Benchmarking of improved image positioning
We want to see the performance impact of the improved image processing (via the CV2 perspective method). To do this, benchmarks have to be made for the old system (rotate/shift image) and the new system.
My proposal is to test for difference in time to do the image positioning, as well as the improved accuracy after positioning. With the latter I want to compare the corner markers found in the positioned image using both methods to the 'perfect' exam. So far I've found the new method to work better than the previous, but this needs to be quantified.
I think it's best to only benchmark the actual calls related to positioning; e.g. not take into account the time that it takes to find the corner markers