27 Replies Latest reply on Feb 1, 2008 11:46 AM by Bill Lamp

    Need for 3rd-party Noise Reduction Software?

      After reading Real World's Camera Raw for CS3 by Bruce Fraser and Jeff Schewe, and what was written about it's noise reduction capabilities, I'd like to ask if it's still necessary to get a 3rd-party noise reduction plug-in/software like Nik's Dfine 2.0 or Noiseware or NoiseNinja?

      Any feedback or thoughts would be greatly appreciated.

      Thanks.
        • 1. Re: Need for 3rd-party Noise Reduction Software?
          Level 1
          For lower ISO shots, I tend to do it in Camera Raw...but for higher ISOs (or images that have required LOTS of lightening) I tend to do it in NoiseWare. So, to answer your question needs more info. As I said in the book, there are times when you need heavy duty noise reduction and you'll only find that in a stand alone product.
          • 2. Re: Need for 3rd-party Noise Reduction Software?
            Level 1
            Thanks for the super quick reply!

            Do you have any experience or thoughts on Dfine 2.0? My understanding is that Dfine applies the noise reduction on a separate layer...would you say that's more non-destructive, and thus a good thing?

            Does NoiseWare apply their nosie reduction on a separate layer too?

            I've done abit of reading on the various plug-ins, and am torn between Dfine and NoiseWare.

            Thanks!
            • 3. Re: Need for 3rd-party Noise Reduction Software?
              Level 1
              >Does NoiseWare apply their nosie reduction on a separate layer too?

              Nope...but it's easy enough to duplicate a layer and then run the filter. I often use Blending Options to concentrate the noise reduction only to the shadows which are often the noisiest places..
              • 4. Re: Need for 3rd-party Noise Reduction Software?
                Hudechrome-sd9sPI Level 2
                I use Dfine and find it an exceptional program.

                One of the important considerations in any noise reduction is it's effect on small details. More often than not, the noise reduction removes both noise and detail to the extent that if you attempt later to increase separation in the shadows, you run into posterization and a resultant additional "noise" i.e. there is tonal differences but not recognizable detail or better stated, no information. I was appalled when I took a digital photo, applied the Infrared B&W conversion, used available tools like Shadow/highlight to separate the values in the shadows, only to wind up with, at the end, with areas that look as if they were hammered with buckshot. I crept back to the RAW file and started over.

                So, my approach is much milder now. If you use Dfine, they advise to do absolutely no sharpening at all before noise reduction, which is reasonable, but not an absolute. I look, instead of shadows, to areas like uniform sky and apply the lightest noise reduction as possible. Then in PS, I usually go to areas like massed trees in the distance or closeup fine detail and erase back the noise reduction on the Noise layer. Alternatively, you can use the sliders to cut back overall, and of course, a combination of both.

                It's worth it to avoid coming to the final stages, doing sharpening for print, only to find the trees are lumps of gray with no definition, or a hillside of a volcano that looks like mud.

                I have found no level of noise reduction that doesn't exhibit even a small amount of detail reduction. There have been arguments about the zero point of the Luminosity noise reduction slider removing detail as well, countered by the argument that at that level, you don't know for certain what is noise and what is detail, but confronted by the above examples, especially when all you need to do is click off the noise layer and see the detail return, all argument ceases!

                So, be careful out there! :-)
                • 5. Re: Need for 3rd-party Noise Reduction Software?
                  Dirk Williams Level 1
                  "Hey, let's be careful out there."
                  Sergeant Phil Esterhaus
                  • 6. Re: Need for 3rd-party Noise Reduction Software?
                    Hudechrome-sd9sPI Level 2
                    Ah, yes. That's where I heard it first! :D
                    • 7. Re: Need for 3rd-party Noise Reduction Software?
                      NR is supposed to be done before capture sharpening. How then 3rd party NR software like Noiseware(that I'm considering buying) should be used in conjunction with ACR sharpening? Noiseware is a Photoshop plug-in but can it be used from ACR?
                      • 8. Re: Need for 3rd-party Noise Reduction Software?
                        Dirk Williams Level 1
                        One way to do it:

                        Do everything you need to do in ACR but do not do any sharpening (ZERO). When you go to Photoshop apply the Noiseware. Then save file as a Tiff and go back to ACR through Bridge and apply your sharpening. Then just click "Open Image" again and open it in Photoshop and save.
                        • 9. Re: Need for 3rd-party Noise Reduction Software?
                          Mathias Vejerslev Level 2
                          With never versions of ACR, you can restrict sharpening to edge details, where it matters the most. Thus you wont be sharpening your backgrounds with capture sharpening.
                          • 10. Re: Need for 3rd-party Noise Reduction Software?
                            Level 1
                            Dirk, Mathias thanks. Sounds like both suggestions are nice workarounds. I'll definitely try them.

