I think that's the purpose of the finder patterns. From a Source image I imagine you would locate the finders, find the edges, unskew the source image, and use

I think that's the purpose of the finder patterns. From a Source image I imagine you would locate the finders, find the edges, unskew the source image, and use the finders to correct the rotational orientation, and then crop the image. Maybe binarize the new image to make the next step easier. At that point, it's a matter of working out the size of each cell to generate a grid to drop on top of the image. Then you can sample each cell of the grid. If it's light, 1. If it's dark 0. Now you have the pixel map. I am sure it's more complex, but I think a lot of this sounds like stuff you could do with openCV