*Small bug fixed (decidedly non-power-of-2 images sometimes had important wavelet-coefficients zeroed out). Thanks, Christian Lange!*
Wavelet code updated to version 2.5.2 (including source). Improved the quantiser-selection for bit-limits (i.e. file-size) a LOT (was a bit stupid before – I only optimised for the smallest difference in maximum and minimum mean square error) -> much better results for same size, changed compile-settings slightly (should be faster) and added a bitplane / quantiser display to the GUI. Get it here!
Category Archives: wavelet
Improved the wavelet-code a bit
Added support for loading texture- and cube-maps and changed the GUI so that it can be used to view files from the command-line.
Better, smaller, faster
Fixed the bug and updated the wavelet code to 2.5. Includes much faster compression, smaller memory footprint and a web-browser plugin w/ progressive downloading. Check the code-page for source and details. Also take a look at the demonstration-page to test the plugin and / or applications.
Warning
there’s a bug in my wavelet-decompression browser plugin (using multiple threads). i’ll release version 2.5 as soon as i fix that (with the source ;))…
Wavelet code now up to version 2.5
Will be up on the page soon (i.e. tonight or probably tomorrow). Sneak preview available here. The browser-plugin still has some multi-threading issues I need to fix (although IMO it should work… I ain’t allocating critical sections for nuthing! ;)).
Integer Math in CAS
got this week off (all of 4head does in fact), so i started to do some more research on my wavelet-code. spent two days in mathematica trying to analyse error-propagation in the wavelet-transform, proved it is symmetric (just did that as a check), and now it turns out it doesn’t work for some reason (just kinda, but it should be exact)…argh!
wavelet kompressor 2.1
wavelet.zip updated to version 2.1. it had serious bugs in the quality based quantiser-selection (one was related to not returning lossless “solutions” when they were asked for, the other very often returned large solutions, although better (i.e. smaller) ones were available).