                            Jeff, don't you think it seems logical that stronger NR should be built into ACR. That would streamline the workflow quite a bit. You would be done with ACR and proceed to detail and output editing in PS.
                            • 11. Re: Need for 3rd-party Noise Reduction Software?
                              Dirk Williams Level 1
                              You know, I'm of thinking that Adobe leaves in some weaker features in their software so 3rd party companies stay in business. Maybe Adobe gets a fee from these 3rd parties to let them make plugins. Call me crazy but it sure looks that way to me Blind
                              • 12. Re: Need for 3rd-party Noise Reduction Software?
                                Level 1
                                >You know, I'm of thinking that Adobe leaves in some weaker features in their software so 3rd party companies stay in business. Maybe Adobe gets a fee from these 3rd parties to let them make plugins.

                                Horsecrap...first off, if you knew anything about 3rd party plug-in development you would know that historically, 3rd parties have developed stuff that ended up in Photoshop and thus eliminated the market for the plug-ins. So, to accuse Adobe of doing the opposite is silly.

                                Second, a 3rd party who develops a plug-in had all their time/effort dedicated towards that one thing. As a result, they HAVE to work to make it better that what you can get for free inside of Photoshop. The plug-in has to be clearly superior otherwise, why bother?

                                As it relates to serious noise reduction, doing it in the demosaicing is always going to be a problem because CR and raw processors in general are global editors-meaning EVERY pixel gets treated the same. Being able to run a 3rd party plug-in locally means you can hit the noise harder where it's needed and back it off where it's not. That changes the development on the use of a plug-in vs. a built in process.

                                If you think Adobe is intentionally _NOT_ doing something, you're wrong...
                                • 13. Re: Need for 3rd-party Noise Reduction Software?
                                  Dirk Williams Level 1
                                  Jeff...first off, I wasn't accusing Adobe of doing anything of the kind. I basically was saying "it looks that way". I could add, "seems like", or "it appears like". Nothing I said was meant to be factual. Plus I had a smiley face at the end of my post Blind

                                  But I'm glad you cleared that up for us anyway :-)
                                  • 14. Re: Need for 3rd-party Noise Reduction Software?
                                    Level 1
                                    Jeff, what's then the point of having professional grade sharpening in CR? I'm still going thru my learning curve but it seems sharpening in CR can be selective re: midtones and edges
                                    • 15. Re: Need for 3rd-party Noise Reduction Software?
                                      Level 1
                                      >what's then the point of having professional grade sharpening in CR?

                                      Capture sharpening isn't a "local" sort of correction, there's also a degree of sharpening in the demosaicing processing anyway. So, for the purposes of capture sharpening (not intended for final output) Thomas WANTED Camera Raw to improve (since it hadn't changed much if any from the early implementations). And Thomas was influenced by Bruce Fraser's multi-pass sharpening approach which made Camera Raw the natural place for capture sharpeningwhich can be adjusted globally and parametrically.

                                      With noise reduction it far more ISO/Camera/settings sensitive and often VERY difficult to determine the noise/micro detail cutoff. As a result, it's often better to be able to noise reduce the final image on a layer with a layer mask to only hit those areas where you know it's noise and not image detail. That's the way I use NoiseWare...on a layer, blending into the shadows (and really into blues) to reduce the noise selectively.

