I can't help noticing that when I reduce the resolution of digitised images they look a lot fuzzier than the original, and need to have the "sharpen" filter applied to them. I can understand the need for sharpening after increasing the resolution, but to have to do it after reduction just seems to run against common sense.

Does anyone know the technical reason why?

-- Pete Andrews (p.l.andrews@bham.ac.uk), February 28, 2000

Answers

Hi Pete,

Well, I've never written a general purpose algorithm to do scaling, but...

I think it's because the same core mathematical process is being used to go in either direction, averaging. The difference is what you do with the averages. You scale up by creating new pixels that are assigned values averaged "between" the existing pixels and inserting them into the image matrix[well actually, you create a new matrix as you go along and discard the old one when done... :-)] You scale down by averaging pixels in the existing image and then just keeping the averaged pixels. It's a twitch more complex than that, but it's a simplified example that works well in terms of 50% reduction or 200% enlargement. An actual algorithm would probably use the ratio of pixels in the final image to the original to determine just how many pixels would be averaged to create each new pixel of the final image.

Both processes tend to eliminate extremes of shade, intensity, etc. between pixels through the averaging. Since what we percieve as sharpness seems to be sharp transitions in color or brightness, it gets affected.

I think it would actually be a lot stranger for any process if the effect only existed in one direction...

Anyone else have another take on this?

-- Gerald M. Payne (gmp@francomm.com), February 28, 2000.

Thanks very much Gerald, that's good enough for me. It seems obvious now you've pointed it out. Not so obvious why a re-sharpening algorithm isn't combined with the reduction though.

-- Pete Andrews (p.l.andrews@bham.ac.uk), February 29, 2000.