I have a smallish file that I have to use large in a calendar. It has
obvious pixellation. Is anyone willing to help me process this into a
larger size?
Thanks
Ah, that brings back memories - I used to enjoy playing with it.
If you had the original program and knew the parameters you
had fed into it, it may be possible to generate a new similar image
with a higher resolution - but, sorry, I can't help...
--DR
I've used Genuine Fractals a few times, but the paint-by-numbers appearance
of the detail enlargement isn't to my liking. It's also very costly for
what it claims to do. As well being rather convoluted in functionality, not
easy for a novice to use.
You might want to search out any utilities that use the S-Spline upsampling
technique. One of my favorites being BenVista PhotoZoom Pro.
While not having S-Spline methods, Qimage (for printing) and Cleanerzoomer
are both adequate upsampling utilities too. As are some editors that
include more than the common bicubic resampling methods. Some subjects work
better with certain resampling algorithms than others (Bell, Pyramid,
Triangle, Lanczos, etc.). There's no real cut & dried answer in this
regard. But S-Spline capable utilities are usually the safest bet for
upsampling.
You might like to also apply a Fourier Transform utility after upsampling,
commonly marketed as focus-sharpening utilities. One of my favorites is
marketed under the name of Focus Magic. When not applied too strongly it
can tighten up some of those soft edges created when upsampling an image
greatly. Creating some false detail where there was none before. Focus
Magic, when used together with S-Spline upsampling first, can create some
quite believable detail. Sometimes it's best to apply a Fourier Transform
in several smaller and weaker stages, rather than one large strong step in
focus adjustment. Applied to strongly and you'll get annoying, what they
call, "ringing artifacts". You'll learn to recognize what this is the first
time you play with a Fourier Transform tool. It's easy to get those ugly
artifacts if you aren't careful.
Any tool is only as good as the talent of the person manipulating that
tool. This applies to all editing software (noise removers, resampling
tools, basic editing, etc.) as well as the cameras in hand.
How big is your original? (X pixels x Y pixels)
How big do you want the Calendar image to be? (X inches x Y inches)
Do you have Photoshop? Any other image editor?
Bob Williams
GF is mostly marketting hype. You cannot magically blow up tiny bits of
image and obtain more detail from nowhere in the style of Bladerunner.
In some limited instances where natural textures like trees or rocky
crags that are roughly self similar fractals GF can invent plausible
looking "new" but fake detail. These conditions are seldom met in reality.
Your best bet is to upscale the image by factors of 2 with the best
interpolation function your image processing package provides (some
would argue for multiple upscalings of 1.1x) and then apply unsharp
masking to taste on the final image to bring the edges back to
sharpness. It might be good enough if you are *very* lucky.
You are unlikely to be satisfied with the results however it is done.
Regards,
Martin Brown
It's 600 x 450 pixels, 72 pixels/inch resolution. I have Photoshop 5.0
LE, and have enlarged the image using that program, judiciously
applying sharpening etc. It still looks like crap. I need to use it in
a calendar at about 30-40% bigger than its actual size. I'm sure you
see the problem. I need to use it at 9 x 5.23 when it should be used
at about 6 x 3.5.
I can e-mail it to anyone who can help.
My e-mail is:
petraioprime
@
yahoo
dot
com
You're not going to sell many Calendars!
BugBear
LOL. It's a company calendar showing bridges we have designed. This
photo was taken by the county engineer with a primitive digital camera.
A 6" x 3.5" print from the above pixels will itself look like crap,
never mind at a larger size. You really need to retake or substitute
something else.
--
john mcwilliams
Ah, sorry, I was thinking of a different program, although I
do own Genuine Fractals also (wanna buy it? ;-). I've been
able to get decent horizontals for DVD covers from SD
video frames, but for anything larger than about 1.5x (with
compatible image material), the results are not great - and
your photo does not look like it would work well trying
to do anything with this program... You may be able to get
around this by having a graphic artist turn the image into an
illustration based on the photo (which would also likely
result in a "prettier" image for the calendar...).
