[Numpy-discussion] Syntax trouble in translating matlab code to numpy

Oliver Siemoneit oliver.siemoneit@web...
Thu Apr 12 13:53:00 CDT 2007

Dear numpy experts!

I'm just trying to port some matlab algorithm to python which allows an 
image correction for color blind users. If I succeed this bit of code might 
be part of the MoinMoin wiki, the wikisoftware scipy.org uses (see for more 
on that http://moinmoin.wikiwikiweb.de/AccessibleMoin). Since I'm totally 
new to matlab and numpy it took me already hours to figure the basics out - 
without much success. So I decided to ask here on the mailing list for some 
help. I hope - as an exception - this is ok for you...

Here's the part of the matlab code which should be translated. The last 5 
lines actually cause the trouble (long version, see 
    file_name = 'grandparents' ;
    % transorm matrices
    rgb2lms = [17.8824 43.5161 4.11935; 3.45565 27.1554 3.86714; 0.0299566 
0.184309 1.46709] ;
    lms2rgb = inv(rgb2lms) ;

    %read picture into RGB value
    RGB = double(imread(file_name,'jpeg'));
    sizeRGB = size(RGB) ;

    %transform to LMS space
    for i = 1:sizeRGB(1)
       for j = 1:sizeRGB(2)
            rgb = RGB(i,j,:);
            rgb = rgb(:);
            LMS(i,j,:) = rgb2lms * rgb;

And that's what I did with numpy and PIL in Python. The last 2 lines seems 
to be wrong. But how to do it correct?

    im = Image.open(fpath) # fpath is the path to the image
    RGB = numpy.asarray(im) # checked: RGB is of shape[y,x,r/g/b]
    # Colorspace transformation matrices
    rgb2lms = 
    lms2rgb = numpy.linalg.inv(rgb2lms)

    # Transform to LMS space
    LMS = numpy.zeros_like(RGB)
    for i in range(RGB.shape[0]):
        for j in range(RGB.shape[1]):
            rgb = RGB[i,j,:2]
            LMS[i,j,:2] = numpy.dot(rgb2lms, rgb)

This code fails in the last line with the error message "ValueError: 
matrices are not aligned". But how to align them?

Any help would be highly appreciated. Many thanks in advance!

With regards,
Oliver Siemoneit

More information about the Numpy-discussion mailing list