CIELAB values from TIFF image

hello,

I'm trying to read CIELAB values from TIFF image in Matlab with:

A_lab= imread ('A', 'tiff');

what I get is A_lab<1x33x3 unit8> with values [0,255] which are not CIELAB values that I want (positive & negative integer and non-integer).

Have any idea?

thanks

Answer 1

[Walter Roberson]

http://www.mathworks.com/help/techdoc/ref/imread.html says,

"If a file contains CIELAB color data, imread converts it to ICCLAB before bringing it into the MATLAB workspace because 8- or 16-bit TIFF CIELAB-encoded values use a mixture of signed and unsigned data types that cannot be represented as a single MATLAB array."

See http://www.mathworks.com/help/techdoc/ref/imwrite.html for a description of ICCLAB encoding

Answer 2

[Thomas]

this page might help

http://robotics.stanford.edu/~ruzon/software/rgblab.html

Answer 3

[KAE]

Here is what I used on a 16-bit CIELab TIFF file based on the 16-bit ICCLAB documentation here. Or there are conversions within encode_color.m which comes with the Image Processing toolbox, though some coefficients are different,

aUint16 = imread(fileData);
% Warning: Converting CIELab-encoded TIFF image to ICCLab encoding. 
aDouble = double(aUint16); % Converting to double so can have decimals

Lab(:,:,1) = aDouble(:,:,1)*100/65280; % L
Lab(:,:,2) = aDouble(:,:,2)/256 - 32768; % a
Lab(:,:,3) = aDouble(:,:,3)/256 - 32768 ; % b

% Values are integers in the range [0, 65280]. L* values are multiplied by 65280/100. 
% 32768 is added to both the a* and b* values, 
% which are represented as integers in the range [0,65535].

[KAE]

And substitute these 3 lines for 8-bit CIELab TIFF,

Lab(:,:,1) = aDouble(:,:,1)*100/65280; % L
Lab(:,:,2) = aDouble(:,:,2)/256 - 128; % a
Lab(:,:,3) = aDouble(:,:,3)/256 - 128 ; % b

By the way all these constants are powers of 2, for example 65280 = 2^16 - 2^8.

[Roger Breton]

Thanks! Will experiment with this code...

Answer 4

[John]

IMREAD converts CIELAB to ICCLAB (it should also warn you that it is doing this). If you want the raw CIELAB values as stored in the TIFF, try the Tiff class instead.

[Opiuz]

thanks John, I get almost I want by this:

input_class = class(A);

cielab = double(mapped_lab3d);

if strcmp(input_class, 'uint8')

cielab(:,:,1) = cielab(:,:,1) ./ (255/100);

cielab(:,:,2) = cielab(:,:,2)-128;

cielab(:,:,3) = cielab(:,:,3)-128;

end

the thing is L* values are pefect (- and+ non(integers))

but a* and b* are only integers(also - and +)

what can I do that they are given by non-integer value?

[Opiuz]

sorry*

A=mapped_lab3d

[John]

I'm not really a colorspace expert, so I'd suggest looking at the "Adobe Photoshop TIFF Technical Notes" dated March 22, 2002, which goes into a bit of explanation as to the interpretation of the different channels as retrieved from a CIELab TIFF.

[KAE]

I had trouble with trying to do this with the tiff class, possibly because the CIELab tiff files are a mix of signed and unsigned integers, and had to use imread instead.

% The a* and b* values seemed wrong with this approach
t = Tiff(fileScan, 'r');
aUint = read(t); 

% This seemed to work better, though Matlab warns:
% "Converting CIELab-encoded TIFF image to ICCLab encoding"
aUint = imread(fileData);

[Roger Breton]

This code works for me :

aUint8 = imread('LabImage.tif')
aDouble = double(aUint8); 
cielab(:,:,1) = aDouble(:,:,1) ./ (255/100);
cielab(:,:,2) = aDouble(:,:,2)-128;
cielab(:,:,3) = aDouble(:,:,3)-128;

The only 'problem', as noted by opiuz, is that, a* and b* lose their fractional values...

Answer 5

[seackone]

Hi, I want exactly the same! I've got a tiff file and looking for their lab values? Any ideas?

[Image Analyst]

Try rgb2lab() for arbitrary "book" values. Those will not be the same as the true LAB values you'd get from measuring your sample on a spectrophotometer. If you need that, start a new discussion thread (it's complicated) after reading the attached tutorial.