10 Replies Latest reply on Jun 4, 2012 2:03 PM by ignitionrocks

    grid like interference pattern with scaling

    (Andrew_Shanks) Level 1
      This is a strange little issue that has been brought to my attention by colleagues after they had some shots rejected in tech check (when they gamma slammed the heck out of the footage and saw these artifacts, ...when you know they are there, you can see them without any crunching of levels).

      Basically when doing a scale up or down on a layer (more visible on the smaller scales) you get this grid-like interference pattern happening.
      I have tried it on numerous packages. After Effects, Combustion, and Photoshop all had the same grid patterning when scaling up this 1080p footage by 104%. Nuke had it at its default cubic filtering, but could be eliminated by using Parzen or Notch). Shake was the only application that seemed to do the transforms without creating the grid interference (am guessing it uses a different transform filter by default).

      Here is a link to the image, I have marked little yellow squiggles at the start of the grid lines to help see them, plus the shot has been graded to make the grids obvious).


      From what I can see it is something to do with After Effects (and the other apps) use of bicubic sampling. My questions are, has anyone else come across this, and if so, do you have a fix? My only suggestion to the others was an adjustment layer with a 1 pixel fast blur followed by an unsharp mask filter (at default settings), over the top of the worst clips. Not ideal as the image becomes slightly softer, but at least the grid is dealt to enough to be passable.

      Its a real head scratcher as I have never encountered this before in AE, ...and it seems to depend on the footage, as other clips seem to scale with no artifacts.


        • 1. Re: grid like interference pattern with scaling
          Todd_Kopriva Level 8
          Sorry if this is a silly question, but can you confirm that the pattern wasn't in the original footage?
          • 2. Re: grid like interference pattern with scaling
            Level 1
            In this case the project files may be of some help. How is the grain applied? Some noise generators must be in multiples of 64 pixels. It looks to me like this could be part of the problem.

            I've never seen exactly this problem, but I have seen similar problems with earlier versions of some noise and grain plug-ins.
            • 3. Re: grid like interference pattern with scaling
              (Andrew_Shanks) Level 1
              It isn't in the original footage (the source is a Genesis sourced DPX sequence). There is no grain added, this is the source footage scaled (with colour correct to make things more obvious, ...you can still see the grid without the colour correction, but for posting I thought it best to make it as clear as possible). we're 99.99% positive now, that this is just to do with the bicubic filtering when doing transforms in After Effects (Rick is kind of on the right track there, in the way bicubic samples in blocks).

              Since I posted this earlier today, I have found that if you use the resize image command in Photoshop and switch the filtering to 'smooth bicubic' the grid becomes softened as to almost disappear, you have to look hard to see any trace of it (to the degree that I think it would pass tech checks). Would be great if After Effects had the option to change what filter you used in a similar fashion (or even had an additional transform filter for these tricky situations, which allowed transform filter changing).

              I have been conversing with a friend in Canada who has since done his own tests in Shake using a grey solid, grain, then resized using different tranform filters. He has found it does occur in Shake too (but again this must be image specific because in the example I have posted it doesn't seem to have the grid problem). His results using different transform filters were:

              BOX: Brutal. Moving on...
              SINC: Very good, but too sharp
              GAUSS, QUAD, TRIANGLE: Slight Gridding & quite soft
              DIRAC, IMPULSE: Blergh, visible stretching at seams.
              LANCZOS: Slight gridding, but sharper than GAUSS
              MITCHELL: Halfway between LANCZOS & GAUSS

              One of those strange things where maybe we should not be looking so close at these images (higher scalings are fine, it is the ones within 10% that are an issue).

              Still looking for ideas for work arounds if anyone has them?
              • 4. Re: grid like interference pattern with scaling
                Mylenium Most Valuable Participant
                Workarounds: Apparently resizing plug-ins such as Resizer and Instant HD Advanced. All these considerations aside, I'm wondering if it could not possibly be some issue with uneven luminance in the source footage, perhaps due to some compressed intermediate processing while recording from the Genesis. In my lo-fi world access to such high profile equipment is limited, to put it mildly, but I could imagine that some disc recorders would still compress even if receiving a perfectly clean signal. Or perhaps some incorrect on the fly PAR tweaking when routing the signal throug hanother device (i.e. back and forth conversion by the preview monitor)? Maybe AE just brings out the bad more than other tools. I dare say it's certainly a combination of conditions, not a single cause.

                • 5. Re: grid like interference pattern with scaling
                  (Andrew_Shanks) Level 1
                  Thanks Mylenium, I usually live in the lowfi world as well (TV-land with my toes occasionally dipping into the film world), and it is just so weird I have never come across this before with any of those (more compressed) formats (certainly have never had anything rejected). I think you are right about the resizing just accentuating something that is there already (such as compression, maybe even patterns in the cameras sensors, etc), much in the same way that pushing the levels in a grade on compressed video will show up the compression artifacts. I know that they are going through HDCAM-SR tapes, and at some point I can see that the footage has become 4:2:2 by the noise introduced on the chroma channels (the camera is 4:4:4 in theory). We have some Phantom test footage which I will test today to see if it has the same issue (different post pipeline so not likely to rule anything out because so many variables have changed). I also can lay my hands on some Genesis footage from another facility, so will test that in the next day or two.


                  • 6. Re: grid like interference pattern with scaling
                    (Andrew_Shanks) Level 1
                    quick update, Phantom footage has same issue at 104% scaling, doh! Will give the Magic Bullet Instant HD demo a whirl to see how it fairs.
                    • 7. Re: grid like interference pattern with scaling
                      Mylenium Most Valuable Participant
                      Mmh, weird. Then perhaps indeed a 104% scaling causes some values to be perfect evens, bringing out the pattern in regions where the subsampling pattern coincides with those even values... It's strange, though. I never have noticed anything like that before, either.

                      • 8. Re: grid like interference pattern with scaling
                        (Andrew_Shanks) Level 1
                        Just to conclude this thread, and help others who might strike this weirdness, we ended up learning to work around the transform issues (a little blur here or there does wonders too), but also I found this fxguide article which goes over this whole issue, with various compositing appliations and transform algorithms, plus has some tips for the various applications at the end of it (for AE it is basically to limit nesting moves in precomps). Here's the link:

                        • 9. Re: grid like interference pattern with scaling
                          craulmedia Level 1


                          -Excerpt From FXGuide

                          tips for After Effects

                          The important thing to remember is that a pre-comp is exactly that -- a pre-processed composite. If you want to avoid unnecessary image degradation, limit the nesting of composites within composites.

                          If you want to avoid resampling an image in After Effects, you can move the layer in 1 pixel incremements.


                          So much for the discussion of not having "smart folders" in AE timeline instead, uh, pardon for the precomp guys, additional to precomping = good for using one composit over and over again, i know... (smart)folders are a rock solid concept JUST similar to smart objects in PS because in PS tough you have the "Precomping" philosofy there then instead using a folder itself as the receiver for scaling, filters, whatsoever properties.
                          No need for a layer to be affected by the folder? just pull the layer(s) out and you're done.

                          sounds like a walk in the park? its not. i could not even write the first three correct letters for the programming script achieving this.

                          i'd better save some bucks every month and wait till christmas for CS5 and hopefully it will drive me to a major upgrade of Ae, i've never missed any, but with CS4 i had to, driving my anger due to the lack of "since then" implementations... but, keeping my chin up, can't wait, good luck ppl out there!
                          • 10. Re: grid like interference pattern with scaling

                            andrew, did you ever resolve the grid issue back in 2009, any new solutions?


                            we're experiencein the same issue and curious.

                            thanks, ralph smith