Your opinion on the new resampling algorithms in CS?

RA
Posted By
Rafael_Aviles
Dec 19, 2003
Views
437
Replies
13
Status
Closed
Has anyone done a comparison of the new Bicubic (Smoother) and Bicubic (Sharper) resampling algorithms? Do they give noticeably different results? Which one do you prefer for photographic images?

Thanks!

Rafael
Related Tags
new opinion resample

MacBook Pro 16” Mockups 🔥

– in 4 materials (clay versions included)

– 12 scenes

– 48 MacBook Pro 16″ mockups

– 6000 x 4500 px

RA
Rafael_Aviles
Dec 21, 2003
(Bump)

Anyone?
DM
dave_milbut
Dec 21, 2003
I haven’t done any tests, per se, but I’ve been using chris cox’ advice:

Chris Cox – 06:37pm Nov 26, 2003 Pacific (#3 of 4)

Dave –

Bicubic for simple transforms

Bicubic smoother for upsampling

Bicubic sharper for downsampling

But that is only a rough guideline – some images may do better with other choices.

And it certainly ain’t any worse! 🙂
HG
Howard_Goldberg
Dec 21, 2003
Bicubic for simple transforms
Bicubic smoother for upsampling
Bicubic sharper for downsampling

How do these choices compare with programs like Genuine Fractles?
DM
dave_milbut
Dec 21, 2003
I dunno. It’s the weekend. Maybe on monday chris or one of the other guys from adobe will be able to comment.
PC
Pierre_Courtejoie
Dec 21, 2003
What? they don’t work on weekends? <G, D & R>
RA
Rafael_Aviles
Dec 21, 2003
Thanks!

And have a ver happy holiday season!

Regards,

Rafael
Q
Quale
Jan 28, 2004
For downsampling the advice is to use bicubic sharper. Is bicubic sharper much like bicubic plus some USM, or is bicubic sharper a different enough algorithm that would give better results, retain more detail without having jagged edges, than using bicubic followed by a sharpening of some sort?
BO
Burton_Ogden
Jan 28, 2004
Rafael,

There is a good comparison of upsamplers at

Digital Photo Interpolation Review < http://www.americaswonderlands.com/digital_photo_interpolati on.htm#Kneson%20(S%20tandard)>

I have been using LizardTech’s Genuine Fractals Pro, but from this review it looks like Shortcut Software’s S-Spline Pro, now renamed as PhotoZoom Pro, is worthy of consideration. Be sure to expand the comparison photos on the website above, because you can’t tell too much from the thumbnails, or from the original expansions of them. You have to go down to the lower righthand corner of the photos to activate the expander icon to bring the photos to full size, and then you have to scroll over them to see all the details.

This comparison site was an eye-opener for me. I had assumed that Extensis’ new product would be competitive, but when you take a really close look, it is quite disappointing. Likewise for some of the other contenders.

— Burton —
CC
Chris_Cox
Jan 28, 2004
Bicubic sharper is different enough that it will get better results than post processing with USM (because the interpolation kernel can take into account the original image pixel values).

Also, that review doesn’t take into account the many resampling improvements in Photoshop CS.
Q
Quale
Jan 29, 2004
And my need is to downsample. Everybody wants to sell you something to upsample. What happens when you have a nice big scan and need half the resolution…I don’t want degrade the image more than necessary while downsampling.

I am scanning large format slides in at 4000 dpi, but only need 2000 dpi for output (at this time). Do you get better results downsampling from 4000 to 2000 dpi as compared to originally scanning at 2000 dpi? If the 4000 dpi scan is able to resolve film grain, will downsampling show some of that grain? Less dpi’s won’t show grain.
BO
Burton_Ogden
Jan 29, 2004
Quale,

And my need is to downsample. Everybody wants to sell you something to upsample.

It’s true that most people need to upsample. But many, if not all, of those products can also be used to downsample. I know firsthand that is true for Genuine Fractals.

If the 4000 dpi scan is able to resolve film grain, will downsampling show some of that grain? Less dpi’s won’t show grain.

I think it is highly debatable whether scanning at a lower resolution is a solution for grainy film. I think some people would say that scanning at a lower resolution is worse for grain. This grain vs resolution thing is almost achieving the status of an urban myth.

However, theoretical arguments frequently generate more heat than light. The best way to resolve the question is side-by-side comparisons, like in the web site I referenced. Right offhand, I don’t know of a place where that has already been done for downsampling with various products. But it wouldn’t be rocket science to downsample a test image or images with several competitive products. Do you have a sample image you would like to submit for tests? Judging the results might be somewhat subjective.

— Burton —
Q
Quale
Jan 29, 2004
I have two scans of the same image. One is scanned at 4000 dpi, the other at 2000 dpi. I downsampled the 4000 dpi scan to 2000 dpi using each of the available PSCS options. I did not yet try downsampling with third party plugins as suggested. I applied a little USM after downsampling to the bicubic and bicubic smoother methods so as to make all images the same apparent sharpness. Comparing the downsampled results, the bicubic sharper looks subtly better than the others. It’s not a big difference, but worth the effort to click the bicubic sharper option. Also, the bicubic sharper downsampled method compares most like the image scanned in originally at 2000 dpi.

My preliminary results suggest that if one knows the final output dpi, scanning at that final resolution works well. There isn’t an improvement in image quality by scanning at a higher resolution, then downsampling. Perhaps if noise reduction were performed at the higher resolution prior to downsampling (higher resolution was 4000 dpi in my case) the result might be better than just scanner at the intended output resolution (2000 dpi in my case).
BO
Burton_Ogden
Jan 29, 2004
Quale,

Perhaps if noise reduction were performed at the higher resolution prior to downsampling (higher resolution was 4000 dpi in my case) the result might be better than just scanning at the intended output resolution (2000 dpi in my case).

I agree with that approach. I use Neat Image a lot, and also Focus Magic. I think a dose of Neat Image and Focus Magic could help before downsampling. Also, I think it would be worthwhile to experiment with Genuine Fractals and S-Spline Pro as downsamplers. Although it is good news that Photoshop CS offers the new option for downsampling. As you pointed out, quality downsampling has been neglected too long.

— Burton —

MacBook Pro 16” Mockups 🔥

– in 4 materials (clay versions included)

– 12 scenes

– 48 MacBook Pro 16″ mockups

– 6000 x 4500 px

Related Discussion Topics

Nice and short text about related topics in discussion sections