Architect internals to support more algorithms in a general way

[mhseiden]

Currently, the internals assume that the pixel sorting happens row-by-row, where each line has no notion of the other lines. This does create interesting results, however there are certainly other algorithms that could be added, which need a different processing framework. For example, grid based transforms, column based transforms, 2-D convolutions, etc.

The action item here, is to write more generic functions for the other types of transforms.