Finding closest matching RGB value from array

[General]

Let's say I have an RGB value, "231 56 24", and have an array with many different RGB values.

I need to find a color on that array which is the most similar to "231 56 24". How would I go about doing this? Apparently Blockland uses a similar method to load bricks with similar colors when using a different colorset.

Help would be much appreciated.

[Ad Bot]

[Port]

The cheapest way is finding the value with the lowest vectorDist(color1, color2), but of course, that can yield some strange color matches once in a while.

[General]

The cheapest way is finding the value with the lowest vectorDist(color1, color2), but of course, that can yield some strange color matches once in a while.

No yeah I've already tried that method, I need a method that will never fail at all. It needs to be completely accurate.

[BluetoothBoy]

Grab the absolute values of r1 - r2, g1 - g2, and b1 - b2. Add together the absolute values. The set with the smallest added number is the closest to rgb1. (Note that I'm using r1, g1, etc only for examples sake; I know that's not how your variables work, being complete rgb values.)

[General]

Grab the absolute values of r1 - r2, g1 - g2, and b1 - b2. Add together the absolute values. The set with the smallest added number is the closest to rgb1. (Note that I'm using r1, g1, etc only for examples sake; I know that's not how your variables work, being complete rgb values.)

Yeah I did already try your method. The issue with it is that it doesn't take into consideration which values are higher in RGB.

By this i mean, let's say, you have 3 RGB values in the array. The first is "255 0 0", second is "0 255 0" and third is "0 0 255".
Let's pretend the value I want to match is "85 85 85", the script will automatically assume that either of the 3 in the array is identical, when really they are not.

[Port]

One thing you could try is converting it to HSV/HSL and then do the same vectorDist method. It'd work at least a bit better than when using RGB.

[General]

One thing you could try is converting it to HSV/HSL and then do the same vectorDist method. It'd work at least a bit better than when using RGB.

I guess so, it would be the best bet at this moment.

Does anyone know how Badspot does it though? It works wonderfully for blockland.

[MARBLE MAN]

I guess so, it would be the best bet at this moment.

Does anyone know how Badspot does it though? It works wonderfully for blockland.

build bot has a "findClosestColor"

http://forum.blockland.us/index.php?topic=219699.0
Link to thread you could have looked up

Code: (Client side) [Select]

function findclosestcolor(%x) //input four numbers between 0 and 1
{
for(%a=0; %a<64; %a++)
{
%match = mabs(getword(getcoloridtable(%a),0) - getword(%x,0)) + mabs(getword(getcoloridtable(%a),1) - getword(%x,1)) + mabs(getword(getcoloridtable(%a),2) - getword(%x,2)) + mabs(getword(getcoloridtable(%a),3) - getword(%x,3))*4;

if(%match < %bestmatch || %bestmatch $= "")
{
%bestmatch = %match;
%bestid = %a;
}
}
return %bestid; //trans may be messed up if you input a nearly-but-not-quite opaque color
}


Code: (Server Side) [Select]

getclosestpaintcolor(vectorscale(%rgb,255))


[$trinick]

You could rate them on an exponential curve from the proper color. I don't know what the results would actually look like but in theory it should be better than vectorDist.

Code: [Select]

function colorMatch(%colorList, %color)
{
for(%i = 0; %i < getFieldCount(%colorList); %i++)
{
%testColor = getField(%colorList, %i);
%rDiff = mPow(mAbs(getWord(%testColor, 0) - getWord(%color, 0)),1.2);
%gDiff = mPow(mAbs(getWord(%testColor, 1) - getWord(%color, 1)),1.2);
%bDiff = mPow(mAbs(getWord(%testColor, 2) - getWord(%color, 2)),1.2);
%totalDiff = %rDiff + %gDiff + %bDiff;
if(%totalDiff < %lowestDiff || %lowestDiff $= "")
{
%lowestDiff = %totalDiff;
%lowestColor = %testColor;
}
}
return %lowestColor;
}

The exponential factor should make it throw out colors that are too different on any combination of axes.

[Ipquarx]

I think Port's suggestion of converting to HSV and doing the vectorDist on them would return the closest matches.