1 2 Previous Next 65 Replies Latest reply on Jan 7, 2009 6:47 AM by (Leo_A._Geis)

    ACR INTERPOLATION UPSIZING

      ACR INTERPOLATION UPSIZING (is it better?)

      I have been off the subject for a couple years but the controversy seems to be about the same. I am a professional, high end, fine art, landscape photographer/print maker and I really want to ensure that my interpolation method is as BEST AS IT CAN BE for very large, high resolution gallery prints (usually the Canon 1DS Mark II or III). I did a very comprehensive research effort last time and basically found a lot of CONTRADICTORY info out there. I then tried all the up-sizing methods and then settled on PS Bicubic SHARPER (a tip by a very reputable fine artist/photographer and also mentioned in a KELBY book). Yes, I know this is unorthodox and controversial!

      As I have gone through the database here on this forum I have all kinds of contradictory answers.

      Some say ACR interpolation creates artifacts.

      Some say it is the SAME as PS Bicubic Smoother.

      I have also read DEKE say it is a "more sound algorithm" to interpolate in ACR at the RAW stage. This I also have found a few saying; that is is best done in ACR for quality, but the improvement is only slight (slight is good for me).

      I am trying to produce some of the best, large, fine art landscape prints I can with my equipment (printed on Fuji Flex paper and printed on either a Durst Lambda or an Oce Lightjet) and I now want to re investigate this subject and get down to the bottom of it (again).

      So... please throw out anything you know here! Is there a benefit to up-sizing in ACR, in terms of sheer quality large prints? Maybe JEFF SCHEWE would even throw in his perspective (again)?

      Thank you for whoever participates in this beat to a pulp controversy. I just want to find out, if in two years, the subject has settled, or the answers changed?
        • 1. Re: ACR INTERPOLATION UPSIZING
          Level 1
          I just read an interesting article that said two key things (at least I think they might be key):

          "The two new methods Bicubic Smoother and Bicubic Sharper are designed to minimise the amounts of sharpening artifacts (smoother) or loss of detail (sharper).

          The up-sampling in ACR can produce more artifacts when used, as opposed to Bicubic Smoother. The example below is from a Canon 1Ds raw file."


          Maybe the true answer for me is a bunch of test printing...
          • 2. Re: ACR INTERPOLATION UPSIZING
            Level 1
            > Maybe the true answer for me is a bunch of test printing...

            If you do, please come back and share your results. I don't do fine art
            printing, but anything you discover would be of great use to me.

            -X
            • 3. Re: ACR INTERPOLATION UPSIZING
              Level 1
              Well, I just did one. I took an image I shot that had a TON of high and mid frequency detail in it (aspen trees). I shot it with the Canon 1DS Mark II last fall.

              One image I interpolated as big as I could in ACR (because of the composition, it was cropped in ACR to 12.8 native mega pixels and then I interpolated it to 19.4 in ACR). Then, in PS, I converted it from Adobe RGB to LAB mode and applied a 1.0 radius 500 amount 0 threshold in USM, on the L channel (it looked good that way).

              I then brought another copy of the same image into PS but at the native resolution (12.8MP) and then interpolated it up (to the exact same size as the other) using my favorite Bicubic SHARPER. I then applied the same exact sharpening in LAB.

              I slapped the two copies on top of each other and carefully went over the image at different zooms (200% seemed to work the best). It took a very careful eye (I do have 20/20 vision) but to my surprise (and sadly to say) my Bicubic SHARPER was slightly LESS defined! It did seem to have slightly stronger halos in a very few, very strong high contrast edge areas (where halos seem to show themselves the most) but overall it slightly LACKED DEFINITION or what I would call a "tightness of the detail".

              I really had to look the ENTIRE image over, but I'm now convinced that ACR beat Bicubic SHARPER.

              Next test will have to be PS Bicubic SMOOTHER, and the others...

              I'll keep everyone posted.

              Sorry about not posting examples. It was really looking over the entire image that made me see the differences (clicking on and off the layer eyeball). I do now advocate giving a test of your own (anyone who is interested in the issue).
              • 4. Re: ACR INTERPOLATION UPSIZING
                MadManChan2000 Adobe Employee
                For the record, Camera Raw's interpolation algorithm is different from both PS's Bicubic Sharper and Bicubic Smoother.

                I am not going to comment on the relative merits, because there are so many possibilities to consider. For example, one could argue that to do a valid comparison you should apply the same output sharpening to an image that has been resampled 2 different ways (which is what you've done). But it's also valid to argue that you should fine-tune the output sharpening based on the method of resampling.

