From: It's NEVER the Camera on
On Tue, 06 Jul 2010 10:40:07 -0700, John Navas <spamfilter1(a)navasgroup.com>
wrote:

>On Tue, 06 Jul 2010 10:25:21 -0700, in
><_budnfYnf52U-q7RnZ2dnUVZ_rKdnZ2d(a)giganews.com>, Paul Furman
><paul-@-edgehill.net> wrote:
>
>>It's NEVER the Camera wrote:
>
>>> Since everyone is into thistles....
>>>
>>> http://farm5.static.flickr.com/4139/4764772879_e644410ff4_b.jpg
>>>
>>> Shot handheld with a with a Fuji Finepix 2600z, 2.1 megapixel, fully
>>> automatic, 3x zoom, P&S camera (about 10 years ago) on a very cloudy and
>>> overcast day. (Note total lack of shadows.) Using a close-up filter
>>> fashioned from a single-element plastic "achromat" salvaged from an old
>>> 126-Instamatic.
>>>
>>> With careful upsampling and applying a Fourier transform at each stage this
>>> can be successfully printed to 18"x13" from a 2.1 megapixel camera. This is
>>> the downsized (and intentionally jpg degraded) image from the 3300x2475
>>> upsampled version used for that very purpose.
>>
>>Nice pic, but 8 megapixels from 2 (given half a megapixel to see). Oh
>>boy. I've printed 3 megapixels at 13x19 - it can be passable flipping
>>through quickly if the content is good.

Here's where you don't realize that content is more important than pixels.
Due to the large subject sizes and bold color-contrasting composition
viewers don't concentrate on looking for every hair on the body of the
beetle or minute depressions in the elytra.

>
>I liked the Fourier transform bit. LOL!
>
>Even with Genuine Fractals, I'd say 13x18 from 2.1 MP
>would be no better than "acceptable" at normal viewing distance.

Try it sometime. I never cared for the oil-painting artifacts that Genuine
Fractals creates. Someone gave me the latest version and I didn't even
bother to install it this time. Don't apply the Fourier-transform at full
strength for each upsampling step. Gently nudge the edges where you want
them. At times using a 2 to 3 pixel radius at 25%, other times a 1 or 2
pixel radius at 50%. A gaussian blur or noise-removal step to subdue the
ringing artifacts at some stages of multi-step upsampling can also help.
The amount used at each upsampling step is subjective and wholly dependent
on the subject matter. Like noise-removal methods, this can't be a "one
standard sequence fits all" operation. I will also mix and match upsampling
algorithms during this process. Bicubic when I need to soften edges for the
next Fourier-transform step; or use a Lanczos, Bell, or Spline upsampling
step when I want to retain the details created by the last
Fourier-transform.

This does of course require that you have pixel-level details to start
with. If your details are spread over 2 or more pixels to begin with you
won't be able to upsample as much. That little Finepix camera has an
amazingly well matched set of optics and sensor. With 1 quick upsampling to
smooth the stair-step pixel edges 8"x10" prints are the norm from it. No
different than a print made from a DSLR where details are commonly spread
over 2 to 3 pixels. Each pixel from that Finepix camera contains discreet
and valid image information. Though this does depend on if your images are
worth looking at. Mine usually are. People never look at the pixels of the
images that I print. If someone is looking at your pixels and not the
content that they convey then you've failed as a photographer.







From: me on
On 2 Jul 2010 17:20:58 GMT, ray <ray(a)zianet.com> wrote:

>Totally untouched photo of a thistle (canadian thistle, I think) I shot
>near Cauldron Linn on the Snake River a couple of days ago.
>
>http://www.rcarter.net/region/thistle.jpg


Nice shot. Now wait for it to go to seed and look for shots of birds
feeding on it.