14 Replies Latest reply on Sep 23, 2009 6:18 PM by OsakaWebbie

    Why is a rendered comp of stabilized video so blurry?

    OsakaWebbie Level 1

      I have two pieces of SD nature footage that I brought into AE CS4 to stabilize from shakes (one was shaky from wind blowing on the tripod, the other from my attempt to get a smooth pan without a really long handle).  Because I wanted the pixels to be as crisp as the original, I did not enlarge the layer to cover up the little gaps on the edges, but instead placed a copy of the same footage unstabilized underneath (my jiggles were quite small, so the exposed edges will never be seen on a TV and only minimally on computers or projectors).  But although the original footage is very crisp and clear, the result of the AE render after stabilizing is rather fuzzy or blurry, even just looking at a single frame as shown in PPro's source pane or monitor pane.  I know that motion blur could be present, but that would affect the original, too, right?  Why is there so much difference, and is there something I can do about it?  I can post short before-and-after examples if that would be helpful.

        • 2. Re: Why is a rendered comp of stabilized video so blurry?
          OsakaWebbie Level 1

          Thanks for the reply.  I have never changed any quality settings from their default "Best" setting, and the graphic I'm positioning is exactly comp-sized (NTSC 720x480), so those two points should not be a problem.  But when I use the Tracker to stabilize the footage, naturally it makes lots of anchor point keyframes at non-integer values - I wouldn't have any idea how to stop it from doing that, and if I did, it would make the result less stabilized, right?  But maybe it would be a subtle difference - I don't know.  Anyway, are those non-integer keyframes likely to be the cause of the blurriness?  If so, is there a way to "round" the keyframe values to the nearest integer or something?

          • 3. Re: Why is a rendered comp of stabilized video so blurry?
            TimeRemapper Level 4

            If you apply the below expression to your Position property, it'll ensure that only Integers will result for the positional values:

             

            [Math.round(position[0]),Math.round(position[1])]

            • 4. Re: Why is a rendered comp of stabilized video so blurry?
              OsakaWebbie Level 1

              Thanks.  I did that (it was the anchor point that needed it, not the position,  but I got the idea), and could see that the keyframe values did end in ".0".  Then I rendered again and compared.  Unfortunately there is very little difference in the clarity between the integer version and the subpixel version - maybe a slight difference, but both are still much blurrier than the original.  Am I just expecting too much?

               

              Specs of my situation:

              • PPro CS3
              • AE CS4 (no Dynamic Link, so I have to render to use my results in PPro)
              • NTSC SD footage and comps
              • Current piece I'm working on: a vertical pan (tilt) down the tree-covered sides of a gorge - very complex image with all the leaves
              • Anchor point keyframes generated by X and Y stabilization
              • About 4-5 position keyframes were manually added after the stabilization in order to follow the tilt
              • No other effects were used in AE - I am just trying to stabilize
              • Render settings: I have tried both Microsoft DV (the normal mode I use due to ease of use in PPro) and lossless - blurriness is the same
              • 5. Re: Why is a rendered comp of stabilized video so blurry?
                TimeRemapper Level 4

                You probably don't, but, just to make sure, do you have motion blur enabled on the layer in the timeline that you're stabilizing?

                Posting example pics would be good.

                • 7. Re: Why is a rendered comp of stabilized video so blurry?
                  Todd_Kopriva Level 8

                  Heh. Steve and I think alike.

                  • 8. Re: Why is a rendered comp of stabilized video so blurry?
                    OsakaWebbie Level 1

                    Nope, to both of you.  Unless that's something that AE would put on by itself.  All I did was bring in the footage to a new comp, duplicate it (to form a background layer that isn't stabilized, to cover the edges), run the stabilization on the top layer, add a few position keyframes to restore the tilt, and render.  Then, after looking for answers and then posting to the forum, I added the expression Steve taught me, rendered and tested again, then tried lossless and tested again.  All four versions are in PPro, where I can compare them easily.

                    • 9. Re: Why is a rendered comp of stabilized video so blurry?
                      TimeRemapper Level 4

                      Todd_Kopriva wrote:

                       

                      Heh. Steve and I think alike.

                      Honored, sir.

                       

                       

                      OsakaWebbie, did you separate fields upon importing your footage, if the source was indeed interlaced?

                      If so, are you rendering interlaced, or progressive?

                      • 10. Re: Why is a rendered comp of stabilized video so blurry?
                        OsakaWebbie Level 1

                        Oh, yeah, and here are example pics as requested.  They are all frame captures in PPro of what I think is the same frame of the video.  It is almost at the end of the tilt downward, but not quite - it's still moving at this point.  (I used a spot where the river showed so that I could see something unique to know that I was getting the same frame on all four, because the original is longer than the comp.)

                         

                        I prepared four frame captures, but I'm only allowed three files, so I zipped them together.

                        • 11. Re: Why is a rendered comp of stabilized video so blurry?
                          OsakaWebbie Level 1

                          OsakaWebbie, did you separate fields upon importing your footage, if the source was indeed interlaced?

