Here is a neat idea on how to improve images using median evaluation.
- Shoot a number of pictures (at least 10) using a tripod (so that the images are accurately aligned). Let us say that the numpy objects are in a list
obj
- For each pixel position+color
(a,b,c)
look at the values ofobj[n](a,b,c)/8
asn
varies and assignnew(a,b,c)
to be the 8 times the value that occurs most often (mode) in this sorted list.
Now view new
as an image and prepare to be amazed at the improvement!
One more thing to try. Do a time elapsed series of photographs where some objects (say people) are moving in the foreground. With enough pictures, this approach may get you only the background and remove the moving objects.