pygame.transform.pixelate

[oddbookworm]

Addressing #2302

TODO:

• tests

[andrewhong04]

Works great.

Before & After

Code

import pygame

screen = pygame.display.set_mode((500, 500))
creeper = pygame.image.load("creeper.jpg")
creeper = pygame.transform.scale_by(creeper, 0.5)
creeper = pygame.transform.pixelate(creeper, 50)

running = True
while running:
    for e in pygame.event.get():
        if e.type == pygame.QUIT:
            running = False

    screen.fill((0, 0, 0))
    screen.blit(creeper, (0, 0))
    pygame.display.update()

original screenshot

[Starbuck5]

Could this effect be replicated with a smoothscale down and a scale up? It could even do that in C code, and it might be faster because those are SIMD optimized?

Initial testing is promising that it replicates the effect, I didn't test performance.

[MyreMylar]

Could this effect be replicated with a smoothscale down and a scale up? It could even do that in C code, and it might be faster because those are SIMD optimized?

Initial testing is promising that it replicates the effect, I didn't test performance.

Shouldn't even need smoothscale for this, a bit of math with pixel_size versus the smallest dimension of the original image then two calls to SDL_SoftStretch should probably do it I'd have thought? Unless smoothscale is now faster than SDL_SoftStretch after recent SIMD-ifying?

Would probably be worth a look at least to see if it is faster than this current method.

[Starbuck5]

I've converted this to a draft due to large open questions about how the implementation should work.

[oddbookworm]

I'll revive this soon, I promise

[MyreMylar]

OK, all looks great to me now. I gave it the full walt.png test:

And it performed as expected. Tests and docs make sense and I like the reuse of existing scaling functions. All in all this is neat! 👍

[Matiiss]

How about we implement this by way of generalizing the blur transforms by adding a stride argument to them? Then we could implement this as a blur with stride equal to the blur radius. Then we'd also have the stride argument for blurs and pixelate could be like a convenience function that simply calls one of the blur functions with stride equal to radius, though I suppose it'd be called pixel size or something.