Re: MP needed for 24x36 print?


"Roger N. Clark (change username to rnclark)" <username@xxxxxxxxx> wrote in message news:43A650F1.6070209@xxxxxxxxxxxx
SNIP 
In my opinion, one should interpolate to maintain 300 ppi
as a minimum.

I agree. 300 ppi equals approx. 5.9 cycles per millimetre, the lower threshold of human visual acuity. Having more pixels, to come closer to 8-12 cy/mm, is a recipe for utter perfection at any size.

SNIP
My work flow was: 
SNIP

3) Image Restoration Using Adaptive Richardson-Lucy Iteration in
   ImagesPlus 2.5, typically 7x7 box, 7 iterations.

Rather than 'giving in' to ImagesPlus' pre-defined PSFs, if you're interested, I'd be willing to provide you with an optimized Custom PSF that you can use in IP as custom input for the RL restoration.

To allow the creation of the PSF I'd need a .CR2, or a typical 15-16 b/ch baseline conversion crop of it, of an approx. 5 degree slanted edge (e.g. http://www.imatest.com/docs/lens_testing.html#target) that was shot with your preferred lens (typical aperture, or best aperture at around f/8) on your camera from about 25-50x the focal length distance. ISO 100 would give best results, even though you would in practice likely shoot at higher settings.

With that I can approximate the PSF that was caused by lens+aberrations, AA filter, and sampling density (assuming perfect focus, tripod, and mirror lock-up), including PS resampling. It can be used as direct input for PSF based restoration algorithms, or as an approximation of a high-pass filter for sharpening.

SNIP
I feel I get a much better image with the Adaptive Richardson-Lucy algorithm than unsharp mask. See:
http://www.clarkvision.com/imagedetail/image-restoration1

Yes, RL rocks (although it is a bit slow with a large number of iterations). Have you tried Smart Sharpening (lens blur) in PS2? The results come close to Adaptive RL, but it's much faster.

Bart 

.

Relevant Pages

  • Re: A "slanted edge" analysis program
    ... so film/sensor non-linearity and gamma adjustments can't influence the calculations. ... SNIP influence the MTF results, ... "for each line of pixels perpendicular to the edge, the edge is differentiated using the discrete derivative "-0,5; +0,5", meaning that the derivative value for pixel "X" is equal to -1/2 times the value of the pixel immediately to the left, plus 1/2 times the ... What's worse, the PSF can change throughout the image, but a symmetrical PSF will already allow to improve image quality. ...
    (comp.periphs.scanners)
  • Re: MP needed for 24x36 print?
    ... > SNIP ... 300 ppi equals approx. ... I'd be willing to provide you with an optimized Custom PSF ... > b/ch baseline conversion crop of it, of an approx. ...
    (rec.photo.digital)
  • Re: How much dec drift, field rot, and periodic error allowable?
    ... Chris wrote: ... > alter the PSF: they make it broader and possibly asymmetric. ...
    (sci.astro.amateur)
  • Re: More performance tests of different AES implementations
    ... > Are you saying that the timing results vary widely from one run to the next ... benchmark was that the number of iterations was too ...
    (sci.crypt)
  • Re: Problem running 10a demo on Omega
    ... >> demo contains an Obeyfile named SetDefaults, ... >> was not in earlier iterations of the demo. ... Prev by Date: ...
    (comp.sys.acorn.apps)