1. RGB轉換為Lab的原理[1]
RGB=>Lab
X = [0.4339100.376220 0.189860] * [R/255]
Y = [0.2126490.715169 0.072182] * [G/255]
Z = [0.0177560.109478 0.872915] * [B/255]
L的計算:
如果 Y>0.008856
L = 116*Y*(1/3) ;
否則
L = 903.3*Y.
a, b的計算:
a =500*(f(X)-f(Y))
b =200*(f(Y)-f(Z))
其中,
如果 t>0.008856
f(t)=t*(1/3);
否則
f(t)=7.787*t+16/116.
2. Matlab實現的代碼(摘自[2])
function [L,a,b] =RGB2Lab(R,G,B)
% function [L, a,b] = RGB2Lab(R, G, B)
% RGB2Lab takesmatrices corresponding to Red, Green, and Blue, and
% transforms theminto CIELab. This transform is based onITU-R
%Recommendation BT.709 using the D65white point reference.
% The error intransforming RGB -> Lab -> RGB is approximately
% 10^-5. RGB values can be either between 0 and 1 orbetween 0 and 255.
% By Mark Ruzonfrom C code by Yossi Rubner, 23 September 1997.
% Updated forMATLAB 5 28 January 1998.
if (nargin == 1)
B = double(R(:,:,3));
G = double(R(:,:,2));
R = double(R(:,:,1));
end
if ((max(max(R))> 1.0) | (max(max(G)) > 1.0) | (max(max(B)) > 1.0))
R = R/255;
G = G/255;
B = B/255;
end
[M, N] = size(R);
s = M*N;
% Set a threshold
T = 0.008856;
RGB =[reshape(R,1,s); reshape(G,1,s); reshape(B,1,s)];
% RGB to XYZ
MAT = [0.4124530.357580 0.180423;
0.212671 0.715160 0.072169;
0.019334 0.119193 0.950227];
XYZ = MAT * RGB;
X = XYZ(1,:) /0.950456;
Y = XYZ(2,:);
Z = XYZ(3,:) /1.088754;
XT = X > T;
YT = Y > T;
ZT = Z > T;
fX = XT .*X.^(1/3) + (~XT) .* (7.787 .* X + 16/116);
% Compute L
Y3 = Y.^(1/3);
fY = YT .* Y3 +(~YT) .* (7.787 .* Y + 16/116);
L = YT .* (116 * Y3 - 16.0) + (~YT) .* (903.3 *Y);
fZ = ZT .*Z.^(1/3) + (~ZT) .* (7.787 .* Z + 16/116);
% Compute a and b
a = 500 * (fX -fY);
b = 200 * (fY -fZ);
L = reshape(L, M,N);
a = reshape(a, M,N);
b = reshape(b, M,N);
if ((nargout == 1)| (nargout == 0))
L = cat(3,L,a,b);
end
註:這裡L,a,b的計算有點巧妙,利用bool型變數將條件陳述式綜合在一條式子裡。
參考資料:
[1] OpenCV手冊.chm 中的“CvtColor”部分.
[2] http://www.mathworks.com/matlabcentral/fileexchange/22274-hill-climbing-color-image-segmentation/content/RGB2Lab.m