                                      That's not to say we won't see improvements in noise reductions-Camera Raw 4 has already had 3 pretty substantial moves in that direction...it's just that the parametric adjustments are currently locked into a global parameter so if you are going to reduce noise, it's better for Camera Raw to do it gently and not too aggressively.
                                      • 16. Re: Need for 3rd-party Noise Reduction Software?
                                        Level 1
                                        Jeff, thanks for the promt reply. Being a technical guy I like to understand what's going on and why.
                                        BTW I'm on my first pass thru your book and am really enjoying it. It's written just the way such a book is supposed to be.
                                        Hope you are going to stay tuned to this forum and keep us moving along.
                                        Now, I shoot a lot in dimly lit arenas and similar venues. This means high ISO and noise. I guess for now I need to improvise and probably go back and forth between PS and CR
                                        • 17. Re: Need for 3rd-party Noise Reduction Software?
                                          Bill_Janes Level 2
                                          >As it relates to serious noise reduction, doing it in the demosaicing is always going to be a problem because CR and raw processors in general are global editors-meaning EVERY pixel gets treated the same. Being able to run a 3rd party plug-in locally means you can hit the noise harder where it's needed and back it off where it's not. That changes the development on the use of a plug-in vs. a built in process.

                                          Well, when an edge mask is involved in ACR, the process is no longer global. The edge mask could be inverted to make a surface mask that could be used for noise reduction. I have communicated with Dr. Jim Christian, the developer of NoiseNinja, and he confirmed that NR would ideally be applied in the raw converter where the NR algorithm gets first crack at the data. Exposure adjustment has a major effect on the noise profile. One can make separate noise profiles for various amounts of exposure adjustment and use auto profile matching. Another approch suggested by Jim is to make arbitrary edits and profile each image separately. These problems would be circumvented by having the NR incorporated into the raw converter.

                                          Additional information is in the NoiseNinja FAQ

                                          Where should Noise Ninja be used in the workflow?

                                          It is usually best to apply noise reduction as early as is practical in the workflow. Post-processing adjustments like sharpening, contrast stretching, and color balancing can alter pixel values and noise levels in unpredictable ways. Depending on the amount of adjustment, this can make it more difficult for Noise Ninja to estimate noise levels. Sharpening, for instance, is a nonlinear operation that can significantly distort the distribution of noise values.

                                          If you are working at low ISO with a large pixel camera in good light, then ACR's built in NR should be adequate and one can proceed to ACR's sharpening. However, if you are shooting at high ISO in available light situations where there is little blue in the illumination and with a smaller pixel camera, then noise can be severe, especially in the blue channel. In this situation, Bruce Fraser suggested turning off all sharpening in ACR and going to a third party NR product such as NoiseNinja. In so doing, one loses all the fancy sharpening features of ACR.

                                          One question: When turning off all sharpening in ACR, what should one do with Clarity, which is a kind of sharpening?

                                          It is interesting to note that Bibble Pro has successfully incorporated NoiseNinja into the raw converter. Nikon Capture NX is written in conjunction with Nik Software, which also makes Dfine® 2.0, a new highly regarded NR program. It is not unreasonable to expect that some of these NR features could be incorporated into NX?
                                          • 18. Re: Need for 3rd-party Noise Reduction Software?
                                            On Sat, 12 Jan 2008 04:44:18 -0800, Bill_Janes@adobeforums.com wrote in
                                            <3c05b358.16@webcrossing.la2eafNXanI>:

                                            > As it relates to serious noise reduction, doing it in the demosaicing
                                            > is always going to be a problem because CR and raw processors in general
                                            > are global editors-meaning EVERY pixel gets treated the same. Being able
                                            > to run a 3rd party plug-in locally means you can hit the noise harder
                                            > where it's needed and back it off where it's not. That changes the development
                                            > on the use of a plug-in vs. a built in process.

                                            >Well, when an edge mask is involved in ACR, the process is no longer global. ...

                                            True, but a sharpening plug-in like Focus Magic could be used instead.

                                            --
                                            Best regards,
                                            John Navas http:/navasgroup.com
                                            • 19. Re: Need for 3rd-party Noise Reduction Software?
                                              Mathias Vejerslev Level 2
                                              >In so doing, one loses all the fancy sharpening features of ACR.

                                              If you prefer ACRs sharpening, it can still open Tiffs and JPGs after 3rd party NR...
                                              • 20. Re: Need for 3rd-party Noise Reduction Software?
                                                Hudechrome-sd9sPI Level 2
                                                The Dfine NR takes the same path as Noise Ninja. First, do the NR. But here is a curious situation: When sharpening at or near final stages, any resulting noise can be minimized by slight blurring. So, shouldn't that be applied first? No. The output is always better if you do the blur (Gaussian) last.