                In general I would argue against using Bicubic Sharper for upsampling because of the likelihood of getting stronger halos over which you have no control. If you want to introduce slight halos for the purposes of compensating for ink bleeding, for instance, it is better done during the output sharpening process, under __your__ control (or under the control of software that understands the output process) as opposed to letting it be folded into an upsampling algorithm that has no idea what you're printing to (or even the fact that you're planning to print the thing at all).
                • 5. Re: ACR INTERPOLATION UPSIZING
                  Level 1
                  Excellent points here Eric!

                  For more accurate testing, each sampling method should require a different (even if slightly) custom sharpening (based on the specific needs of the image after interpolation) to bring out its best, so that should be factored into a "test." Makes accurate testing much more complicated...

                  Why allow uncontrollable halos (which Bicubic SHARPER seems to introduce stronger than the other PS interpolation methods)
                  when controlling halos (hopefully people here understand what sharpening is really doing in terms of causing mini halos to all edge contrast/detail areas) is really what sharpening is all about (when you sharpen masterfully you have utter control of the halos). Another strike against my, used to be, favorite way to up-size.

                  Thanks for pointing these things out Eric.
                  • 6. Re: ACR INTERPOLATION UPSIZING
                    Level 1
                    Bicubic has a small amount of sharpening as a component of the algorithm. Bicubic Smoother has none and Bicubic Sharper as a stronger sharpening component. Smoother is intended for upsampling and produces a more sharpen-able result than either of the other two algorithms.

                    Camera Raw uses a Lancoz upsampling that is not the same as any of the Bicubic algorithms. In general one would want to use Bicubuc Smoother then apply more sharpening carfeully than using Camera Raw's upsampling or Bicubic Sharper. Bicubic Sharper is really a lazy way of upsampling while avoiding the whole sharpening issue. Bicubic Sharper was designed for downsampling because it needs a higher level of sharpening.

                    I suggest you read this and come back with questions... The Art Of The Up Res
                    • 7. Re: ACR INTERPOLATION UPSIZING
                      Bill_Janes Level 2
                      >I suggest you read this and come back with questions

                      That is a good article on upresing, but it makes no mention of deconvolution sharpening (e.g. Richardson-Lucy) or specialized upresing such as Genuine Fractals.

                      I haven't used either method, but Roger Clark claims that one can effectively double the camera reslution with the Richardson-Lucy algorithm.

                      http://www.clarkvision.com/imagedetail/image-restoration1/index.html
                      • 8. Re: ACR INTERPOLATION UPSIZING
                        Level 1
                        >deconvolution sharpening

                        Fine if you can work out the motion blur that you are trying to overcome. It's designed not for "general sharpening" but to undo certain kinds of blur...

                        >Genuine Fractals

                        I got a bridge in the desert I'll sell ya...
                        • 9. Re: ACR INTERPOLATION UPSIZING
                          Level 1
                          Very helpful! Thank you very much. It seems a couple years ago this was much more controversial than now. It seems people have been getting down to the answers. I will modify my interpolation methods to keep them on the cutting edge.

                          I read the artical. A very good artical. It did have a few new things I enjoyed learning. Thank you much Jeff.
                          • 10. Re: ACR INTERPOLATION UPSIZING
                            Bill_Janes Level 2
                            >deconvolution sharpening

                            >Fine if you can work out the motion blur that you are trying to overcome. It's designed not for "general sharpening" but to undo certain kinds of blur...

                            If you had taken the trouble to look at Roger Clark's example in the use of the adaptive Richardson-Lucy iteration, you would have seen that it has other uses than correction of motion blur as you hint at in your reply. Strictly speaking, the R-L algorithm is not a sharpening algorithm but a restoration algorithm. It can actually recover lost detail, whereas the sharpening algorithms merely create the illusion of more detail. However, a point spread function (PSF) describing how an image was blurred is necessary to apply the deconvolution, and this is sometimes difficult to derive. NASA used R-L to correct the spherical aberration present in the early Hubble telescope images. Raw Developer also uses the R-L algorithm, but I don't use a Mac and have not tried it out.

                            http://www.iridientdigital.com/products/rawdeveloper.html

                            As you correctly note, the Smart Sharpen in Photoshop uses deconvolution. It allows one to choose between three PSPs : Gaussian Blur, Lens Blur and Motion Blur. Focus Magic also uses deconvolution to undo the effect of defocus and Jonathan Wienke uses it for capture sharpening with good results:

                            http://luminous-landscape.com/forum/index.php?showtopic=24924&view=findpost&p=191731

                            As described by Roger and in the link below, the PSP is sometimes not critical and good results can be obtained empirically.

                            http://www.flickr.com/photos/jmzawodny/312509696/
                            • 11. Re: ACR INTERPOLATION UPSIZING
                              Level 1
                              >If you had taken the trouble to look at Roger Clark's example in the use of the adaptive Richardson-Lucy iteration, you would have seen that it has other uses than correction of motion blur as you hint at in your reply.

