fitting rectangles into a given shape using Python or Matlab

I have a simple task that needs some programming experience and some basic math skills. Consider the following: There are a few hundred images. Each image is a shape of something, like handwritten digits as in MNIST data or alphabet, etc. The shape has physical dimensions of approximately 3 by 3 micrometer. The task is to fit some smaller rectangles (20 nm < side lengths< 100 nm) into the object shape. The objective is to optimize the size, number and arrangement of the rectangles so that their sum closely resembles the original image.
It seems Gaussian process regression (Bayesian optimization) suits well to solve this problem.