I did the same thing Butzi did (but only posted it here) - except I did
it with a practical image of the sort of thing I photograph, not a test
chart.
And I didn't do the stepwise interpolation myself. I used FM's Stair
Interpolation Pro, which is for any image, and FM Resize Pro for the
camera it was taken with, a 300D.
The results spoke for themselves for me. I don't know if FM does
anything beyond simple 10% stepwise interpolation, but it works better
than any of the three PS bicubic options for me.
And the camera specific Resize Pro is even better
<http://www.moosemystic.net/Gallery/tech/Process/UpSample/resize.htm>.
Be sure to look at the noise in the shadows on the slicker, as well as
the hat creases, cheek and gloves. The RP versions don't look uprezed to me.
If I ever have to uprez a 5D image, I'll be getting Resize Pro for it.
Moose
Joel Wilcox wrote:
> On 12/6/06, AG Schnozz <agschnozz@xxxxxxxxx> wrote:
>
>>> Heh, it's always funny how someone actually just doing a test
>>> can debunk accepted wisdom like this.
>>>
>> The USAF 1951 Lens Chart contains no converging or diagonal
>> lines. If you are comparing raw resolution, I absolutely agree
>> that Stair-Interpolation stinks. However, if you've got lots of
>> lines that can alias on you, a Stair-Interpolation may be your
>> only option.
>>
>> AG
>>
>
> Thanks for that. I feel convinced from looking at *prints* from
> bicubic upsampled files and FM stair interpolated files that the FM
> tool is a Good Thing. It seemed to me that Butzi's using the USAF
> lens chart for his test was a bit like using a brick for a hammer.
> But I couldn't really say why.
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|