Skip to content

Add pygame.transform.bloom function #2872

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 32 commits into
base: main
Choose a base branch
from

Conversation

damusss
Copy link
Member

@damusss damusss commented May 24, 2024

I implemented the bloom algorithm to add it to pygame.transform.
I also added documentation and stubs.

I think this is the best implementation. This is the steps, if someone needs them to review:

  • Create a bright pass filter surface
  • Iterate the source surface and keep only the pixels with the luminance within the threshold
  • Run either the box or blur algorithm
  • On the lines where the blurred color is applied, also blit it to the original surface with additive blend

Please tell me if there are any C improvements, structure improvements or performance improvements.

@damusss damusss requested a review from a team as a code owner May 24, 2024 20:41
@damusss damusss requested a review from a team as a code owner May 25, 2024 11:28
@itzpr3d4t0r itzpr3d4t0r added New API This pull request may need extra debate as it adds a new class or function to pygame transform pygame.transform labels May 26, 2024

float intensity = 1, threshold = 0.5;
int blur_radius = 5;
char blur_type = 'b';
Copy link
Member

@itzpr3d4t0r itzpr3d4t0r May 26, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

personally, I'd prefer gaussian_blur to be default since it's more precise. I believe we should be as precise as possible by default and if the user wants less precision they can change it.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

personally, I'd prefer gaussian_blur to be default since it's more precise. I believe we should be as precise as possible by default and if the user wants less precision they can change it.

I set box as default because gassian blur is 5 or 6 times slower, but I guess the default should be the best looking. I think on a real usage, you would use box_blur for realtime effects (on not-too-big surfaces) while gaussian_blur for cached results. the gaussian bloom of realtime big surfaces is out of question, but the same logic applies to gaussian_blur without bloom

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And on this note you could add a note about all this in the docs.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And on this note you could add a note about all this in the docs.

Wonderful idea

Comment on lines 3670 to 3683
static char *kwlist[] = {"surface",
"intensity",
"luminance_threshold",
"blur_radius",
"blur_type",
"dest_surface",
0};

if (!PyArg_ParseTupleAndKeywords(
args, kwargs, "O!|ffisO!", kwlist, &pgSurface_Type, &src_surf_obj,
&intensity, &threshold, &blur_radius, &blur_type_str,
&pgSurface_Type, &dst_surf_obj)) {
return NULL;
}
Copy link
Member

@itzpr3d4t0r itzpr3d4t0r May 26, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The only mandatory argument is the surface to bloom, I think either just blur_radius or/and intensity should be mandatory as well.
And I think the argumets "order" should be changed a bit and ordered by importance, in case one doesn't use kwargs.
I'd do something like this:
"surface" -> "blur_radius" (mandatory) -> "intensity" (mandatory) -> "luminance_threshold" (optional) -> "blur_type" (optional default="gaussian") -> "dest_surface" (optional).

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The reordering isn't a strong requirement.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah I like this 👍
@itzpr3d4t0r I'll fix this 2 comments once I know if I should keep the buffer or keep the macros (I left a message on the discord server)

@robertpfeiffer
Copy link
Contributor

I see no reason not to merge this, but I think this might be more flexible as three separate operations: Threshold, blur, and blit. At least, I think this could be more useful as a function that computes the bloom overlay with a black or transparent background.

@damusss
Copy link
Member Author

damusss commented May 27, 2024

I see no reason not to merge this, but I think this might be more flexible as three separate operations: Threshold, blur, and blit. At least, I think this could be more useful as a function that computes the bloom overlay with a black or transparent background.

Maybe it would be more flexible but not more performant as the blit is implemented within the blur using only 2 loops. I'd like for @itzpr3d4t0r to have a word about what you mentioned. I can always modify it as requested.

@damusss
Copy link
Member Author

damusss commented May 27, 2024

To add to what I just said, what i think would be best is to keep bloom as a function and maybe add the luminance filter as another function that can reuse the same code this pull request is using.

Co-authored-by: Dan Lawrence <[email protected]>
Comment on lines 287 to 288
Returns a surface where the bright pixels are blurred and added to the original
surface resulting in a bloom effect.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

After this section I think some details on the expected parameter ranges would be useful, particularly on the valid values for blur_type, which I believe are gaussian and box. Without some guidance it is hard to know if values like 'intensity` should normally be in the range of 0-1000, or 0.0 to1.0.

Copy link
Member

@MyreMylar MyreMylar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I gave this the full walt test:

image

with this test program:

import pygame


display_surf = pygame.display.set_mode((1280, 600))
display_surf.fill((255, 255, 255))

walt = pygame.image.load("images/walt.png").convert_alpha()

walt_2 = pygame.transform.bloom(
    walt, blur_radius=2, intensity=1.5, luminance_threshold=0.6
)
walt_3 = pygame.transform.bloom(
    walt, blur_radius=15, intensity=4, luminance_threshold=0.4, blur_type="box"
)

pygame.init()

clock = pygame.time.Clock()

running = True

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

    display_surf.fill((255, 255, 255))
    display_surf.blit(walt, (0, 100))
    display_surf.blit(walt_2, (417, 100))
    display_surf.blit(walt_3, (834, 100))

    pygame.display.flip()

    clock.tick(60)

I think it is a useful effect that developers will use. They'll probably use it, and all of the blurs more if we can eventually get them SIMD accelerated in some manner, but I have no problem adding this among them. Especially as it is reusing a lot of the code from the blurs.

My main criticism is that the input parameters are quite loosely defined right now in the documentation with no expected ranges, and not much description of what they effect in the output.

@damusss
Copy link
Member Author

damusss commented Oct 5, 2024

My main criticism is that the input parameters are quite loosely defined right now in the documentation with no expected ranges, and not much description of what they effect in the output.

you have a point. I'll improve the documentation and explain what each argument does

Copy link
Member

@MyreMylar MyreMylar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice docs, LGTM 👍

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
New API This pull request may need extra debate as it adds a new class or function to pygame transform pygame.transform
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants