Oh that's easy. Consider the following image.
A = imread('cameraman.tif');
Now you're done.
Seriously though. This is a nonsense question. The "gray level values" of a grayscale image is the image. There's nothing to "scan" or "detect". Was it about finding average gray levels? What about extreme values instead of the mean? Would those be local or global measures? Histogram counts? Optimal threshold levels? GLD? GLCM? If there were something else intended by the question, it's been lost.
The slowest way to be disappointed is to ask for something so apparently ridiculous that everyone ignores you. You want the conversation to start with efforts toward creating the solution to an understood problem. You don't want the conversation to start with people asking you to explain your question. When people see the conversation headed in that direction, they'll avoid it.
