DEEP64 1.1

Short:        IFF-DEEP 48/64 Bit extensions
Author:       info@ar-kleinert.de
Uploader:     info ar-kleinert de
Type:         docs/misc
Version:      1.1
Architecture: generic

 Previously, the IFF-DEEP file format supported
 the following image data formats basically:

  24 bit -> as RGB  8:8:8       true color pixels
  32 bit -> as RGBA 8:8:8:(4/8) true color pixels + alpha

 In addition to RGB also other color spaces are
 supported and the sample order could vary. Basically
 any kind of bit depth is possible - but yet uncommon.
 Bit depths also could vary for each sample.

 For compression, only uncompressed, RLE and DELTA
 are common. For DELTA, instead of single pixels
 (all samples alternating in a row) separate lines
 per sample are stored for a row.

 Now the following - as a consequent logical extension
 of the previously mentioned definitions - is introduced
 by SView5-Library:

 48 bit -> RGB  16:16:16    true color pixels
 64 bit -> RGBA 16:16:16:16 true color pixels + alpha

 With regards to the Alpha channel it is considered legal
 to load/save/interpret only the upper 8 bits, thus leading
 to effectively 56 bit depth - however generation of 56
 bit files is considered uncommon.

 Instead, prior to loading/saving, the 8/16 bit alpha
 channel should be properly scaled up/down to 16/8 bit.

 For compression this means:

  - uncompressed: no change (raw buffer size)
  - RLE:          multiply "bytes per line" by 2;
                  compress on byte-level
  - DELTA:        multiply "bytes per line" by 2;
                  lookup table remains 16*WORD;
                  delta is relative to UWORD instead
                  of UBYTE
  - other:        to be defined