14 Replies Latest reply on Sep 19, 2010 11:02 AM by ACRFREAK

    ACR INTERPOLATION Q

    ACRFREAK

      Simple Question (sorry I have not gotten the latest book yet - if the info is there).

       

      When I use a higher MP DSLR like the Canon 5D Mark II...  I only get one option to uprez (from 21. to 25.2) in ACR 6.1.

       

      I want to know if in this situation (choosing the 25.2), is it going to interpolate with the algorithm that is very similar to Bi-cubic Smoother?

       

      Since I am being open minded to the new information presented on this forum that interpolating in ACR with good Capture Sharpening settings, to the largest size, with the intent to push the boundaries of quality enlargements, that uprezzing in ACR can improve the quality... I want to do a series of tests on the subject.

       

      The test would be:

       

      1: (native 21MP) image, capture sharpened appropriately and then ACR interpolated to max size (25.2), and then in PS brought up larger to a 70 inch print using Smoother (200 ppi Lightjet Flex).

       

      2: (native 21MP) image, capture sharpened appropriately (the same way) and then brought in PS at native size, and then in PS brought up larger to the 70 inch print using Smoother (200 ppi Lightjet Flex).

       

      Of course output sharpening would be applied to both and the testing would look at both the image without output sharpening applied, then with.

       

      Thank you in advance for your feedback.

        • 1. Re: ACR INTERPOLATION Q
          Jeff Schewe Level 5

          ACRFREAK wrote:

           

          When I use a higher MP DSLR like the Canon 5D Mark II...  I only get one option to uprez (from 21. to 25.2) in ACR 6.1.

           

          I want to know if in this situation (choosing the 25.2), is it going to interpolate with the algorithm that is very similar to Bi-cubic Smoother?

           

          ACR is using an adaptive upsampling which is a combo of bicubic and bicubic smoother. As to whether or not the 2X upsample is using a pure bicubic or a hydbred of bicubib and bicubic smoother I can't tell you...the odds are it's a hybrid that can't be exactly duplicated in Photoshop since the adaptive part does an interpolation between both algorithms seamlessly.

          1 person found this helpful
          • 2. Re: ACR INTERPOLATION Q
            ACRFREAK Level 1

            Then would you say with some level of confidence, for my output of large fine art landscape prints (pushing sizes) that there may be some quality benefit of using the ACR interpolation?

             

            Since you said in another thread that the capture sharpening is applied AFTER the interpolation (thats what I thought you said) in ACR and now here that it might not be the Smoother-Like only interpolation, I am thinking, no.

             

            Thank you for your time!

            • 3. Re: ACR INTERPOLATION Q
              Jeff Schewe Level 5

              ACRFREAK wrote:

               

              ...now here that it might not be the Smoother-Like only interpolation, I am thinking, no.

               

               

              What I said is that it would be a hybrid of both bicubic and bicubic smoother–an interpolation that couldn't be exactly reproduce in Photoshop. For that reason, yes, I think the optimal place to do upsampling is in ACR.

              • 4. Re: ACR INTERPOLATION Q
                ACRFREAK Level 1
                function(){return A.apply(null,[this].concat($A(arguments)))}

                function(){return A.apply(null,[this].concat($A(arguments)))}

                Jeff Schewe wrote:

                 

                ACRFREAK wrote:

                 

                ...now here that it might not be the Smoother-Like only interpolation, I am thinking, no.

                 

                 

                What I said is that it would be a hybrid of both bicubic and bicubic smoother–an interpolation that couldn't be exactly reproduce in Photoshop. For that reason, yes, I think the optimal place to do upsampling is in ACR.

                 

                Sorry for the confusion, but I was under the assumption that I would WANT a Smoother-Like algorithm coming from ACR and NOT the hybrid, when pushing the boundaries of fine art print sizes?  So what you are saying is the Hybrid that would likely be the algorithm in this exact case (5D II  21MP  to 25.2MP) would, in your opinion (and choice) be the better option over just using Smoother in PS from a native file size?

                 

                If so, I'll give it a try.