                              Uh huh. . .well, I _DID_ take the time to read Roger's article and noted that it was done in 2005 and ONLY compared Photoshop USM (which I rarely if ever use these days) with ImagesPlus Adaptive Richardson-Lucey algorithm. It ain't a magic bullet...upsampling and trying to make something out of nothing is always going to be a challenge. Far better to start with something shot well with a pixel count appropriate for the intended use. Yes, you can almost always do a successful 2X upsample on good digital captures and even up to 4X is the original capture is technically excellent.

                              As you note, using a deconvolution algorithm is designed more for correcting image defects than for correcting images that don't suffer from such defects. Without and accurate PSF, you are just using an alternative method that may or may not be better than other methods. Trying to correct camera movement or subject movement is really trying to make lemon aid out of lemons. Yes, you may be able to improve some certain aspects and the improvement might be beneficial, but there are other ways of improving image quality like using a shutter speed equal to freesing subject motion or a tripod to eliminate camera shake (heck, even popping the ISO up a couple of notches and using good noise reduction can help).

                              You seem to be a great advocate of deconvolution...but do you use it? Or do you just find it "interesting"?
                              • 12. Re: ACR INTERPOLATION UPSIZING
                                Ramón G Castañeda Level 4
                                I did a series of tests comparing the most recent versions of Genuine Fractals and Photoshop's Bicubic smoother. Photoshop wins easily.

                                There was a time, before Bicubic Smoother, when Genuine Fractals had an edge. That edge now belongs to Photoshop.
                                • 13. Re: ACR INTERPOLATION UPSIZING
                                  Level 1
                                  OK, I finally did my first ACR interpolation VS PS Bicubic SMOOTHER interpolation "test" (previously posted the ACR VS Bicubic SHARPER results on this thread)...

                                  Clearly not a scientific test, but very interesting (to me) for sure.

                                  I took that same aforementioned (in this thread) image that has a ton of high to mid frequency areas (aspen trees in Glacier NP). It was cropped in ACR to a native 12.8 MP (a Canon 1DS Mark II file) and interpolated it as high as ACR would go to 19.4 MP and brought it into PS as an Adobe RGB file where it was converted to LAB and then sharpened with a global USM on the L channel at 500 amount, 1.0 Radius, and 0 amount (the image uniquely looked pretty good that way).

                                  I then brought the native resolution 12.8 shot into PS and interpolated it to 19.4 MP using PS Bicubic SMOOTHER, then did the same LAB conversion and LAB sharpen.

                                  I then slapped these on to each other as separate layers and looked them over very carefully in many different frequency areas.

                                  I did this same test with the UNSHARPENED versions as well.

                                  Interesting!

                                  Still quite a noticeable difference at 200% zoom (but not nearly as much as the difference between ACR and Bicubic SHARPER). To my eye, the ACR image still was a tad "tighter" so to speak. In the sharpened versions, there was a noticeable "ultra fine detail" almost like an "ultra fine sharpening - minus any sharpening halos" in the ACR version (in the entire image). Almost like a very, very slight graininess, but not a "graininess" I would consider negative or a minus to an images overall quality. Maybe the type of "graininess" that would make an image appear less "digital" if you know what I mean. Although the "graininess" was SO fine I'm not sure it would make any noticeable appearance in a print (after the averaging that takes place turning pixels into a continuous tone print). I do wonder though, if it might be a touch more ultra fine DETAIL being preserved? I also wonder if the increased "tightness" would give the appearance of a slightly more detailed print in high to mid frequency areas?

                                  In the UNSHARPENED versions the ACR image looked significantly "tighter" with a slight more ultra fine detail, but ONLY in the higher frequency areas. To my eye the mid and low frequency areas looked identical to the SMOOTHER version.

                                  In the SHARPENED versions I did try to mimic the ultra fine detail (found in the ACR version) by adding just a bit more sharpening at the lowest radius possible on the L channel of the BICUBIC SMOOTHER version, but nothing I could do would make it look the same as the ACR version. Although I could get it fairly close, there was a slight increase of edge halos on the SMOOTHER version caused by the sharpening, that were not present in the ACR version.

                                  Now I know this is not an exhaustive or scientific test whatsoever, but if I HAD to make my own personal conclusions or judgement based on this test alone... it would lead me to believe that I should interpolate using ACR as much as I can. Then If I needed an even bigger print than ACR can produce, I could then take the interpolation further in PS using Bicubic SMOOTHER. I also would guess that on large prints the overall print difference would be slight but noticable to a careful eye. My guess would be that small to mid sized prints may not make any discernible difference at all (say up to 18 or so inches).

                                  Maybe more personal tests will are in order!
                                  • 14. Re: ACR INTERPOLATION UPSIZING
                                    Bill_Janes Level 2
                                    >ONLY compared Photoshop USM (which I rarely if ever use these days)

                                    Isn't USM with blend-if and masks the basis of Bruce Fraser's capture sharpening workflow along with the High Pass filter and the overlay blending mode for output sharpening? What do you use in place of USM for capture sharpening?

                                    >Far better to start with something shot well with a pixel count appropriate for the intended use.

                                    That's true, but with a Bayer array camera, blur filter, and the approximations introduced by demosaicing, some sharpening or image restoration is always necessary, so why not try to use the best method?

                                    >You seem to be a great advocate of deconvolution...but do you use it? Or do you just find it "interesting"?

                                    I find deconvolution interesting, but to tell the truth I haven't figured out how to use the tool to best advantage in practical field photography. The main problem is determining the PSP function parameters. With motion blur, one can often determine the blur axis and distance by visual inspection of the image, and deconvolution is the best method to improve such images. With defocused images, the amount of defocus can often be estimated by inspection of point sources in the image, and the image can be improved to some extent with deconvolution.

                                    The USM filter was a pretty crude tool until Bruce and others taught us how to use it, and the same could be true for deconvolution methods.
                                    • 15. Re: ACR INTERPOLATION UPSIZING
                                      Level 1
                                      > What do you use in place of USM for capture sharpening?

                                      Camera Raw 4.5 or Lightroom 2.0.

                                      See, Bruce had a little something to do with the sharpening that was put into Camera Raw 4.1 and Lightroom 1.1. While he didn't get to finish the work, I did. Same deal with output sharpening...I print out of Lightroom 2.0 with it's Bruce inspired output sharpening.

                                      If you know how to use it (and it's really not all that hard) Camera Raw can do a better job of capture sharpening than PhotoKit Sharpener can. You might consider Camera Raw 4.1 and above to be PhotoKit Sharpener 2.0 (although that will come from PixelGenius in time).
                                      • 16. Re: ACR INTERPOLATION UPSIZING
                                        Level 1
                                        >Maybe more personal tests will are in order!

                                        Yeah, like learning how to use Camera Raw's sharpening capability...
                                        • 17. Re: ACR INTERPOLATION UPSIZING
                                          Ramón G Castañeda Level 4
                                          <u><b>OFF TOPIC:</b></u><br /> <br /> Since you mention <b>Pixelgenius</b>, do they have another email address (other than <community-support@pixelgenius.com>)?<br /> <br /> I have had no responses for my requests for help accessing the forum for about a month.  Below the text of my last email:<br /><br /><span style="font-size:0.75em"><blockquote>All of a sudden the login didn't work, so I requested automated help.  But<br /><br />The new user name and new password are rejected by the site.<br /><br />My user name (until tonight) had always been "Ramón G Castañeda".  In your email quoted below, it has been changed by the forum software (I assume) to "Ram?n G Casta?eda".<br /><br />I've tried it both ways.  It does NOT work.<br /><br />Please advise. </blockquote></span><br /> <br /> Thanks in advance,
                                          • 18. Re: ACR INTERPOLATION UPSIZING
                                            Bill_Janes Level 2
                                            >What do you use in place of USM for capture sharpening?

                                            >Camera Raw 4.5 or Lightroom 2.0.

                                            And isn't USM the backbone of ACR capture sharpening just as it was in Bruce's capture sharpening? If so, then you are still using USM.
                                            • 19. Re: ACR INTERPOLATION UPSIZING
                                              MadManChan2000 Adobe Employee
                                              Hi Bill, CR/LR's capture sharpening has evolved a bit more away from the good ol' USM that's been in PS for a long time. Certainly some of the basic principles of edge sharpening still apply, but there are a lot of differences as well (particularly with regards to the Detail slider) and built-in protections to minimize sharpening of undesirable areas.

                                              (This is not to say that CR/LR has perfect capture sharpening or can't be used to produce ugly halos. There's always room for improvement.)
                                              • 20. Re: ACR INTERPOLATION UPSIZING
                                                01af Level 1
                                                Mark Metternich wrote:
                                                > Some say ACR interpolation creates artifacts.

                                                It really does. See here: http://luminous-landscape.com/forum/index.php?showtopic=27937&view=findpost&p=221560

                                                .
                                                Mark further wrote:
                                                > Some say it is the SAME as PS Bicubic Smoother.

                                                The algorithms are different but the results are virtually identical ... except Camera Raw's ringing artifacts which are absent in Photoshop's Bicubic Smoother.

