Color Matching Algorithm Needed

[Chuck Bruce]

Does anyone have (or know of) a good algorithm for finding the closest color available in a palette? I am using a 256-color palette, with 8 bits each for the r,g, and b components. Allegro uses a weighted squares method, which works OK, but it does not seem to be as accurate as the CPalette::GetNearestColorIndex() function in VC++. I am writing an app for DOS, so I can't use the VC++ function. Any help would really be appreciated.



Jesus is Lord!

[Wade]

Hello, These two functions together work really well. We have a math wiz on our team who gave me this to use once, and it is great.


double CYourClass::ComputeColorDistance(double r1, double g1, double b1, double r2, double g2, double b2)
{
/*double rd = 77*(r1-r2)/255;
double gd = 150*(g1-g2)/255;
double bd = 29*(b1-b2)/255;
double d = sqrt(rd*rd+gd*gd+bd*bd);

return d;*/

double x1 = .412453*r1/255+.357580*g1/255+.180423*b1/255;
double y1 = .212671*r1/255+.715160*g1/255+.072169*b1/255;
double z1 = .019334*r1/255+.119193*g1/255+.950227*b1/255;
double x2 = .412453*r2/255+.357580*g2/255+.180423*b2/255;
double y2 = .212671*r2/255+.715160*g2/255+.072169*b2/255;
double z2 = .019334*r2/255+.119193*g2/255+.950227*b2/255;

double cl1 = 116*pow(y1,1.0/3)-16;
if (y1 <= .008856) cl1 = 903.3*y1;
double ca1 = 500*(Fcn(x1)-Fcn(y1));
double cb1 = 200*(Fcn(y1)-Fcn(z1));
double cl2 = 116*pow(y2,1.0/3)-16;
if (y2 <= .008856) cl2 = 903.3*y2;
double ca2 = 500*(Fcn(x2)-Fcn(y2));
double cb2 = 200*(Fcn(y2)-Fcn(z2));

double ld = (cl1-cl2);
double ad = (ca1-ca2);
double bd = (cb1-cb2);

double dist = ld*ld+ad*ad+bd*bd;
dist = dist / 10100; //4 .05

return dist;
}

double CYourClass::Fcn(double x)
{
if (x > .008856)
return pow(x,1.0/3);
else
return 7.787*x+16.0/116;
}





Good luck!

Wade

[Chuck Bruce]

Thanks, that is a lot better than the standard weighted squares method (which I noticed you had commented out in your code). This is good enough for me to use, but STILL not quite as good as the CPalette::GetNearestPaletteIndex() function.

Thanks for your help.


Jesus is Lord!

[mili123]

Hello, These two functions together work really well. We have a math wiz on our team who gave me this to use once, and it is great.


double CYourClass::ComputeColorDistance(double r1, double g1, double b1, double r2, double g2, double b2)
{
/*double rd = 77*(r1-r2)/255;
double gd = 150*(g1-g2)/255;
double bd = 29*(b1-b2)/255;
double d = sqrt(rd*rd+gd*gd+bd*bd);

return d;*/

double x1 = .412453*r1/255+.357580*g1/255+.180423*b1/255;
double y1 = .212671*r1/255+.715160*g1/255+.072169*b1/255;
double z1 = .019334*r1/255+.119193*g1/255+.950227*b1/255;
double x2 = .412453*r2/255+.357580*g2/255+.180423*b2/255;
double y2 = .212671*r2/255+.715160*g2/255+.072169*b2/255;
double z2 = .019334*r2/255+.119193*g2/255+.950227*b2/255;

double cl1 = 116*pow(y1,1.0/3)-16;
if (y1 <= .008856) cl1 = 903.3*y1;
double ca1 = 500*(Fcn(x1)-Fcn(y1));
double cb1 = 200*(Fcn(y1)-Fcn(z1));
double cl2 = 116*pow(y2,1.0/3)-16;
if (y2 <= .008856) cl2 = 903.3*y2;
double ca2 = 500*(Fcn(x2)-Fcn(y2));
double cb2 = 200*(Fcn(y2)-Fcn(z2));

double ld = (cl1-cl2);
double ad = (ca1-ca2);
double bd = (cb1-cb2);

double dist = ld*ld+ad*ad+bd*bd;
dist = dist / 10100; //4 .05

return dist;
}

double CYourClass::Fcn(double x)
{
if (x > .008856)
return pow(x,1.0/3);
else
return 7.787*x+16.0/116;
}


Good luck!

Wade

Hello Wade,

you computed color difference in this function .
Well this is actually color matching or similarity . but my question is that what the difference mean actually regarding matching .
I want to show color similarity image .do't want to compute the color differences.

look : http://i1096.photobucket.com/albums/...similarity.jpg
This is color difference formula that's what you did in your function and this is actually color matching.
Just tell what this difference mean ?
Also how can i show color similarity image ?

[Skizmo]

This thread is 10 years old.

[Paul McKenzie]

Hello Wade,

How do you know that Wade is even still alive after 10 years?

Regards,

Paul McKenzie