LZW differences between TIFF and GIF
LZW compression

Prev:gif-lzw-uncompress ( seq code-size -- byte-array )


There are some subtle differences between the LZW algorithm used by TIFF and GIF images.

Variable Length Codes
Both TIFF and GIF use a variation of the LZW algorithm that uses variable length codes. In both cases, the maximum code size is 12 bits. The initial code size, however, is different between the two formats. TIFF's initial code size is always 9 bits. GIF's initial code size is specified on a per-file basis at the beginning of the image descriptor block, with a minimum of 3 bits.

TIFF and GIF each switch to the next code size using slightly different algorithms. GIF increments the code size as soon as the LZW string table's length is equal to 2**code-size, while TIFF increments the code size when the table's length is equal to 2**code-size - 1.

Packing Bits into Bytes
TIFF and GIF LZW algorithms differ in how they pack the code bits into the byte stream. The least significant bit in a TIFF code is stored in the most significant bit of the bytestream, while the least significant bit in a GIF code is stored in the least significant bit of the bytestream.

Special Codes
TIFF and GIF both add the concept of a 'Clear Code' and a 'End of Information Code' to the LZW algorithm. In both cases, the 'Clear Code' is equal to 2**(code-size - 1) and the 'End of Information Code' is equal to the Clear Code + 1. These 2 codes are reserved in the string table. So in both cases, the LZW string table is initialized to have a length equal to the End of Information Code + 1.