[18335fall08] matlab notes on problem 3

[Steven G. Johnson]

When you are reading an image into Matlab with "imread", it reads it  
in as an integer array, and you may have noticed that this gives an  
error with SVD etc.  To perform an SVD, you need to convert this to  
floating-point.   You can do this with the "double" command, e.g.:

A = double(imread('myfile.jpg'));

To plot it with pcolor, I noticed that pcolor displays it upside- 
down.  Also, the grid lines are annoying, and the aspect ratio is  
distorted.  You can plot it nicely with:

pcolor(flipud(A)); colormap(gray); shading interp; axis equal

Finally, I should mention that things are more complicated for a color  
image, which is why I recommend sticking to a grayscale image.  For a  
color image, it reads it in as a three-dimensional array where the  
third dimension (for an RGB image) is red, green, blue.  To convert a  
slice of this, e.g. A(:,:,1) into a 2d array for SVD, you would need  
to use the "squeeze" command: squeeze(A(:,:,1)).  Then, I suppose you  
could do SVD compression on each color component individually, then  
combine them back together and write it out to a color file ..... but  
it is much easier to just use a grayscale image.

Steven