next »

[thekill473]

Are there any optimization for rotating a group of pixels by an arbitrary angle. I don't want to use standard image rotation algorithms because the number of pixels in the group has to stay consistent.

Thanks.

[surt]

If you are rotating a grid (in non quarter revolution increments) and returning the result to the grid you are always going to have overlaps, which result in a reduction of the number of pixels, and gaps, which when filled with interpolation result in a increase in the number of pixels.

[Gimym JIMBERT]

What's the goal? good image or precise rotation? looks like @surt expose the problems quite well.

Else there is a algo, that upscale in some hqx filter (use in emulation), rotate, and downscale to have very good visual rotation.

[SF]

Have you considered running the game at a modern resolution, but with scaled up low-res sprites? That would allow you to do fine rotations without much hassle, though you'd lose some visual authenticity if you were going for a retro look.

[powly]

What's the situation - why do you need a consistent pixel count? You could probably just rotate every pixel, see where they land and adjust them to adjacent pixels if you get overlaps. If your groups are sparse enough or some oddities here and there don't matter, it could work very well.

[thekill473]

I wanted to attempt to make a game with powder toy(http://powdertoy.co.uk/) like physics, maybe with a lower resolution. But the pixels could group together to form a bigger object like a clump of dirt. And to make it anywhere near realistic the object would have to be able to rotate and be able to split and merge with other objects of the same material.

[nikki]

nice toy,

perhabs you keep a reference to every particle, then you occasionaly don't see a pixel because it's rotated and aliased away but at a slightly different angle it reappears.

and then see if that's good enough