                                                .
                                                Mark further wrote:
                                                > So ... please throw out anything you know here!
                                                > Is there a benefit to up-sizing in ACR, in terms
                                                > of sheer quality large prints?

                                                No! Just to the contrary, there's a penalty. Avoid upsizing in Camera Raw when you're after the best image quality possible! I keep getting the best results this way:

                                                1) Convert to RGB using Camera Raw at native size and without any
                                                i capture sharpening.

                                                2) Import the RGB image into Photoshop and upsize there using Bicubic Smoother.

                                                3) Apply
                                                i sharpening for source
                                                using Bruce Fraser's method. Also apply
                                                i sharpening for content
                                                if required. By the way,
                                                i sharpening for source
                                                and
                                                i sharpening for content
                                                combined is called
                                                i capture sharpening.

                                                4) Apply other post-processing tasks as required.

                                                5) Apply
                                                i sharpening for output
                                                and print.

                                                -- Olaf
                                                • 21. Re: ACR INTERPOLATION UPSIZING
                                                  Level 1
                                                  > If so, then you are still using USM.

                                                  Well, considering that Camera Raw is working in linear gamma (which DOES really help with sharpening) and the fact that the Detail slider is a bi-spacial frequency adjustment between high and medium frequencies and acts as a halo suppression and the fact you can combine an on-the-fly adjustable edge mask, I wouldn't call Camera Raw's sharpening USM. The only time it's even similar to USM is when the Detail slider is at 100something I rarely do.

                                                  Nope, it ain't plain old USM. It's an exotic hybrid built in the minds of Thomas Knoll and Mark Hamburg and following Bruce's concepts of capture sharpening.
                                                  • 22. Re: ACR INTERPOLATION UPSIZING
                                                    Level 1
                                                    Thank you very much Olaf Ulrich

                                                    I don't want to get into sharpening too much here (this thread is mainly being about comparing the various interpolation methods). But I have been trying to get to those ACR vs PS answers, in relationship to sharpening, in the other ACR thread "How much to do in ACR vs PS?"

                                                    Your info here is basically, more or less, what I have come to believe too. But in the other thread where we are hitting on sharpening in ACR, folks are trying to tell me "capture sharpening" in ACR prior to uprezzing a file massively in PS will produce a superior result. I'm 100% fine with that if it is true, but I wonder where this is comprehensively explained? Maybe it is in the new ACR book, which I will get and read soon. If you would like to throw your voice into the mix on that other thread, feel free. It might benefit us all.
                                                    • 23. Re: ACR INTERPOLATION UPSIZING
                                                      JimHess-8IPblY Level 3
                                                      Mark,

                                                      Do you realize who Jeff Shewe is?
                                                      • 24. Re: ACR INTERPOLATION UPSIZING
                                                        Level 1
                                                        Hi Jim.

                                                        Yes, I know who Jeff is. I think he is simply brilliant.

                                                        I'm not trying to be disrespectful here, just trying to get answers. I'm attempting to find the best resources that satisfy questions I have.
                                                        • 25. Re: ACR INTERPOLATION UPSIZING
                                                          Bill_Janes Level 2
                                                          Eric Chan wrote:

                                                          >CR/LR's capture sharpening has evolved a bit more away from the good ol' USM that's been in PS for a long time. Certainly some of the basic principles of edge sharpening still apply, but there are a lot of differences as well (particularly with regards to the Detail slider) and built-in protections to minimize sharpening of undesirable areas.

                                                          Jeff Schewe wrote:

                                                          >that the Detail slider is a bi-spacial frequency adjustment between high and medium frequencies and acts as a halo suppression and the fact you can combine an on-the-fly adjustable edge mask, I wouldn't call Camera Raw's sharpening USM. The only time it's even similar to USM is when the Detail slider is at 100something I rarely do.

                                                          Perhaps this is going a bit off the topic of this thread, but Eric and Jeff raise some interesting points. Bruce Fraser's capture sharpening is readily understood. It uses the USM (modified by amount, radius, and threshold) together with layer blend-if to protect the high and low luminance pixels and an edge mask to help restrict sharpening to the edges.

                                                          I had assumed that ACR's Amount and Radius sliders to operate on an USM or similar convolution based algorithm, but this may not be the case. The mask seems to be analogous to Bruce's method and I don't know where the blend-if operation occurs in ACR sharpening.

                                                          The operation of the Detail slider is somewhat puzzling. Perhaps its methodology is proprietary, but can one of you give any indication of how it works?
                                                          • 26. Re: ACR INTERPOLATION UPSIZING
                                                            Level 1
                                                            >but can one of you give any indication of how it works?

                                                            Well, I pretty much said it..."the Detail slider is a bi-spacial frequency adjustment between high and medium frequencies and acts as a halo suppression".

