Extensive support for ROI (Region of Interest) encoding, including both max-shift and distortion weighting methods and the ability to specify regions of interest using either a simple geometric description, or an auxiliary mask image. In the latter case, the mask image is thresholded and automatically scaled to match the dimensions of each image component undergoing region of interest encoding. Like the rest of the Kakadu package, all processing is performed incrementally to efficiently utilize memory resources. The mechanisms provided by Kakadu to support ROI encoding should be easily extended to accommodate interactive applications (e.g., a radiologist marking up medical images with regions deserving special attention)
MMX optimizations implemented under both Microsoft Visual C++ and gcc -- Linux versions of the demo executables now incorporate these optimizations.
Upgraded to the new ANSI C++ standard I/O package. This improves namespace protection and standardization across multiple platforms, but take note that versions of gcc prior to version 3.0 do not support the standard in this regard. The demo executables were all compiled under gcc version 3.0 without errors or warnings. Same is true of Microsoft Visual C++.
Various minor bug fixes and incidental improvements.