• Shorten Huffman Code Better?

    From Shelwien@21:1/5 to All on Mon Mar 25 01:59:27 2019
    shortening my compression technique's Huffman code header.

    deflate uses the following to compress length tables in its headers:
    - RLE with 3 extra symbols for different runlength scales
    - secondary huffman coding for RLE symbols
    - optimized reordering for RLE huffman-code-length table
    - type1 blocks with precomputed huffman tables

    However I think that you can just as well use a good statistical
    model and arithmetic coding for the headers.
    If its only headers, it won't affect overall speed anyway.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Harry Potter@21:1/5 to All on Mon Mar 25 16:04:15 2019
    I am already using a form of RLE: a repeated value costs only one bit.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From gautier_niouzes@hotmail.com@21:1/5 to All on Mon Apr 1 08:21:53 2019
    - type1 blocks with precomputed huffman tables

    Side note: it's surprising that neither PKWare nor WinZip seemed to consider to define the type 3 block (so far, unused) for further precomputed tables in some "Deflate++" format. For instance if block type is 3, the next byte would be a choice in 256
    different huffman tables.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)