MATLAB normally deals with matrices of floating point numbers. An image is typically represented by an array of small integer values, pixel value or greyscale values, which have a limited dynamic range and special rules for arithmetic.