                          If so, are you rendering interlaced, or progressive?

                          Yes, my footage is interlaced, but I'm not familiar with separating fields.  What is that, and should I be doing it?  I'm pretty new to AE - it wouldn't surprise me if I'm missing some basic steps in usage.

                           

                          During render, I'm selecting render settings of "DV Settings" and output module "Microsoft DV NTSC 48Hz" with audio output deselected.  I don't see the word "interlaced" or "progressive" appearing anywhere in the settings (but I could be missing it - it's after 3 a.m. here), but the Field Render is set to Lower Field First, so that's a clue.  And when I render, I see it count the frames with L and U alternating after the frame number.  When I did the lossless test, I allowed the default settings of "Current Settings" and "Lossless" - the Field Render for "Current Settings" is Off, so perhaps that type is progressive.

                           

                          Because of your question, I decided to try a render with all the same settings as I normally use, but with the stabilized layer hidden, so that only the unstabilized background layer is rendered.  It looks fine (same frame capture attached), so it definitely seems to be something about the stabilization that is causing the problem.

                          • 12. Re: Why is a rendered comp of stabilized video so blurry?
                            TimeRemapper Level 4

                            Yes, you need to separate fields for your imported footage.

                            I'm guessing that your footage is Lower Field First due to it's NTSC SD nature.

                            You are rendering interlaced (Lower Field First), which is fine.

                             

                            A different issue is that of the codec you're choosing when you render.

                            When making intermediate renders for footage, always try to render uncompressed, or losslessly compressed (i.e. Quicktime Animation @ Best/Full quality).

                            If you render to a lossy format (i.e. DV), you'll be causing degeneration, and there's no need for that.

                            • 13. Re: Why is a rendered comp of stabilized video so blurry?
                              OsakaWebbie Level 1

                              Okay, I seperated fields on the piece of source footage in my Project panel (and checked the box to preserve edges as well), but I have a question: Do I need to rebuild the comp, pulling the footage from the Project panel to the comp again and redoing the stabilization, or is it already now using the separated data as it is and all I need to do is render again?  I expected more dramatic action when I hit okay on the dialog box, but nothing visually happened (no hourglass or anything).  How do I know the field separation is working?

                               

                              Yeah, I know about the degradation using DV, and I knew someone would give me the lecture about it.   But I'm willing to trade one generation of DV loss for at least two advantages: (a) in PPro, lossless footage renders very slowly and previewing is very choppy, whereas DV is the native format in PPro so it doesn't need extra processing, and (b) I'm storing these stabilized raw footage clips (some of them quite long, like sunrises that are timelapsed later) in a large master collection on an external drive alongside other DV clips straight from my camera that didn't need stabilization, for possible use in various projects - the lossless files are huge and cumbersome.  In fact, since I always use the clips in PPro rather than straight from AE to MPEG2-DVD, it might be that the rendering that PPro is obviously doing for previewing lossless AVIs means that it's doing the intermediate conversion to DV anyway, just for its internal editing processing - if so, then I'm losing the same generation regardless.  Anyway, I'm not the kind of purist that many of you are about these things - if it looks okay to my eyes, I don't lose sleep over knowing that I added a generation of DV-level compression if it makes my life easier.  Everything I do ends up on a DVD, so the MPEG2 compression will be much more significant than a generation or two of DV.

                               

                              Test results: If the answer to my question above is that it is sufficient to just change the settings in that dialog box about separating fields and there is no need to rebuild the comp, then something else is still wrong.  I did a test render after doing only that, and just to alleviate your concerns about compression generations, I chose Best Settings/Lossless and simply changed the Field Setting from None to Lower Field First.  It's just as blurry in PPro as the rest of them.  I attached another frame capture.

                              • 14. Re: Why is a rendered comp of stabilized video so blurry?
                                OsakaWebbie Level 1

                                I haven't heard any replies in a while, and I'm doing my final edits to my current project today, so it would be nice to have a little more discussion about this.  In regards to the issue of separating fields, I may have discovered why changing the setting didn't seem to make any different in my case - it may be that I have had separated fields all along.  Notice that in the same help file referenced earlier in this thread, there is this sentence (I didn't catch it the first time I read the page): "After Effects automatically separates fields for D1 and DV video footage items."  In theory, that should mean that AE is doing what I expect would be the most logical thing: handle incoming footage in the way that it can be best manipulated.  Correct me if I'm wrong.

                                 

                                I burned a test disk with my footage, and interestingly enough, it doesn't look as bad in the final product than when I'm looking at frames in PPro.  So I may just ignore what it looks like in PPro and put up with the rest, although a real solution would be better.

                                 

                                P.S. I can't seem to get the latest screen capture file (the TIF) I attached to do anything in my browser (offering download like the previous ZIP is what I would expect, but it just sits there) - can you guys get it, or did something go wrong?  So far no one has said, "Hey, we can't get the screen capture", so I assumed it was fine until just now.