True about resizing ...

Lot was written about the true resolution of different scanners - let me take as an example one of the flat-bed scanners - where the true resolution at the setting of 4800 dpi is only about 2000 dpi. But what happens when we resize the final scan?

To make this all more clear let me consider two different scanners at the same scanning spi - say 4000:

scanner A) The true resolution is very close to spi of the scanners (good quality drum scanners or other high end ones)

scanner B) True resolution is about 1/2 of the scanners spi (that is 2000). In other words only (approximately) 1/4 of the produced data contains true information - the rest is just ballast.

Now - what happens to the information in such a scan (where the true resolution lags behind the spi) if we DOWNSIZE the file (for example with Photoshop) - say by linear factor 2 (factor 4 in the total amount of pixels). The scan (A) will loose 3/4 of the original information, but the resulting image will have the same "information density" as the original.

But what happens to the scan (B)? Will the software miraculously manage to throw away the "useless" 3/4 of the image and leave the 1/4 full of information? In such a case - after the downsizing - the scan (A) should be more-less identical to the scan (B) ??? Or the resizing software "randomly" decides which data/pixels to keep and we will end up with a scan that is not any sharper than the original. That would be the best and the worst scenario respectively.

But what truly happens? :eek: ;)

-----
The above was triggered by the review of one of the Plustek 7600i scanner where it was found out, that true resolution @ 7200 spi is about 3250, but @ 3600 spi only 2600. While this is not the same as what I discussed above, it just got me thinking..

I don't really know the answer to your question, but I can tell you approximately what downsizing involves. It uses some averaging system to reduce the number of pixels rather than just omitting pixels. The averaging invovles neightporing pixels and there are different ways to do it. Some are faster and some are slower but producin better results. Often you can choose which method to use.

Even if a scanner is said to return 4800spi but only actually returns 2400spi, there is no such thing as "ballast"; every pixel wil be a representation of the image seen by the scanner. When folks say that there is no advantage in scanning at the higher (4800) resolution, it usually indicates that, in the resultant file, there is very little perceived extra detail to be seen over the lower resolution. Sometimes that lack of perceived detail can simply be down to the inability of the lens to resolve the detail but what happens is that the pixels in the unclear areas are simply an average of the colour values both in and around the detail.

Perhaps an example would help: Draw a tic-tac-toe game on a piece of paper, fill the nine squares with a random mixture of O and X. Now, imagine tha the scanner can only resolve a pixel as small as the whole tic-tac-toe game. What happens is that the resulting pixel becomes an "average" (best guess) of the nine elements in that area.

And so it is when down sizing a file; well as a simple analogy anyway. The downsizing takes a "bunch" of pixels and tries to make sense of what will look best when the nine squares are reduced to one, but taking into account all the other tic-tac-toe games adjacent to the one that it is trying to compress.

As for the difference in capability between, say, an Imacon and an Epson flatbed, Although the Imacon is said to be able to resolve up to 6300spi, in actual fact, that is only for 35mm film, 4x5 resolution is only stated as being 2040spi; which is similar to that obtainable on an Epson V700. The main perceived difference seems to be the Imacon's ability to extract a bit more shadow detail.