multiply filter for PImage?

Programming Questions

ericsoco

in Programming Questions • 1 years ago

tint() appears to be additive -- regardless of colorMode(), the parameters to tint() are always added to each pixel.

is there a multiplicative version of tint()? i ask because i'd like to desaturate an image. i can make a copy of the PImage, desaturate it completely with filter(GRAY), and then blend() the two. but it seems like that process could be simpler.

e.g. as3's ColorTransform class has both a multiplier and an 'offset' for each channel; the offsets correspond to the params to tint().

http://help.adobe.com/en_US/FlashPlatform/reference/actionscript/3/flash/geom/ColorTransform.html

but as far as i can tell there's no equivalent in processing for the multipliers.

maybe there's an obvious answer, but i'm not finding it...help?

Replies(7)

PhiLho

Re: multiply filter for PImage?

1 years ago

" it seems like that process could be simpler" " as3's ColorTransform class has ..."
Processing isn't Flash, and lot of things are done at a lower level than in AS3.
Why do you care if the process could be simpler? You code it once, then the computer runs it...

The filters in Processing are a bit primitive (in both sense of the term), and coders often have to combine them, or even act directly on the pixel array to get the effects they want.
Now, perhaps somebody know them better than me and will give you a better answer.

ericsoco

Re: multiply filter for PImage?

1 years ago

of course processing's not flash...i wasn't trying to suggest that processing emulate flash, just to mention that an API for multiplicative color manipulation is useful in flash. and maybe that would be a good thing for processing.

i'll try digging into the filter implementations to see what i can find, and perhaps post a feature request. just asking first to see if i'm missing anything obvious...

davbol

Re: multiply filter for PImage?

1 years ago

>> is there a multiplicative version of tint()?

yes - it's called tint()!

iow, your premise and question appear to be worded backwards. tint() multiplies, and there is no equivalent to tint() that is additive.

for example, tint(128) will make your image half as bright as it originally was, because it's being multiplied by 128/255th's. tint(128,255,255) will make just the red channel half as bright, etc.

or another way of looking at it is: you can "tint" white darker, but you can't "tint" black lighter - why? multiplying by a n/255 fraction (n on [0,255]) will never *increase* the value. (so a nitpicker might argue that tint() might have been better named shade(), oh well, color theory notwithstanding)

agree that it's too bad that an additive offset isn't also available, as bias+scale is a very useful rgb operation, equivalent (in non-strict terms) to a simultaneous contrast/brightness adjustment. but it would be a per-image operation, as tint() is, and there are "fancier" methods available in processing...

it's easy to enough to use blend(...,ADD) and blend(...,MULTIPLY) with suitable overlay images to achieve the desired effect, and then some! because now you're in charge of the effect, per pixel!

for instance, try blending (try with both ADD and MULTIPLY) with a white-to-black gradient. that type of effect (and many others) cannot be achieved with either tint() or flash-style bias+scale method, even if it existed in processing, because those both operate per image, not per pixel. something like this:

size(500,500);
PImage hgrad = loadImage("hgrad500.png");
PImage vgrad = loadImage("vgrad500.png");
PImage sourc = loadImage("sourc500.png");
sourc.blend(vgrad,0,0,500,500,0,0,500,500,MULTIPLY);
sourc.blend(hgrad,0,0,500,500,0,0,500,500,ADD);
image(sourc,0,0);

ericsoco

Re: multiply filter for PImage?

1 years ago

@davbol -- ha! you got me. you're right, tint() *does* multiply color channels. and, thanks for the suggestions for other manipulations to try.

i think my question stems from a misunderstanding of how HSB colorMode() works. i was attempting to desaturate an image with just tint(), by changing the saturation parameter. however, it appears that changing the saturation channel is essentially applying a multiplier to the image's pixel values based on the hue parameter; with saturation set to 255 (i.e. tint(N, 255, N);), the image stays the same. lowering the saturation uses multiplication to reduce color channels to isolate just the hue specified by the hue param.

blah blah. sorry for the misinformed question.

for future reference, simplest way to partially desaturate an image is to blend it into a filter(GRAY)-ed version of itself:

PImage img;
void setup() {
size(200, 200);
img = loadImage("jelly.jpg");
}
void draw() {
image(img, 0, 0);
filter(GRAY);
tint(255, 255 * mouseX/width);
image(img, 0, 0);
}

davbol

Re: multiply filter for PImage?

1 years ago

>> misunderstanding of how HSB colorMode() works

>> appears that changing the saturation channel

Looks like you've found a solution, and this thread is getting old, but if still following and if it's of any help:...

tint() operates directly and exclusively on RGB values: there is no saturation (or hue, or brightness) "channel" in processing. You can use colorMode(HSB) to specify colors, but they're always stored as RGB. So, tint() stores the tinting color internally as RGB, and operates on an array of stored RGB pixels.

You could (with some minimal work) write your own routine to process pixels in HSB-space: loop through all the pixels one at a time, convert to HSB, adjust as desired, reconvert back to RGB and stuff back in pixel array. Except for the colorspace conversion back and forth, tint() causes exactly that kind of per-pixel read-modify-write behind the scenes.

hth

ericsoco

Re: multiply filter for PImage?

1 years ago

right, i understand that, should have said saturation *parameter* instead of *channel*.

@davbol, can you point me to the part of the code in processing's source that applies the tint changes? i've had some trouble finding it. i assumed it was in PGraphics, but i don't see any pixel operations...maybe because the tint implementation is abstracted out to PGraphics subclasses. just curious to know exactly how these operations are done.

ericsoco

Re: multiply filter for PImage?

1 years ago

oh actually, answering my own question -- just found it in PGraphicsJava2D.update(). i'm guessing this kind of thing is offloaded to the GPU in GL, but PGraphicsJava2D helps me understand what's going on.

Top Reply