                                                            Now, what that means might be tougher to understand. Camera Raw since 4.1 and above incorporated multi-frequency sharpening via the Detail slider. The main radius setting is used when Detail is set to 100%. As you lower the Detail setting you taper off the high frequency radius and move toward the medium frequency (which for the life of me I can't remember the factor). But the effect is a halo suppression unlike anything in a standard USM (no, it's not like threshold).

                                                            As far as the blend-if, the shadows and highlights are rolled off (not unlike PhotoKit Sharpener). Since the sharpening is being done in linear, the numbers don't really relate at all however. In point of fact, the radius setting in Camera Raw isn't like the radius in USM either because of the linear space. I believe Hamburg stated that the difference is a factor of 1.4 or 2 (can't remember which) so that means .5 in Camera Raw is more like .35 or .25 radius.

                                                            Granted, Bruce was party to the consulting done for Camera Raw 4.1 even if he wasn't around for the final algorithms (our loss) but bringing Thomas Knoll (who advocated PhotoKit Sharpener's logic) and Mark Hamburg into the mix had a major impact on designing a more modern set of algorithms which you can't even do in Photoshop. Which is why I do advocate using Camera Raw's sharpening as a capture sharpening process in combination with proper noise reduction (which is intermingled with the sharpening) as well as CA correction.
                                                            • 27. Re: ACR INTERPOLATION UPSIZING
                                                              Level 1
                                                              A hypothetical question:

                                                              I have a Canon 1DS Mark II CR2 file (landscape photo). It was shot with very good glass, tripod, mirror lock up, great exposure, the works...

                                                              The image is almost exclusively very detailed, high frequency to mid frequency and evenly balanced between these frequencies. The image also has a lot of color variation. The image has no noise problems at all.

                                                              The image is to become a 40 inch, 300ppi, Adobe 1998, Tif file printed on Fuji Flex (poly) using an Oce Lightjet.

                                                              The image will be rezzed up from the native 16.6 MP to 98 MP in PS using Bicubic Smoother (not that this is the ideal ppi or size).

                                                              Careful advanced custom layer USM sharpening will be applied to the image in LAB mode as the last comprehensive steps in work flow before converting to Adobe RGB, converted to 8 bit, and sending the file to print.

                                                              Question #1:

                                                              In the above scenario what would be a general ideal way or guideline to approach capture sharpening in ACR (with some general setting ideas) before such a massive interpolation, to produce the absolute best possible quality print?

                                                              Question#2:

                                                              What multiple resources would you refer me to, to understand the ideal way to capture sharpen in ACR so that I can produce a higher quality print than only using USM as a last step?

                                                              Thank you beforehand anyone who can help me with these questions.
                                                              • 28. Re: ACR INTERPOLATION UPSIZING
                                                                Level 1
                                                                >In the above scenario what would be a general ideal way or guideline to approach capture sharpening in ACR

                                                                Adjust the image at 1:1 to look "good"...the odds are you'll want 40-60 amount, radius .6-.8 and detail between 25 and 50 with a layer mask in the 10-20 range. Then upsample and then do additional sharpening (I would NEVER go to Lab except to do certain color recovery steps that can only be done in Lab). After getting the image to the desired output size I would do output sharpening for the media. I use Lightroom for most print output otherwise I would use PhotoKit Sharpener's output sharpening.

                                                                Are you referring to 40" on the longest side? If so, that's only a 240% increase in size which really shouldn't be that hard. I personally would do some Super Sharpening (from PhotoKit Sharpener) and at the end of the process (before output sharpening) I would add a touch of photo grain (also from PKS). And I've done this and larger output on a regular basis.

                                                                But, the 8 bit Adobe RGB kinda worries me...I would be keeping it in Pro Photo RGB 16 bit up to the point where I transform it for the final output profile...you do have a custom profile for the output, right? And this business of converting to Lab for a well exposed good RGB image isn't something I would stoop to, not if the image was a good full range color image to start with.
                                                                • 29. Re: ACR INTERPOLATION UPSIZING
                                                                  Level 1
                                                                  Thanks very much Jeff, this gives me a lot to think about and something to work with and ultimately compare to what I am already doing. I very much appreciate you taking the time. I will certainly try this as a base.

                                                                  I will add just a few comments to your reply:

                                                                  Adjust the image at 1:1 to look "good"...the odds are you'll want 40-60 amount, radius .6-.8 and detail between 25 and 50 with a layer mask in the 10-20 range. Then upsample and then do additional sharpening (I would NEVER go to Lab except to do certain color recovery steps that can only be done in Lab).

