| 摘要 |
<p>In a multiresolution image processing system images are stored in files that contain thumbnail data as well as a full image data structure. The image data is preferably wavelet or wavelet-like transform coefficients, generated by applying a wavelet or wavelet-like transform to an image multiple times. Data representing mid-level resolution images are generated on the fly by extracting from the full image data structure only the data needed for the user or application selected resolution level. If the user has selected a subset of the image for viewing at a higher resolution level, a corresponding mid-level resolution image is constructed by extracting from the full image data structure the data needed for the user specified image portion at the user or application selected resolution level. The full image data structure is preferably encoded and stored in a manner allowing the image data for mid-level resolution images to be efficiently extracted without having to compute or recompute any image coefficients. By storing only thumbnail data and the full image data in an image file, and producing image data structures for other resolution levels on the fly, the storage requirements for image files are reduced. Since the image data structures for other resolution levels can be generated with minimal computational resources in a small fraction of a second, in most implementations the user will not be adversely affected by the non-inclusion of the other resolution levels in the image file.</p> |
