speedo

Is there a way to remove all colors from a matrix except orange? I've tried chromakeying and went through the color balancing tutorials, but there isn't anything putting me on the right track.

I just want the matrix to be all black except for any orange parts.

removing-all-colors-except-orange

really nice solution. I like both of them - elegant and clean. I was completely on the wrong path trying to accomplish this. Thanks.

What exactly do you mean by "You could also imagne a non-binary filter where the multiplier would increase with the nearness of the orange hue." What is an example of a non-binary filter?

really nice solution. I like both of them - elegant and clean. I was completely on the wrong path trying to accomplish this. Thanks. 

What exactly do you mean by "You could also imagne a non-binary filter where the multiplier would increase with the nearness of the orange hue." What is an example of a non-binary filter?


I'm interested as well in a way to feather the edges of the color selection "where the multiplier would increase with the nearness of the orange hue." Still at square 1: trying to plug variables into the expr code:

In the previous examples, pixels do either have a color close to orange and then they go untouched or their color is not close to orange and then they are set to black (so it's binary).

But you could imagine a patcher where there is a continuum of possibilities between a black pixel and an untouched pixel if you multiply the luminosity of each pixel according to the closeness of its hue to orange. The difficult part is to define "closeness" in maths.

any reason you don't want to use jit.chromakey?

No, I didn't know about it, but I am also an expr-addict...

This is great Roald, lots of food for thought! How would one go about sending values dynamically to the jit.expr in your original patch? i.e replace the 2. and 3. multipliers with variable inputs.

"Here I chose a range where red is worth 2 x green but smaller than 3 x green (to avoid red pixels) and where blue is lower or than 10%."

Here is a slighly modified version you can adapt to any color.

To replace a value with an input, change the number of inputs with @inputs and use in[x] as a replacement for each variable with x being the input number.

The trick is that each variable must be transmitted to jit.expr as a matrix (see how it's done with the exponent).

Out of curiosity I tried to alter the first patch you posted-- the binary non-hsl version -- to allow variables in the jit.expr message

(2. * in[0].p[2]) && (in[0].p[1] < (3. * in[0].p[2])) && (in[0].p[3] < 0.1 )) * in[0]">

by using jit.expr @inputs 3 and the message

(in[1] * in[0].p[2]) && (in[0].p[1] < (in[2] * in[0].p[2])) && (in[0].p[3] < 0.1 )) * in[0]">

and sending values via setcell messages through

it seems like it only accepts values between 0. and 1.

great stuff!
Out of curiosity I tried to alter the first patch you posted-- the binary non-hsl version -- to allow variables in the jit.expr message

 (2. * in[0].p[2]) && (in[0].p[1] < (3. * in[0].p[2])) && (in[0].p[3] < 0.1 )) * in[0]">

 (in[1] * in[0].p[2]) && (in[0].p[1] < (in[2] * in[0].p[2])) && (in[0].p[3] < 0.1 )) * in[0]">

and sending values via setcell messages through 

it seems like it only accepts values between 0. and 1.
i.e. no change between 1. and 2. 


Removing all colors except orange?