--DR
Or hire an illustrator to do a photorealistic rendering based on the photo.
This will probably cost more than just reshooting if the subject is still
accessible.
That's probably the best idea yet. A decent arist with the photo and
Adobe Illustrator could produce something that would look far better
than the photo magnified far more than is sensible.
--
Ray Fischer
rfis...@sonic.net
Have you tried using a photo editor to do an 'artistic effect' - like
'clothify' or 'impressionist' or 'cartoon' - lots of possibilities
several of which will likely end up looking a lot better than what you're
trying to accomplish.
>On Dec 24, 3:09 am, Bob Williams <mytbobnos...@cox.net> wrote:
Here's using that even smaller 500x375 version posted at flickr. Better
results could be obtained with the slightly larger original.
Yours
http://farm5.static.flickr.com/4055/4210457549_b4014207ec.jpg
Upsampled to 1800x1350 for 300dpi printing. A 360% enlargement.
http://farm3.static.flickr.com/2679/4211407297_b383284354_b.jpg
(flickr again downsizing it to 1024x768)
However, for a calendar you could easily get away with 150dpi, upsampling
to only 900x675 pixel dimensions.
Upsample to 6"x4.5" at 300 dpi using B-Spline algorithm. Not S-Spline this
time, B-Spline was better for this type of subject, when knowing I was
going to again later use a Fourier Transform tool on it. No unsharp-mask
sharpening. Then applying the Fourier Transform in 3 steps.
4 pixel dia. at 75%
2 pixel dia. at 75%
2 pixel dia. at 75%
(Those are not cut 'n dried rules for transform steps, it's just what
seemed to work okay with so little time and effort devoted to this)
Compare the images side-by-side of the workers at the edge of the bridge to
see what kind of plausible results can be munged out of so few pixels.
http://farm3.static.flickr.com/2696/4212201408_0430234f5e_o.jpg
Half-tone printing in most magazines and calendars, at that image size, has
far less detail than this does. Besides, you say it's for a company
calendar. Most of those will be used for dart-board practice by any
employees or thrown in the trash or used to light their fireplaces when
they get it home. Make sure you credit the photo with original pixel
dimensions to get yourself off-the-hook, should anyone whine about the
image quality.
I wouldn't suggest using this one I that uploaded, as it was done with your
even-lower-resolution image from flickr and I used a high JPG compression
on it so as to not waste bandwidth. Not to mention that flickr downsized it
and JPG compressed again.
Use the above method and try it yourself with your own tools. I have no
intentions of offering further help than this. Unless you give me ten times
that portion of your hourly paycheck or salary for the time you would have,
and already have, spent doing this yourself.
smalish filkes do not work as well as large files. You can't make a
whip out of a shit.
use as is.
On 12/24/09 9:24 AM, in article
e680fd7d-fafd-4675...@h9g2000yqa.googlegroups.com, "UC"
<uraniumc...@yahoo.com> wrote:
Should have hired a pro.
Impossible. Was taken by county engineer during construction of a
bridge several years ago.
>>> It's 600 x 450 pixels, 72 pixels/inch resolution. I have Photoshop 5.0
>>> LE, and have enlarged the image using that program, judiciously
>>> applying sharpening etc. It still looks like crap. I need to use it in
>>> a calendar at about 30-40% bigger than its actual size. I'm sure you
>>> see the problem. I need to use it at 9 x 5.23 when it should be used
>>> at about 6 x 3.5.
>> A 6" x 3.5" print from the above pixels will itself look like crap,
>> never mind at a larger size. You really need to retake or substitute
>> something else.
> Impossible. Was taken by county engineer during construction of a
> bridge several years ago.
I understand that now. So, either substitute something else, or your
next best bet was suggested by John Clark.
--
john mcwilliams
In stead of GF you might give the demo version of Qimage a try. It
contains some of the best upsampling algorithms. But given the size of
the original you stated elsewhere, it sounds like you are asking for a
miracle.
Yeah, I need a miracle. LOL