No-one knows. Stored-Program computers have only existed for about 70 years. 70 years is not nearly long enough to determine the best way to calculate anything as ill-defined as the "edge" of objects in images. It will probably take another several hundred years, maybe more. Perhaps when we have finally created a complete model of the eye and the vision system we might be able to optimize a computer algorithm. You can be sure that the best way will not look anything like what you would program now.
Mind you, it is likely to be at least another 6000 years before we even figure out what "best" means when it comes to multiple objectives. For example, for some people, a difference of half of a pixel in a single location on an image would be worth spending a few million dollars and ten years of computation for, whereas other people might say that if it takes more than 2 years to write the program then the algorithm is useless to them.
