[Tig] new on the TIG wiki

[Bob Friesenhahn]

On Wed, 9 May 2007, peter swinson wrote:

> I agree with Bob that if the sampling is filtered or the image is 
> filtered to below the Nyquist sampling size then the sampling edges 
> will disappear, but that is just like saying OK I will throw away 
> all the fine resolution of the film.

So you prefer to see something which is not really there. :-)

Due to the semi-random nature of point sampling with film grain and 
sampling noise, it makes quite a bit of sense to sample more points 
and then reduce the image down to a more manageable size (with 
appropriate filtering).  There is more actual detail since each pixel 
is a much more accurate interpolation of the image value than a simple 
point sampling can offer.  It seems that this is becoming popular now.

Now there are datacines that scan (point sample) at 4K but output at 
2K.  According to Peter's information, there is still much detail to 
be gained by scanning at 6K or more.  It remains to be seen what image 
file resolution is necessary in order to provide all of the actual 
detail possible from the higher resolution scan.

I think that the best way to scan an image is likely to sweep a beam 
across the film in both horizontal and vertical directions, and 
interpolate the final pixel value based on the value obtained in both 
directions.  The analog to digital conversion path should contain a 
low pass filter designed to eliminate frequencies higher than can be 
represented by the sampling rate.  CRTs are likely not repeatable 
enough for this but perhaps lasers are.

Bob
======================================
Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/