                                                                  I have found that I can often get outstanding color variation/separation in LAB compared to RGB by allowing for extra room in the file to increase the saturation using Dan Margulis's "Man from Mars" technique. I have done a lot of testing and quite often the results are nothing shy of mind boggling. I have examples to show.

                                                                  After getting the image to the desired output size I would do output sharpening for the media. I use Lightroom for most print output otherwise I would use PhotoKit Sharpener's output sharpening.

                                                                  So is there something in Lightroom that is not in PS CS3 that you would use, that you could not do in CS3? Can you get the desired results there?

                                                                  Are you referring to 40" on the longest side? If so, that's only a 240% increase in size which really shouldn't be that hard.

                                                                  You are right and I do upsizing with great results much bigger, I just wanted to throw a size out there that would not distract from the issue by being controversial.

                                                                  I personally would do some Super Sharpening (from PhotoKit Sharpener) and at the end of the process (before output sharpening) I would add a touch of photo grain (also from PKS). And I've done this and larger output on a regular basis.

                                                                  I'm tempted to look into these products.

                                                                  But, the 8 bit Adobe RGB kinda worries me...I would be keeping it in Pro Photo RGB 16 bit up to the point where I transform it for the final output profile...

                                                                  Yes 16 bit and all that, but why Pro Photo RGB? Or if you do not want to open up that colorspace can of worms, where should I read about why that would be a better approach, rather than Adobe RGB?

                                                                  You do have a custom profile for the output, right?

                                                                  Yes.

                                                                  And this business of converting to Lab for a well exposed good RGB image isn't something I would stoop to, not if the image was a good full range color image to start with.

                                                                  Again, after reading Dan's book "Photoshop LAB Color" and doing a bunch of tests, I came to absolutely love the often dramatic power of the color variation/separation that I can easily get there. I can not seem to duplicate it in RGB, let alone ACR. In landscape photography, I have come to consider it having an "edge" of sorts.
                                                                  • 30. Re: ACR INTERPOLATION UPSIZING
                                                                    Level 1
                                                                    >Yes 16 bit and all that, but why Pro Photo RGB? Or if you do not want to open up that colorspace can of worms, where should I read about why that would be a better approach, rather than Adobe RGB?

                                                                    Because most modern output devices can print colors not contained in even Adobe RGB. Oranges, yellows and reds for sure. Look at a gamut plot of Adobe RGB vs an EPson K3 printer (such as the 3800 or 7880) and see all the color that the print can print that the ARGB color space clips. As for the Lightjet...I'm not sure but if you have a profile you can compare it in Color Sync Utility or ColorThink...

                                                                    Personally I would not use Adobe RGB because output devices will only grow larger in gamutI just got a beta unit of the new EPson 7900 (which I _CAN_ talk about) and will be doing profiles this week. It has a 10 color ink set including orange and green and I already know that Adobe RGB would _NOT_ be able to contain those colors...so, why limit the gamut of final archive RGB images to the relatively paltry size of Adobe RGB for the future?
                                                                    • 31. Re: ACR INTERPOLATION UPSIZING
                                                                      Ramón G Castañeda Level 4
                                                                      The late, lamented Bruce Fraser pointed me in the direction of 16-bits ProPhoto RGB years ago, and that has been the starting point of my workflow ever since. Jeff Schewe has provided many more valuable insights.

                                                                      Now, Jeff can't engage in straight self promotion, but as a totally uninterested, unbiased observer, I'll do it for him: :D

                                                                      For the price of a decent, modest lunch, you can download and view an excellent series of QuickTime videos titled From Camera To Print - Fine Art Printing Tutorial from The Luminous Landscape.
                                                                      • 32. Re: ACR INTERPOLATION UPSIZING
                                                                        01af Level 1
                                                                        In this thread's post #20 above, I recommended
                                                                        i not
                                                                        to use Camera Raw's sharpening feature for images that are bound to get upsized. I just want to make it clear that I like Camera Raw's sophisticated sharpening (after some familiarization) and I'm using it all the time---except when upsizing. My tests (not very scientifical) seem to indicate that sharpening
                                                                        i after
                                                                        upsizing yields better results than upsizing a pre-sharpened image. That's why the shortcomings of Camera Raw's upsizing algorithm force me to abandon Camera Raw's sharpening for images that are going to get upsized in Photoshop.

                                                                        Now Jeff recommends to capture-sharpen first (i. e. in Camera Raw) and then to upsize (and to add some more shapening after the upsizing). Ummm. Seems that means more testing for me ...

                                                                        Jeff, would you capture-sharpen an image for upsizing just the same way as an image that's not supposed to get upsized?

                                                                        -- Olaf
                                                                        • 33. Re: ACR INTERPOLATION UPSIZING
                                                                          Bill_Janes Level 2
                                                                          >Yes 16 bit and all that, but why Pro Photo RGB? Or if you do not want to open up that colorspace can of worms, where should I read about why that would be a better approach, rather than Adobe RGB?