                                                I'm talking minimal changes here, like a radius of one and an amount no greater than 10. Usually 5.

                                                This is my findings for my work.

                                                I also use another technique: Double shooting exactly and summing the two images at 1/2 gain. This happens to be easily done right in the camera with the D80. Also in CS3 Extended.

                                                I see improvements even at ISO around 100. Of course, static only and not every subject would actually show this improvement to the unaided eye. But in photographing a timber beam construction of Port Orford cedar, the wonderful surface quality (which defies description!) came through using this process better than NR and better than nothing at all.

                                                NR really should precede the image itself. Once you apply NR to an image you are damaging content. The summing method decreases the ambient noise of the detector without reducing information.

                                                This is an old technique I learned doing instrumentation for nuclear research.
                                                • 21. Re: Need for 3rd-party Noise Reduction Software?
                                                  Bill_Janes Level 2
                                                  >also use another technique: Double shooting exactly and summing the two images at 1/2 gain. This happens to be easily done right in the camera with the D80. Also in CS3 Extended.

                                                  >I see improvements even at ISO around 100. Of course, static only and not every subject would actually show this improvement to the unaided eye. But in photographing a timber beam construction of Port Orford cedar, the wonderful surface quality (which defies description!) came through using this process better than NR and better than nothing at all.

                                                  Lawrence,

                                                  Your method sounds much like stacking, which is widely used in astronomy, but can also be applied to some static objects in normal photography. To stack images with a double exposure does not seem like a good idea to me, since you are collecting no more than the full well number of photons. In astronomical, stacking, one takes multiple images each approaching full well and them sums them so that many more photons have been collected than could be contained in one full well.

                                                  >NR really should precede the image itself. Once you apply NR to an image you are damaging content. The summing method decreases the ambient noise of the detector without reducing information.

                                                  Agreed, but if you are taking 300 images at a basketball game under adverse conditions, there is going to be noise. It is best to start with a good camera for this purpose like the Nikon D3 or Canon 1D MII or III. However, many of us are forced to use what is available. Some loss of detail may not be noticed that much because the pictures are not often that sharp to begin with because of motion blur, shallow depth of field, camera shake and lens aberrations present when shooting wide open.

                                                  Under these conditions, I find that NoiseWare works quite well. One could use more advanced techniques to localize the NR like Jeff suggested, but with large numbers of images, this becomes problematic.

                                                  Another point, not discussed here thus far but covered by Bruce in his Real World Sharpening book, is that NR can produce images that can not be sharpened without bringing back the noise. One would think that an integrated NR-Sharpening program could reduce this problem.
                                                  • 22. Re: Need for 3rd-party Noise Reduction Software?
                                                    Level 1
                                                    On Sun, 13 Jan 2008 10:07:26 -0800, Bill_Janes@adobeforums.com wrote in
                                                    <3c05b358.20@webcrossing.la2eafNXanI>:

                                                    >Lawrence,
                                                    >
                                                    > Your method sounds much like stacking, which is widely used in
                                                    > astronomy, but can also be applied to some static objects in normal
                                                    > photography. To stack images with a double exposure does not seem
                                                    > like a good idea to me, since you are collecting no more than the
                                                    > full well number of photons. In astronomical, stacking, one takes
                                                    > multiple images each approaching full well and them sums them so that
                                                    > many more photons have been collected than could be contained in one
                                                    > full well.

                                                    It actually does work because it averages down random noise,
                                                    particularly in shadow areas, and is non-destructive, unlike even the
                                                    best noise reduction. The catch is image alignment, because rotating
                                                    images can decrease sharpness.

                                                    --
                                                    Best regards,
                                                    John Navas http:/navasgroup.com
                                                    • 23. Re: Need for 3rd-party Noise Reduction Software?
                                                      Hudechrome-sd9sPI Level 2
                                                      In nuclear physics, at least back in the vacuum tube days, this technique was used with photomultiplier tubes, pointed at the event detector, usually some material emitting light as the detector was bombarded with particles. The flash was seen by both PM tubes, which were connected to preamps, then to co-incidence detectors. Pretty straight forward electronics, and it worked well.

