third-order polynomial regression for the RGB components

please can someone help me to find "third-order polynomial regression for the RGB components"

Answer 1

[John D'Errico]

It is not obvious what you want, but my guess is you are doing some sort of calibration or mapping between color spaces. Perhaps mapping from one RGB space to another, or from RGB to another space, like L*a*b* or XYZ.

My polyfitn tool does that for you, although it only allows you to fit one model at a time. These mappings are actually three such models, done independently. Thus to map from RGB_in to RGB_out, you would build one model for each output variable. Thus effectively...

R_out = fr(RGB_in)
G_out = fg(RGB_in)
B_out = fb(RGB_in)

Having said that, polynomial models are frequently used. This does not say they are the best solution, only the most easily built. Sadly, sufficiently often you will find them inadequate for the purpose. In which case direct mappings to build lookup tables directly are the best way. Effectively, this is a low order spline model, usually piecewise linear.

Answer 2

[Image Analyst]

imresize() does that - bicubic interpolation. Is that an example that will work for you?

[Image Analyst]

I also have a background correction that does it. It uses John D'Errico's polyfitn with an order of 4 but you can change it to 3 if you want:

polynomialOrder = 3;

Look for the line above in the code.