0.22beta:
 - Fixed some buffer overruns in huffmanDecoder*() that were causing seg faults
   when decompressing corrupt images. The decompressor shouldn't crash even if
   fed complete garbage now.
 - Optimizations to multiplication macros in iDCT
 - Optimizations to saturation code in iDCT

0.21alpha:
 - Removed floating-point iDCT
 - Added faster integer iDCT. Old iDCT is available with "make oldidct"
 - Removed linkage to math library (was only needed for floating-point)
 - Rewrote huffmanDecoderY/UV functions; should be slightly more efficient 

0.20alpha:
 - Makefile: Removed "accu" and "olddct" targets
 - d518.c: Removed proprietary variables and linkages
 - ov518lib.c: Removed proprietary variables and linkages
 - ov518lib.c: Merging of Y and UV Huffman decoder functions
 - ov518lib.c: Fixed bug where UV quantization table was being used for Y
               dequantization. The iDCT now works as well as OV's version.
 - ov518lib.h: CompInfo now stores location of quantization table (no more
               global variables! Hooray!)

0.19alpha:
 - Makefile: Replaced "integer" target with "float" (integet is now default)
 - Makefile: Added -lm link option to build in math library (for older systems)
 - ov518lib.c: Fixed bug in floating-point iDCT; quality is better than 
               proprietary iDCT now
 - ov518lib.c: Replaced integer iDCT with one that has equal quality to
               floating point iDCT (but is MUCH faster)

0.18alpha:
 - Makefile: Added "integer" target
 - d518.c: Fixed incorrect check of fopen() and fread() return codes
 - d518.c: cinfo.rawLen is set to actual number of bytes read
 - d518.c: Disabled misleading "Image is %d bytes" message
 - grabimage: exit if any command fails
 - ov518lib.c: Fixes to DCT_8x4() comment
 - ov518lib.c: Added experimental integer iDCT. Disabled by default

0.17alpha:
 - First public version
