Computer Graphics with Photoshop

Author

resizing question

vivienhughes2000@yahoo.com

2005-12-19, 6:14 pm

I need to increase the size of some images I already have on disc
(can't re-scan them as originals are lost). I may have to double the
file size, which is quite an increase.

My question is: would a specialist program like Genuine Fractals or
S-Spline give noticeably better results than simply resizing them in
Photoshop (mine is Photoshop 6) using bicubic interpolation? I'm asking
this because I really don't want to go to the expense of one of these
programs if I can avoid it.

And if using PS6, is it best to just resize in one jump or in small
increments?

Thanks in advance

Viv

Bill Hilton

2005-12-19, 6:14 pm

> Viv writes ...
>
>would a specialist program like Genuine Fractals or
>S-Spline give noticeably better results than simply resizing them in
>Photoshop (mine is Photoshop 6) using bicubic interpolation?
>And if using PS6, is it best to just resize in one jump or in small
>increments?

Here's an example comparing the "small increments" method (commonly
known as 'stair interpolation') to several other methods, about the
time of Photoshop 6 ... http://www.fredmiranda.com/SI/ ...

/\\BratMan/\\

2005-12-20, 6:14 pm

<vivienhughes2000@yahoo.com> wrote in message
news:1135006990.080925.287490@f14g2000cwb.googlegroups.com...
>
> I need to increase the size of some images I already have on disc
> (can't re-scan them as originals are lost). I may have to double the
> file size, which is quite an increase.
>
> My question is: would a specialist program like Genuine Fractals or
> S-Spline give noticeably better results than simply resizing them in
> Photoshop (mine is Photoshop 6) using bicubic interpolation? I'm asking
> this because I really don't want to go to the expense of one of these
> programs if I can avoid it.
>
> And if using PS6, is it best to just resize in one jump or in small
> increments?
>
> Thanks in advance
>
> Viv
>
Make your own action for it.

Open an image you want res up.

Open the Actions palette and select "New Set".

Name it "Creeping Increase" or whatever you like.

Now select New Action and call it "1.5 Increase"

Next go to Image/Size and put a check in "Resample" and "Constrain
Proportions"

with Bicubic and change the width drop down to "percent" and enter 110.

Now go to "Image Size" do the same 3 more times.

Then "Stop Recording"

Next "New Action" Name it "2.0 Increase"

Now "Edit/Purge All"

Next select the "1.5 Increase" in the Actions Palette

and press the Play button.

select the "1.5 Increase" again,

press the Play button again.

Stop Recording.

"New Action" Name it 3.0 Increase.

"Edit/Purge All"

Select and play the "2.0 Increase"

Select and play the "1.5 Increase"

Stop Recording.

"New Action" Name it 4.0 Increase.

"Edit/Purge All"

Select and play the "3.0 Increase"

Now "Image Size" 110 percent.

And again "Image Size" 110 percent.

Stop Recording.

Save the Action.

Lorem Ipsum

2005-12-20, 6:14 pm

Hello. I'm Stupid. So tell me about this iterative upsampling method.

Why would one upsample in 10% iterations rather than a specific percentage
of the image size with each step, thereby changing the % interation with
each step? What stays constant (or remains insignficant) about the
bit-pattern that makes 10% work?

Or are folks just making intuitive guesses, hoping for the best, or possibly
thinking that somehow they are stealthily creeping up on the laws of nature
so she doesn't notice?

Leonard Lehew

2005-12-20, 6:14 pm

On Tue, 20 Dec 2005 08:16:24 -0600, "Lorem Ipsum" <nospam@nospam.com>
wrote:

>Hello. I'm Stupid. So tell me about this iterative upsampling method.
>
>Why would one upsample in 10% iterations rather than a specific percentage
>of the image size with each step, thereby changing the % interation with
>each step? What stays constant (or remains insignficant) about the
>bit-pattern that makes 10% work?
>
>Or are folks just making intuitive guesses, hoping for the best, or possibly
>thinking that somehow they are stealthily creeping up on the laws of nature
>so she doesn't notice?
>
This method has been around quite a while. As far as I know, it is
empirical in origin. Someone tried it and found that the observed
result was better than a single up-sampling. Sounds a bit like "snake
oil" to me, too, but I have tried it, and, to my eye, it produces
slightly better results than the single up-sampling approach.

I also think that GF produces still better results. The difference is
noticable to me in a side-by-side comparison, but it is not miraculous
-- just a bit better. GF and its ilk provide alternative up-sampling
algorithms to the built-in ones.

These methods are still trying to "guess" what the missing pixels
should be. It takes some experimentation to see what produces the most
pleasing results for you.

Leonard

John

2005-12-20, 6:14 pm

"Lorem Ipsum" <nospam@nospam.com> wrote in message
news:11qg4hoo7juuf0e@news.supernews.com...
> Hello. I'm Stupid. So tell me about this iterative upsampling method.
>
> Why would one upsample in 10% iterations rather than a specific percentage
> of the image size with each step, thereby changing the % interation with
> each step? What stays constant (or remains insignficant) about the
> bit-pattern that makes 10% work?
>
> Or are folks just making intuitive guesses, hoping for the best, or
possibly
> thinking that somehow they are stealthily creeping up on the laws of
nature
> so she doesn't notice?
>
>
Thank goodness - I'm glad I'm not the only one who thinks like that. To me,
it's like suggesting that it is more accurate to measure a distance of 10
metres using a 30cm ruler rather than a surveyor's tape. However, how do you
explain the comparisons in Bill Hilton's link? I can't think of any
scientific explanation.

I know the OP is using Photoshop 6, but does anyone know how Photoshop CS
'Bicubic Smoother' differs from 'Bicubic'? This is usually recommended for
upsampling, and IME produces fewer artifacts than straight bicubic.

--
John
Replace 'nospam' with 'todnet' when replying.

Bill Hilton

2005-12-20, 6:14 pm

> Lorem Ipsum writes ...
>
>Hello. I'm Stupid.

That explains a lot of your posts ...

>So tell me about this iterative upsampling method. Why would one
>upsample in 10% iterations rather than a specific percentage of the
>image size with each step, thereby changing the % interation with
>each step?

The search for the Magic Formula of Resizing was a bigger topic a few
years back than it is now and people tried all sorts of things, like
1%, 2%, 5%, 10%, 50% and varying the % between 25/33/50 in steps (and
of course doing it in one big jump). It's easy to just write an Action
for each of these (if you're not Stupid) and run them and compare the
results. For whatever reason 10% steps seemed to give the best results
on the highest % of images in an acceptable amount of time.

The original poster was asking for a cheap way to resize using
Photoshop 6 (which doesn't have 'bicubic smoother') and the Miranda
link shows comparisons of various methods for doing this, some better
and cheaper than others. If you have a better suggestion then why
don't you provide it?

>Or are folks just making intuitive guesses, hoping for the best, or possibly
>thinking that somehow they are stealthily creeping up on the laws of
>nature so she doesn't notice?

It's easy enough to run tests and compare results, unless of course
you're too Stupid :)

Bill