spillerrec / animeraster Goto Github PK

The mirrored filters of paeth and strange might not be correct. Try them on mirrored images, and see if they produce the same result as the non-mirrored counterparts.

To accurately determining the optimal configuration, we want to test different configurations on a large sample of files. We need to programmatically select configurations and write the results to a .csv file.

To more easily support all the ways we can transform the saved data and perform analysis, we want to create a common format and pass that structure around. This will be very helpful when we want to test different configurations and compare them, as we don't have to add the test code for all the compression types.

Converting RGB to YUV like JPEG-LS might be useful. Method explained somewhat here:
http://en.wikibooks.org/wiki/Data_Compression/Multiple_transformations#color_transforms

This is a 26-bit transform, we might want to try out this 24 bit transform:
http://stackoverflow.com/a/12146329/2248153
Might not be as effective, but it is easier to fit into the current scheme to test if it is worth it.