                                                                          The critical factor here is does aRGB contain the full gamut of your raw image without clipping? If so, there is no need to use ProPhotoRGB, and aRGB would offer better gradation for a given bit depth. Pages 13 and 14 of the Camera Raw book by Jeff and Bruce Fraser is a good place to start your reading. Many rather innocuous scenes contain colors outside the gamut of aRGB, and it is best to use ProPhoto if you standardize on one color space for your work flow.
                                                                          • 34. Re: ACR INTERPOLATION UPSIZING
                                                                            MadManChan2000 Adobe Employee
                                                                            Bill, I would revise your description by saying that the optimal color space would be the smallest one that holds the final gamut of your edited images (as opposed to the gamut of the raw image; many folks prefer to add saturation to their image, i.e., artifically expand the color range).

                                                                            That said, the tradeoff is that it's less convenient to keep switching working/output spaces on an image-by-image basis. So one way to look at the issue is: should you (1) stick with a larger space like ProPhoto for convenience's sake, with the understanding that many images may fit into a smaller space like Adobe RGB or sometimes even sRGB or (2) choose a working space on an image-by-image basis to optimize the precision, even though it's less convenient to do so.

                                                                            In my opinion, it's far better to go with #1 because as long as you're in 16 bits it's very unlikely you will __perceive__ quantization artifacts. In contrast, if you go with #2 it's highly likely you will __not perceive__ smoother tonal gradations but it __is__ much more likely that you'll become more annoyed by the need to keep switching working/output spaces for each image.

                                                                            From a user's point of view, given the way CR/LR is designed, you may as well stick with ProPhoto RGB as a working space because CR/LR internally uses RIMM (which has PP RGB primaries).
                                                                            • 35. Re: ACR INTERPOLATION UPSIZING
                                                                              Level 1
                                                                              >
                                                                              > (1) stick with a larger space like ProPhoto for convenience's sake, with the understanding that many images may fit into a smaller space like Adobe RGB or sometimes even sRGB

                                                                              This is what I've decided to do in my workflow. My current problem WRT
                                                                              colorspace is a technique for getting consistent results when doing a
                                                                              'colorspace compression' from ProPhoto to sRGB in Ps when prepping for output to
                                                                              web or jpgs for a client. But that's no longer a CR problem :)

                                                                              -X
                                                                              • 36. Re: ACR INTERPOLATION UPSIZING
                                                                                MadManChan2000 Adobe Employee
                                                                                The real solution is intelligent, image-dependent gamut mapping. For now, this process is in the hands of users, rather than in the hands of the CMM.
                                                                                • 37. Re: ACR INTERPOLATION UPSIZING
                                                                                  Level 1
                                                                                  Conclusion:

                                                                                  After 2 all day tests on interpolation methods from ACR, to PS Bicubic Smoother, Bicubic, and Bicubic Sharper, I think I'll have to conclude (at least today) that the differences are so slight that we very well may be splitting the cells that make up hairs. I am not sure (yet) whether these ever so slight pixel variances would make any difference at all in a large print. Eventually I will try some costly print tests...

                                                                                  Having said that, I now have to add that I have come away with a renewed passion for re-mastering my already comprehensive sharpening workflow which IMO makes a HUGE difference in a print. Much, much more than interpolation methods.
                                                                                  • 38. Re: ACR INTERPOLATION UPSIZING
                                                                                    Level 1
                                                                                    OLAF

                                                                                    My tests (not very scientifical) seem to indicate that sharpening after upsizing yields better results than upsizing a pre-sharpened image. That's why the shortcomings of Camera Raw's upsizing algorithm force me to abandon Camera Raw's sharpening for images that are going to get upsized in Photoshop.

                                                                                    Now Jeff recommends to capture-sharpen first (i. e. in Camera Raw) and then to upsize (and to add some more shapening after the upsizing). Ummm. Seems that means more testing for me ...


                                                                                    That is exactly what I have been trying to get at, on that other thread. I'm sort of having to take it by faith that ACR capture sharpening is better in this upsizing application. I too would like to see some tests, and some real world proof. If it has already been done, then great! But if not, we'll have to do it ourselves.

                                                                                    Can anyone refer us to some results that confirm the ACR capture Sharpening is better in this regard?
                                                                                    • 39. Re: ACR INTERPOLATION UPSIZING
                                                                                      Tim Lookingbill Level 1
                                                                                      Why not just sharpen enough in ACR to eliminate the softness along edge detail until you start seeing artifacts like halos and noise?

                                                                                      I believe that's the best you can do.
                                                                                      1 2 Previous Next