Results 1 to 7 of 7

Thread: A lil bitmap test...

  1. #1
    Member
    Join Date
    May 2007
    Location
    Poland
    Posts
    85
    Thanks
    8
    Thanked 3 Times in 3 Posts
    OK, heres a little lossless bitmap compression test i've made.
    It's about the best or most popular (IMO) programs for compressing images losslessy. Today as HDD space is plenty so lossy compression is sometimes superfluous. Some of us have access to unaltered digital content (e.g. RAW digtal camera files) which needs to be archived and stored as original.

    Programs used:
    7-zip v4.45 beta - I decided to be simple and just used default settings
    PNG - for this I chose Irfanview 4, compression level 9 (max)
    Optimized PNG -automated optimizing script (credit goes to JensRex from HA)
    JLS (loco) - HP free implementation, default settings
    CoBALP- I used the version provided by Stephan Bush, all default with and without color conversion

    Images - I tried to pick various images with different complexity and dimensions and origin.

    The test:




    comments will come later

  2. #2
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,954
    Thanks
    359
    Thanked 332 Times in 131 Posts
    And how about PIM?

  3. #3
    Member
    Join Date
    Jan 2007
    Location
    Moscow
    Posts
    239
    Thanks
    0
    Thanked 3 Times in 1 Post
    If you choose CoBALP i don't see any reason why BMF, MRP, SBC and UHIC are not tested... Two first easely beat CoBALP in my tests. Yes, MRP is slow like hell, but anyway... Besides, i'd like to see bit depth and actual numbers of colours in your image set. And times taken for compression and decompression if it is possible. Thanks for nice test.

  4. #4
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    Nice Test!

    I do agree with Ilia though! I would love to see PIM included in those test results.

  5. #5
    Member
    Join Date
    May 2007
    Location
    Poland
    Posts
    85
    Thanks
    8
    Thanked 3 Times in 3 Posts
    Thanks for comments. I'll answer them a little later (need to check something first...)

    Now, a few conclusions that can be made from this small test:

    1. For simple images like drawings, screenshots it is better to use LZ77 method (7z LZMA, PNG). Large areas of identical color can be exploited by such algorithms and in extreme situations images can be compressed to just a few % of their original size - look at 'ars' , 'winrar' and especially 'explorer' (less than 1,5% of original size!).

    2. For natural images like photos things are different:
    Long patterns for achieving significant compression are hard to find. Other aproaches (LOCO-I, CoBALP) work much better for them. Complexity of natural images prevents LZ77 algorithms to perform well.

    3. Fractal-like images are somwhat in between and both aproaches seem to achieve similar results.

    and
    4. CoBALB color conversion improves compression immensely and should be used. The exception in the test was image named 'hell' for with conversion actually gave bad results. It could be that this image is very special (it is mostly red) and this anomaly won't repeat.

  6. #6
    Member
    Join Date
    May 2008
    Location
    France
    Posts
    78
    Thanks
    436
    Thanked 22 Times in 17 Posts
    .... your screenshot of results is not size-optimized

  7. #7
    Member
    Join Date
    May 2007
    Location
    Poland
    Posts
    85
    Thanks
    8
    Thanked 3 Times in 3 Posts
    @encode
    Sorry I didn't test PIM. Now I see that it has a new bitmap filter . Notice however that I even did not include WinRar, nor 7z PPMD method and of course myriads of other compression programs that are available (including my favorite CCM) that would be nice to test too. I'd like to add more but I'm too lazy .


    @nimdamsk
    Yes, you are of course right about including CoBALP and not others. Actually I totally forgot about BMF and I remember it always gave marvelous results. You say it should beat CoBALP. It is certainly possible! Now that you mentioned it, I might try it for curiosity. My incentive to do this was to test CoBALB against some of the programs I myself use from time to time for compressing images losslessy and some established algorithms (loco-jls). It turned out after the testing that CoBALP won and by far, both in total and weighted comparison. Really, I did not expect this! This is amazing algorithm, especially for natural images (of course color conversion helped very much but its not CoBALP fault that the other compression programs did not exploit this possibility).

    About color depth - Sure, I'll add it. Also a correction to my test in this regard needs to be made...

    As for compression times - no sorry .


    @Mike
    Believe or not I did recompress this PNG with the exact script used the test Originally it was like 15kb or something. I like this script very much, it is convenient and reasonably fast . It also has an insane mode which uses PNGOUT but is slooow as paint peeling (thus I never use it). How much bytes did you still manage to squeeze?

Similar Threads

  1. Test set: bitmap
    By m^2 in forum Data Compression
    Replies: 28
    Last Post: 13th January 2009, 17:44

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •