Pages: 1 [2]   Go Down

Author Topic: CS6 Bicubic Automatic: Dissatisfied User  (Read 18743 times)

Vladimirovich

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1311
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #20 on: March 09, 2013, 10:37:58 am »

Hi,

I've tried it, but it still creates more aliasing artifacts than a simple Gaussian pre-blur plus Photoshop Bi-cubic resize does. It seems the plugin's emphasis is on correct color/brightness/contrast (which is better because Photoshop doesn't simply allow to resample with linear gamma).

Cheers,
Bart

what about a simple Gaussian pre-blur plus this one instead of Photoshop Bi-cubic ?
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #21 on: March 09, 2013, 10:50:33 am »

what about a simple Gaussian pre-blur plus this one instead of Photoshop Bi-cubic ?

Hi,

That's always a possibility, but it kind of defeats the purpose of a resizing plug-in. Why doesn't the plugin also do a proper Low-pass filtering when downsampling (since there are better ones that a blurry Gaussian) one could ask? Applying a manual pre-blur step also requires to calculate the required radius based on the amount of downsampling, typically something one would want to automate.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Vladimirovich

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1311
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #22 on: March 09, 2013, 11:21:13 am »

That's always a possibility, but it kind of defeats the purpose of a resizing plug-in.
well, as far as I understand author had a very simple purpose - to make a free PS plugin (so that you will stay within PS w/o going outside) better than what PS embedded ("It seems the plugin's emphasis is on correct color/brightness/contrast") variety of resizing methods do in some aspects and not worse in any others, not the absolute perfect one (also note that it is just several days old - in terms of it being released to the public)...
« Last Edit: March 09, 2013, 11:22:58 am by Vladimirovich »
Logged

MarkL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 475
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #23 on: March 13, 2013, 02:40:53 pm »

I thought I was the only one, bicubic sharper is horrible. All of photoshop's up and downsampling methods are less than ideal, it is irritating to use to dump an image into different software/use plugins to do it well.
Logged

fike

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1413
  • Hiker Photographer
    • trailpixie.net
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #24 on: May 14, 2013, 01:24:17 pm »

So after learning a bit from this thread and doing some experimenting, I have had a change of heart.  It isn't that bicubic sharper is bad, it is that it shouldn't be used for large magnitudes of downsampling in one step.  I have had some really good results taking an image that is say 8,000 pixels wide and downsampling to 7,000 and then 6,000 and then 4500 and gradually going down to the destination size (which is usually web display).  The resulting sharpness is better than I usually get from my old sharpening workflow for web publication (in the past I sharpened at original resolution, downsampled, and then resharpened at web resolution).
Logged
Fike, Trailpixie, or Marc Shaffer

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #25 on: May 19, 2013, 06:46:50 pm »

Given all the complexity of resizing an image in photoshop , I wonder what Lightroom 4 does? This software is deceptively simple, you just plug in print size and behind the scenes it does rescaling and output sharpening. What is going on in there? Does anyone REALLY know?
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #26 on: May 19, 2013, 06:55:05 pm »

Given all the complexity of resizing an image in photoshop , I wonder what Lightroom 4 does? This software is deceptively simple, you just plug in print size and behind the scenes it does rescaling and output sharpening. What is going on in there? Does anyone REALLY know?

Near as I can figure out from the outside, pretty much Bicubic Automatic. That wasn't the case with Lightroom 3.

Jim

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #27 on: May 20, 2013, 10:03:45 am »

So if I want deconvolution sharpening (as discussed here: http://www.luminous-landscape.com/forum/index.php?topic=45038.0) I shouldn't even use photoshop. Rather, export to tiff when done and open in photozoom or onOne resize- do the resizing-then sharpen in that app and print from there?
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #28 on: May 20, 2013, 10:32:08 am »

Near as I can figure out from the outside, pretty much Bicubic Automatic. That wasn't the case with Lightroom 3.

Actually, that's not quite right, if I remember a conversation with Eric Chan properly. He said that Lr 3 did use bicubic interpolation, but it did it while the image was in a linear (gamma = 1) representation. That gave a softer result than Lr 4, which does the interpolation while the image is in a gamma-compressed form.

Maybe that's too much information...

Jim

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #29 on: May 20, 2013, 10:50:04 am »

So if I want deconvolution sharpening (as discussed here: http://www.luminous-landscape.com/forum/index.php?topic=45038.0) I shouldn't even use photoshop. Rather, export to tiff when done and open in photozoom or onOne resize- do the resizing-then sharpen in that app and print from there?

Hi,

It would be best to do proper Deconvolution Capture sharpening at the Raw conversion stage, and thus avoid halos, if the lens characteristics are known. Then do the upsampling with a dedicated algorithm for up-sampling, and if necessary (depends on the algorithm) do additional (deconvolution) sharpening at the output size to compensate for the up-sampling losses, and pre-compensate for the output medium losses.

Cheers,
Bart
« Last Edit: May 20, 2013, 01:12:50 pm by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #30 on: May 20, 2013, 11:10:04 am »

Bart,

You just trebled the time I would spend out of Lightroom . The back and forth questions the use of Lightroom at all. Could you refer me to a link that would hand hold me through your suggested workflow? I am quite puzzled as to how to perform, view and measure the three types of sharpening outside of Lightroom. Is this something I could do in gimp? Or is a dedicated app like resize preferred? Then of course I have to do the actual 'darkroom work' - exposure, color, noise adjustments for which photoshop or gimp or Lightroom can be used.
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #31 on: May 20, 2013, 11:48:05 am »

Bart,

You just trebled the time I would spend out of Lightroom . The back and forth questions the use of Lightroom at all.

I'm working through this issue myself. There are a lot of things I like about Iridient Developer. I like it so much a got a Mac just to run it. However, Lightroom's raw development is plenty good enough for most of the work I do. That's true for resizing for printing, and some specialized processing as well.

So I'm thinking that I'll do all my cataloging, and most of my developing and printing in Lr. For those 4 or 5 images a month that I really care about, I'll go to the trouble of redeveloping in Iridient Developer, tweaking in Photoshop, carefully resizing for printing, etc.

That's the way I already work, except for the redeveloping.  I already sometimes go back to the raw files and start over when I get a new print order; almost always I find I can do a better job knowing what I know about he image from working on it the first time, but starting with a figurative clean sheet of paper.

Jim

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #32 on: May 20, 2013, 02:20:37 pm »

Bart,

You just trebled the time I would spend out of Lightroom . The back and forth questions the use of Lightroom at all.

Hi,

Lightroom by itself offers a decent enough quality for most users. However, if you want better quality, then you'll need to use dedicated tools, when they do a better job than Lightroom.

Quote
Could you refer me to a link that would hand hold me through your suggested workflow?

Since the Capture sharpening dialog of Lightroom (and many other Raw converters) does not really help users enough (IMHO) with finding the correct settings, I've created a tool that can assist in getting a grasp on better settings:
http://www.luminous-landscape.com/forum/index.php?topic=68089.0

Unfortunately, the up-sampling routines of the current Lightroom (v4.4) are not as good as some of the other offerings. They are not bad, but just not as good as some others. For most significant up-sampling jobs, it is hard to beat the S-Spline Max algorithm from Benvista's Photozoom Pro, since it actually increases edge resolution. For more modest up-sampling jobs, the routines offered by ImageMagick (in particular with the '-distort Resize' filtering) offer very good up-sampling quality, with fewer artifacts, which allows more sharpening. I've analyzed/compared a few methods here, based on the Fourier transforms of an up-sampled White noise image that contained all spatial frequencies before up-sampling.

Quote
I am quite puzzled as to how to perform, view and measure the three types of sharpening outside of Lightroom. Is this something I could do in gimp? Or is a dedicated app like resize preferred? Then of course I have to do the actual 'darkroom work' - exposure, color, noise adjustments for which photoshop or gimp or Lightroom can be used.

As I said earlier, you can do most of that in Lightroom, but there are some things that can be improved by utilizing specialized tools. The Gimp is getting better, but it really requires a 16-bit/channel workflow to maintain the quality at a high level. The specialized tools will complicate the workflow, but that's unfortunately the toll that apparently has to be paid for additional quality. Not everybody is willing to pay that price, and not everybody has to (unless they demand better quality). One has to repeatedly ask oneself if the diminishing returns are worth the trouble. For those seeking the best, it is.

Excellent results can be achieved with a workflow consisting of a Rawconverter like e.g. RawTherapee (also offers Richardson-Lucy deconvolution sharpening), TopazLabs photoFXlab with specific filter plugins for editing (if one requires Photoshop's adjustment layer capability), ImageMagick for up-sampling, and e.g. Q-image Ultimate (does also offer pretty decent upsampling, and 'DFS' output sharpening) for printing.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #33 on: May 20, 2013, 08:00:28 pm »

I was looking at  your tool here (http://bvdwolf.home.xs4all.nl/main/foto/psf/SlantedEdge.html)
Should I use an edge that goes from complete black to complete white? Does the number of pixels over which the transition occurs have an optimal amount?  Obviously there are numerous edges one can chose in an image.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: CS6 Bicubic Automatic: Dissatisfied User
« Reply #34 on: May 20, 2013, 08:54:17 pm »

I was looking at  your tool here (http://bvdwolf.home.xs4all.nl/main/foto/psf/SlantedEdge.html)
Should I use an edge that goes from complete black to complete white? Does the number of pixels over which the transition occurs have an optimal amount?  Obviously there are numerous edges one can chose in an image.

Hi,

Good questions.

I prefer the edge to go from dark to light for numerical analysis, not black to white. The reason is that it would allow to detect a halo (under/over-shoot). That would give more information than a totally clipped image.

Having said that, that also means that the source image can in principle go from black to white, because the captured image will suffer from some lens glare that reduces the brightness range to be captured to quite a bit less. However, printing a slanted edge target also requires to be able and accurately print the fine transition between light and (paper) white. One should likewise also prefer and avoid too much ink diffusion at the black end. It's one of the reasons that the ISO cater for a relatively modest contrast target.

As for the angle, there have been various tests and they do tend to confirm the choice that the ISO made for their recommendation, an approx. 5-6 degree slanted edge (1 in 10 pixels equals a 5.71 degrees angle) produces a good compromise. One of the reasons is that the edge should be close to a horizontal/vertical edge in order to test the pixel grid at orthogonal sampling directions close to perfectly horizontal/vertical. The other consideration was to allow an adequate oversampling for computational accuracy, but not to over-do it. The ISO procedure effectively aims for 1/4th of a pixel accuracy, my method aims for 1/10th of a pixel, but with a variance due to shooting the target not perfectly square (which could mean more/less than 5.7 degrees).

Cheers,
Bart

P.S. When creating a target to be photographed, the main consideration for creating a sharp edge on an inkjet printer is that one should try to avoid too much density which could result in ink diffusion, or very nearly whites which on an inkjet will be represented by isolated droplets. That's the reason why the slanted edges of my resolution target, go from very dark gray (not black) to paper white.
« Last Edit: May 21, 2013, 05:13:43 am by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==
Pages: 1 [2]   Go Up