At the moment, all of the above routines are written as separate programs. In future, I hope to combine them into one smaller and faster program. This will also allow me to add a layer of optimization around the last several steps (everything after the extrapolation) that uses the total or maximum pixel error given the parameters estimated as its cost function, and the positions of the corners as the optimized parameters in some way. In other words, the optimizer will change the positions of the detected corners slightly in some parameterized way (such as by a variable amount along the gradient) to try to minimize the error in final point locations. This step is just an extension of Coordinate Adjustment above.