It depends on the particular image type. these two links could be useful for you.

http://link.springer.com/chapter/10.1007%2F978-3-642-10268-4_3#page-1

and

http://rsb.info.nih.gov/ij/plugins/fraclac/FLHelp/Images.htm

You can read the image as a 2D matrix where each element represents the pixel color value. Then you have two options 1) Use the pixel value as a the 3rd dimension and perform fractal analysis in 3D or 2) Use the pixel value as the mass at that x,y coordinate and perform fixed-mass analysis in 2D.

Here is a sample Matlab code that reads a gray-scale mammogram image and plots it as a 3D scatter. You can then use the resulting point distribution in a fractal dimension estimation. I prefer the 2nd option because for mammograms the color represents the tissue density hence it is more analogous with the fixed-mass estimation interpretation.

If you use the MATLAB, You can use the script:  

BW = im2bw(I, level)

the script will convert the gray-scale mammogram image to a binary image. I indicate the matrix of the images, 'level' argument is the threshold value, you can use the script:

level = graythresh(I)

to obtain the 'level' value.