copy() color problem

Core Library Questions

mwhyro

in Core Library Questions • 3 months ago

Hello, I've noticed what seems to be a bug using copy on any surface or PImage. It seems that when using copy() I'll always loose one R, G and B value.

In other words, 255 will be 254 in the copy, 254 will become 253 and so on.

Here's a sketch to illustrate the behaviour:

void setup() {
size(300, 300, P2D);
background(0);
//setup the color white
color white = color(255,255,255);
//draw an ellipse with that white color
noStroke();
fill(white);
ellipse(150,150,50,50);
//sample the white color drawn
color white_sample_1 = get(150,150);
println("White sample #1 = " + red(white_sample_1)); //outputs: 255.0
//now copy the entire image to the same area again
copy(0,0,300,300,0,0,300,300);
//sample the white color
color white_sample_2 = get(150,150);
println("White sample #2 = " + red(white_sample_2)); //outputs: 254.0
//now copy the entire image to the same area again
copy(0,0,300,300,0,0,300,300);
//sample the white color
color white_sample_3 = get(150,150);
println("White sample #3 = " + red(white_sample_3)); //outputs: 253.0
}

This is causing me a lot of headache and problem using processing to process changes in images and sampling by colors exactly. Is this somehow logical and intended, please give me a quick run-down on why that is!

I'm using Processing 2.0

Thanks for any help on this!

Replies(5)

GoToLoop

Re: copy() color problem

3 months ago

My experiment on that:

//size(300, 300, JAVA2D); // works flawlessly
size(300, 300, P2D);

noLoop();
smooth();
noStroke();
background(0);
fill(-1);

final int cx = width >> 1, cy = height >> 1;

ellipse(cx, cy, 50, 50);
final PImage screen = get();
println( hex(get(cx, cy), 2) );  // FF

copy(0, 0, width, height, 0, 0, width, height);
println( hex(get(cx, cy), 2) );  // FE

copy(0, 0, width, height, 0, 0, width, height);
println( hex(get(cx, cy), 2) );  // FD

copy(0, 0, width, height, 0, 0, width, height);
println( hex(get(cx, cy), 2) );  // FC

background(screen);
println( hex(get(cx, cy), 2) );  // FF

When using default JAVA2D, everything works!

But in P2D & P3D, copy() keeps making color white more "opaque" the more it's used!

However, saving an original screen copy w/ get(), there's no bug as copy()!

Very similar behavior happens in Processing 1.5.1 as well.

Only JAVA2D (maybe old OPENGL too) has expected behavior!

mwhyro

Re: copy() color problem

3 months ago

Thanks for the tip GoToLoop!

I tried it with JAVA2D and it worked like you said. I'm a bit worried as to how this will impact performance.

This seems like a bug in the P2D,P3D and OPENGL render modes

GoToLoop

Re: copy() color problem

3 months ago

Each rendering engine has its advantages and disadvantages when dealing w/ certain types of drawing.

They differ greatly between Processing v1.5.1 & 2+.

I believe you noticed that method get() can be an alternative to copy() in many situations.

And works on all renders!

mwhyro

Re: copy() color problem

3 months ago

Yeah it's just a matter of me using scaled copies of the image to do blob detection and finding motion. But I'll see if I can't find an alternate way to do that without using copy() if JAVA2D is slower.

Thanks for your help!

datguy

Re: copy() color problem

3 months ago

I tried a slight variation that had a very interesting effect:

void setup() {
// size(300, 300, JAVA2D); // works flawlessly
size(300, 300, P2D);
background(0);
ellipse(50, 50, 50, 50);
}
void draw() {
copy(0, 0, width, height, 0, 0, width, height);
}

It creates a circle that slowly moves, grows, and fades away.

Top Reply