossy or lossless compression should be used and at what resolution level.
• The use of selected features of the highest (coarse) level of the multiresolution decomposition for indexing and search.
• The extraction of metadata will be an important issue, and depends on our decisions concerning the metadata model that we will adopt.
• How to setup the storage devices, and how to efficiently access the subimages from these devices. Note that to ``reconstruct'' the images from the wavelet coefficients, and computational resources must be provided either at the storage or at the user location for this purpose.