                 

                Thanks.

                • 5. Re: ACR INTERPOLATION Q
                  Noel Carboni Level 7

                  Interesting to see you getting into the territory I've been mapping out...  As you know, I asked a similar question about upsampling.

                   

                  I think the real answer is that you have to try it yourself and see what results you can get.  And don't stop just by looking at the pixels output from ACR...  Take it to the next step and post-process your image(s) to see how much you can make of the results.

                   

                  One thing is certain:  If you have experience with past versions of ACR, you must unlearn some of what you have learned.  The 2010 process doesn't respond the same ways the prior process did.

                   

                  -Noel

                  • 6. Re: ACR INTERPOLATION Q
                    ACRFREAK Level 1
                    function(){return A.apply(null,[this].concat($A(arguments)))}

                    Noel Carboni wrote:

                     

                    Interesting to see you getting into the territory I've been mapping out...  As you know, I asked a similar question about upsampling.

                     

                    I think the real answer is that you have to try it yourself and see what results you can get.  And don't stop just by looking at the pixels output from ACR...  Take it to the next step and post-process your image(s) to see how much you can make of the results.

                     

                    One thing is certain:  If you have experience with past versions of ACR, you must unlearn some of what you have learned.  The 2010 process doesn't respond the same ways the prior process did.

                     

                    -Noel

                     

                    I agree, TESTING!

                     

                    Funny enough I just tested a very high frequency only image of a slot canyon wall (sandstone) shot with a top DSLR and top lens and top f/stop.  I pushed the 21MP image to 70 inches at 200 ppi.  One image was max interpolated in ACR, one was native size in ACR.  Both capture sharpened with 40 - 0.5 - 100 - 0.

                     

                    Once the images got into PS CS5 I used Smoother to go to the 70 inches on both.

                     

                    At 100% they look the same.

                     

                    At closer zooms, the opposite as I would suppose was true!  The image brought into PS at native size (no ACR interpolation) was absolutely noticeably more detailed!  I have no idea WHY.  And I don't suppose it will be the same with every image!

                     

                    But this is, so far, going against what I am hearing about ACR interpolation.

                     

                    It will be interesting to me to see how further testing goes on other images of a variety of types...

                    • 7. Re: ACR INTERPOLATION Q
                      Noel Carboni Level 7

                      Interesting.

                       

                      There are no doubt many different ways to get there.

                       

                      When you think big picture, last year we were happy with the 2003 process.  This year we have the improved 2010 process and various different things to try, and our images are noticeably better than they were last year even without perfecting every step.  Cool stuff.

                       

                      -Noel

                      • 8. Re: ACR INTERPOLATION Q
                        Bill_Janes Level 2

                        Noel Carboni wrote:

                         

                        Interesting to see you getting into the territory I've been mapping out...  As you know, I asked a similar question about upsampling.

                         

                        I think the real answer is that you have to try it yourself and see what results you can get.  And don't stop just by looking at the pixels output from ACR...  Take it to the next step and post-process your image(s) to see how much you can make of the results.

                         

                        One thing is certain:  If you have experience with past versions of ACR, you must unlearn some of what you have learned.  The 2010 process doesn't respond the same ways the prior process did.

                         

                        -Noel

                        Noel,

                         

                        You are quite correct that Process 2010 is quite different from Process 2003. Upsampling is a major concern by those who arr making large prints, but with current high MP dSLRs, downsampling may be needed more often than upsampling for the most common printing tasks. To avoid alaising in downsampling, one  must first apply a low pass filter such as the sinc filter. Photoshop's downsizing was criticized in a recent LuLa thread:

                         

                         

                        http://www.luminous-landscape.com/forum/index.php?topic=40687.0

                         

                        My question is how does downsampling in ACR compare to that of Photoshop.

                        • 9. Re: ACR INTERPOLATION Q
                          Noel Carboni Level 7

                          Please let me ask a leading question:

                           

                          Assuming we're optimizing for best print quality here...