                                                      I first devised a curve in Photoshop to apply to each frame, then layered them, then used Luminosity as the blending mode. I first considered underexposing, but that won't work because you lose shadow detail that way. In an analog system, so long as no information dropped below threshold, cutting th gain in half then summing does the trick. No simple method seemed available to do that digitally, so I devised a curve to accomplish it. Later, I saw that the D80 also gave me a path as it's multiple exposure system did the same thing and had the gain reduction built in as an adjustment.

                                                      The key is perfect alignment, and of course is unavaliable to dynamic subjects. But for some applications, it really works well.

                                                      I looked at some of the techniques used in astronomy for sharpening, but never saw anything about NR, although I would expect to see something. The sharpening I tried was called Richardson-Lucy convolution, which worked sometimes and sometimes not. I have a tech pan scanned neg which took to R-L sharpening very well, but with normally grained images, it was not a good choice. And the iteration times required made it problematic.
                                                      • 24. Re: Need for 3rd-party Noise Reduction Software?
                                                        Bill_Janes Level 2
                                                        >I looked at some of the techniques used in astronomy for sharpening, but never saw anything about NR, although I would expect to see something.

                                                        Laurence,

                                                        Thanks for the explanation of your method. It does sound different from stacking. If I reinterpret it correctly, information present in one image and not the other is considered noise, whereas stacking merely averages out multiple frames.

                                                        Stacking as used in astronomy is primarily to increase the signal to noise ratio, as explained in a post by K Wiley

                                                        The Lucy-Richardson and other deconvolution techniques are interesting, but as you mention are difficult to use in general photography because it is no simple matter to obtain the proper point spread function (PSP) to accomplish the deconvolution. In astronomy, one can use a nearby star to get the PSP.
                                                        • 25. Re: Need for 3rd-party Noise Reduction Software?
                                                          Hudechrome-sd9sPI Level 2
                                                          i...it is no simple matter to obtain the proper point spread function (PSP) to accomplish the deconvolution.

                                                          Yeah, tell me about it! Tech Pan worked because the grain is so small as to actually allow a valid psp attempt. But the final result was not any better than other methods.

                                                          It's been several years since I tried it so I don't remember all the things I did. I saw a series of images saved for which I set different parameter as a test, but there was no data saved, so I deleted all except the one that looked optimum.

                                                          BTW, I detected an error posted with my method. Luminosity is not the blend mode. Overlay is what I used. I was getting tired when I wrote that.

                                                          I would expect that stacking and my method (and Nikons) would both be considered a method to increase the SNR, as the signal sums to 1 and noise would decrease 3db.(Theoretically!).
                                                          • 26. Re: Need for 3rd-party Noise Reduction Software?
                                                            Level 1
                                                            On Sun, 13 Jan 2008 15:31:01 -0800, John Navas wrote in<br /><de7lo3pqi2t7jnhckpodcjl37pujv93gr7@4ax.com>:<br /><br />>On Sun, 13 Jan 2008 10:07:26 -0800, Bill_Janes@adobeforums.com wrote in<br />><3c05b358.20@webcrossing.la2eafNXanI>:<br />><br />>>Lawrence,<br />>><br />>> Your method sounds much like stacking, which is widely used in<br />>> astronomy, but can also be applied to some static objects in normal<br />>> photography. To stack images with a double exposure does not seem<br />>> like a good idea to me, since you are collecting no more than the<br />>> full well number of photons. In astronomical, stacking, one takes<br />>> multiple images each approaching full well and them sums them so that<br />>> many more photons have been collected than could be contained in one<br />>> full well.<br />><br />>It actually does work because it averages down random noise,<br />>particularly in shadow areas, and is non-destructive, unlike even the<br />>best noise reduction.  The catch is image alignment, because rotating<br />>images can decrease sharpness.<br /><br />See "Noise Reduction By Image Averaging" at<br /><a href=http://www.cambridgeincolour.com/tutorials/noise-reduction.htm><br /><br />-- <br />Best regards,<br />John Navas     <a href=http:/navasgroup.com>
                                                            • 27. Re: Need for 3rd-party Noise Reduction Software?
                                                              Bill Lamp Level 1
                                                              See slide/negative multi-pass multi-scan. There the different captures are made from one image to reduce noise.

                                                              But, the multiple exposure method is in use.