Michael Pound

Haha - well it was worth a go :) This looks great though, I think massively pushing the gamma is going to do strange things unless we can ultimately implement...

I may have to have a go at writing a plugin sometime :) I think if we're careful the version with only one integral image shouldn't be much slower, this...

That paper looks like a cool way to produce the initial summed area table, which is a large proportion of the processing required to perform a filter. After that it's...

I agree, this is absolutely a convolution, just with only weights of zero or one which makes summation much faster. Actually there is a suggestion for aliasing on the original...

Ooh, subtracting with weights actually is really neat, saves messing about removing these pixels from the rectangles. For an aliased pixel of weight t you subtract val*(1-t) from the overall...

I've been messing around with a function to return some rectangles from the RLE of a bitmap. It's ok for a first attempt I think, works with a variety of...

My lack of knowledge of Direct2D is showing :) I agree that I don't think you need access to the full table as long as you can visit neighbouring tiles,...

@georgedorn As per a discussion I was having with @bertie2, I've created a new repo which will host the community changes to this code, to keep this version in line...

Very nice :) I'm fairly happy with these kinds of optimisations, with only a few caveats right now: 1) I clearly didn't think through what would happen when I post...

This is a great idea! I may do this tomorrow if I have time :)