With different sets of points each time, right? Stating the obvious (sorry, but gotta say it just in case), but don’t do this multiple times for a single set of points.
Besides that, it will be hard to give super constructive advice without more context-
- where do the points come from?
- Is there any structure to them?
- Why are you doing this many times /
- Is there any relationship between the points each time you do this?
Without that kind of information it will be hard to give great feedback. That being said, here are some thoughts.
Idea 1 (easy, but probably not a huge benefit)
If you need the bounding box, compute the min/max over all dimensions in one loop.
Idea 2 (parallelize)
Have multiple threads, each finding and returning the bounding box of a subset of your points. It’s fast to find the smallest containing bounding box of several boxes.
Other (context dependent) ideas
If there’s some algorithm generating these points which itself probably has to loop over an image? If so - keep track of the bounding box from within that algorithm.
Idea 4 approximate (maybe not great idea)
If you have tons of points and don’t need an exact answer - randomly sample, and compute on a subset only. Obviously this works best on quantities that are not sensitive to single big values (like min and max are).
Idea 5 (really depends on what you’re doing)
Just saying this so I don’t forget -
Imgbli2 has a kd-tree implementation (see here), that I often use when need to search sets of points by distance.