                           

                          For printing, which often sports dpi values in the ten-thousanths of an inch with modern hardware, why wouldn't you just upsample to the nearest multiple of the "best" or "native" ppi (e.g., 600 or 1200)?  I can't help but think that the print would only improve - if microscopically - provided the input data was good.  The human eye and brain have a wonderful ability to "feel" quality that can't be directly seen.

                           

                          Modern 64 bit computers certainly seem up to the task of crunching through the larger datasets.

                           

                          -Noel

                          • 10. Re: ACR INTERPOLATION Q
                            Bill_Janes Level 2

                            Wouldn't that merely pass off the task of downsizing to the printer driver rather than doing it with the processing software? And, according to Fraser and Schewe, passing more then 720 ppi to Epson printers can actually degrade print quality.

                            • 11. Re: ACR INTERPOLATION Q
                              ACRFREAK Level 1
                              function(){return A.apply(null,[this].concat($A(arguments)))}

                              function(){return A.apply(null,[this].concat($A(arguments)))}

                              Noel Carboni wrote:

                               

                              Please let me ask a leading question:

                               

                              Assuming we're optimizing for best print quality here...

                               

                              For printing, which often sports dpi values in the ten-thousanths of an inch with modern hardware, why wouldn't you just upsample to the nearest multiple of the "best" or "native" ppi (e.g., 600 or 1200)?  I can't help but think that the print would only improve - if microscopically - provided the input data was good.  The human eye and brain have a wonderful ability to "feel" quality that can't be directly seen.

                               

                              Modern 64 bit computers certainly seem up to the task of crunching through the larger datasets.

                               

                              -Noel

                               

                              In my case I use the Lightjet on Fuji Flex paper.  On the machine I use I can print a file that comes in at a max res of 305 ppi / Tiff / 8 bit.  With a top end DSLR and lens and technique, this works great up to about 24 - 40 inch prints (to my eye) up-sampling, depending on the image content.  At some undefined point around the (maybe) 30 - 40 inch range depending on the image, the images actually look sharper/better detail when then submitting 200 ppi images printed at the 200ppi setting (still interpolating).

                              • 12. Re: ACR INTERPOLATION Q
                                tshirley3

                                I have a queston that goes along with the original question.

                                When using a camera like the Canon 5D Mark II you only get one size larger to upsample Why is this. If a smaller sensor camera gives you

                                several why can't this be done for the 21 mega pixel cameras. I'm also assume the native resolution would be the sharpest or is it something

                                you wouldn't notice. One person I know always opens all their files at the hightest size.

                                 

                                I also heard somewhere that CS5 and Camera Raw use the same math to resample and so it doesn't matter where you do it.

                                • 13. Re: ACR INTERPOLATION Q
                                  Noel Carboni Level 7

                                  I'm one such person (who converts to the highest resolution possible in ACR).  I've already requested that higher image sizes be added in post 562 of the "Camera Raw Feature Requests" thread:

                                   

                                  http://forums.adobe.com/thread/311487?start=562

                                   

                                  You might consider adding your request to that thread as well.

                                   

                                  -Noel

                                  • 14. Re: ACR INTERPOLATION Q
                                    ACRFREAK Level 1
                                    function(){return A.apply(null,[this].concat($A(arguments)))}

                                    tshirley3 wrote:

                                     

                                    I have a queston that goes along with the original question.

                                    When using a camera like the Canon 5D Mark II you only get one size larger to upsample Why is this. If a smaller sensor camera gives you

                                    several why can't this be done for the 21 mega pixel cameras. I'm also assume the native resolution would be the sharpest or is it something

                                    you wouldn't notice. One person I know always opens all their files at the hightest size.

                                     

                                    I also heard somewhere that CS5 and Camera Raw use the same math to resample and so it doesn't matter where you do it.

                                     

                                     

                                    I have wondered that too.

                                     

                                    Now after my own personal testing (and do your own) on big enlargements I believe I am seeing better results with Bicubic Smoother than ACR (as apposed to what has been implied).

                                     

                                    It is not the same math, because if you do the tests you will see differences.