- The libjpeg and TurboJPEG APIs and associated programs now fully
support lossy JPEG images with 12-bit data precision, without the need
to recompile libjpeg-turbo from source. (12-bit data precision is not
fully SIMD-accelerated. See
https://github.com/libjpeg-turbo/libjpeg-turbo/issues/199.)
- The libjpeg and TurboJPEG APIs and associated programs now fully
support lossless JPEG images with 8-bit, 12-bit, and 16-bit data
precision. (Lossless JPEG does not share any algorithms with lossy JPEG
and is thus not SIMD-accelerated at all.)
- Arithmetic coding is now supported with 12-bit lossy JPEG images.
- Completely overhauled the TurboJPEG API to add the most useful and
requested features from the libjpeg API and to address numerous other
shortcomings.
- Significantly sped up the computation of optimal Huffman tables.