Sample Video - Need Advice

[Richard Hunter]

I was going to post the video below to show how nice XL2 images can be, then I noticed some really horrible artifacting that has me very worried.

If you look at the nameplate on the old motor car, it starts off fine, but then as the shot pulls back, the slanting edges turn really nasty. Is there any way to stop this happening in future, or at least reduce the effect? I was shooting 25p and had vertical detail set to Low. I also had the Century Optics 0.7x converter fitted on the 20x zoom lens.

I'd appreciate any advice.

Richard

http://www.jaegercat.com/~richard/OldMercedesLowRes.wmv

[Advil Dremali]

Link doesn't work bro.

[Richard Hunter]

Sorry about that. Should be OK now.

Richard

[Barry Goyette]

Richard,

You've just discovered the source of a lot of conversation around here a few months back. Do a search for "moire" and you will find angst, bitterness and even a little truth about the subject.

In a nutshell, all digital capture devices will exhibit this type of problem at the appropriate "sensor frequency (resolution) to image pattern frequency" ratio. The xl2, because of its higher resolution and progressive scan may cause this more than other camera's you've used. But all of us who were debating this last august-september had the opportunity to see NBC's HD cameras doing it even worse during the Summer Olympics. So at least you're in good company.

Some things you can do: You've done one of them, lowering the vertical detail, but from my experience this doesn't seem to help much. Second is to lower the in-camera sharpening. Especially important in 4:3, as the sharpening seems to be tuned higher in this mode. Third you can use a softening filter to take the edge off the finer detail. (all of these will have the effect of lowering the sharpness of the xl2's spectacular image --disappointing, huh?)

Lastly, and most importantly, you can act like a knowledgeable videographer and be aware of situations like the one you showed us. Angled black lines against lighter backgrounds will produced this kind of artifact to some extent in all DV cameras. Like you said with the placard, it looked fine until you started zooming...hmmm...maybe it was the zoom that cued you into the problem...now you know that zooming in situations like this might cause a problem...you're aware...now avoid.

All cameras formats have limitations that we, as professionals, need to learn about. The solution to your problem lies in knowing when to apply the above fixes...every shot doesn't need them, just the ones that cause the problems. (note also my previous posts regarding the effect of the output device relative to moire pattern effects).

Good luck.

Barry

[Richard Hunter]

Hi Barry. Thanks a lot for the valuable info in your reply. It makes me think that maybe I made a big mistake in buying this camera. The fact is, for the videos I make., I like to shoot this kind of style and having to avoid zooms because of such effects is a very unattractive prospect for me.

My previous camera, which I still have, is a Sony VX2000, and I've never seen this type of problem with it, at least not to the extent that it was total unacceptable. Hopefully this means that 4:3, 50i PAL DV video is not inherently as susceptible to these artifacts.

I will try out some tests using 16:9 mode. I don't think 4:3 has additional sharpening as you say, it's just that the 16:9 mode causes a softening of the image beacuse the same number of pixels has to represent a wider image. However, I doubt if this will help much since it is only effective in the horizontal axis.

I will also try to compare results in 50i mode. If there is a significant difference, and I find that I have to choose between zooming and progressive scan, I think I'd have to go for interlaced and keep the zooming capability. For what I do, this would be more important to me.

Thanks again.

Richard

[Barry Goyette]

Richard

Your sony, being an interlaced only camera will not show this problem to the extent of the xl2 or DVX100. The higher resolution progressive scan is really the culprit (especially when coupled with a lower resolution output device -- ie standard NTSC set (or PAL in your case), LCD monitor, or the camera viewfinder.) Even your posted clip being lower res than the actual DV footage will probably make more of the problem than the actual footage (viewed at full resolution) does--or at least it will show it differently. (Example:The original poster who brought this issue to the fore months ago, showed us deinterlaced images (halved vertical detail) to illustrate the moire he was seeing, because the moire didn't exist in the actual footage).

The only way to eliminate the problem is to eliminate the "detail" that is causing it in the first place...you can go interlaced as in your vx2000 or xl2 in 50i mode (cuts the vertical detail about 50%) or soften the footage as described before.

Don't misread my comments on zooming. It doesn't mean you can't zoom. Just to be aware that some types of image detail might cause the problem. I zoom all day long, and rarely do I see this problem. To explain the issue another way, I own a 22mp still camera that exhibits moire issues from time to time (because it doesn't have the anti-moire "softening" filter that most consumer digital still camera's have. The established technique among pro still photographers to eliminate moire on these high resolution cameras is to change the zoom, or camera distance to adjust the image pattern frequency as it hits the chip. By running through a zoom on the xl2, you are essentially trying out all the potential moire points, which is what is causing the artifacts you are seeing.

Regarding choosing between progressive and interlaced relative to zooming, you may be making the right choice. Interlaced will certainly give you a smoother zoom, albeit one with less detail. Another option would be a minimal softening filter to take the edge off the detail (if having the look of progressive scan is important to you).

Barry

[Richard Hunter]

Hi Barry. Yes, that all makes sense, thanks. Regarding the resolution of the clip, actually it is not low-res, despite the clip name. It is rather low data rate - the first version I made juddered on playback on my PC so I created this one at lower data rate to play more smoothly. The resolution is 720x576, same as the original DV. As such, the moire effect is exacly the same when I look at the captured DV AVI on a PAL monitor, I just didn't notice it until after I checked the wmv. (I usually pay more attention while editing, but I wasn't really editing in this case :) )

I will try more softening in 25p mode and see how it turns out. There are also a lot of random "twinkly bits" in most of my 25p shots that are probably due to the same problem. Hopefully these will also be reduced by the softening.

Richard

[Barry Goyette]

Richard,

Sorry, my WM9 settings were wrong...I was viewing at half size. I think something is going on in the windows media encoding, As the problem seems much more prevalant in full size .Are you able to view your DV file side by side with the WMV on your computer? (not on the pal monitor). I'm curious if you see the same thing.

I'm no expert regarding PAL, but I know that on a NTSC monitor, the xl2 can show some moire artifacts that I don't see when I view the DV file on my computer. It almost looks like WM is deinterlacing your file to get the lower bit rate, which from my tests in the past will introduce a lot of these problems. Here's a clip of mine thats got lots of hard edges in it (not much in the way of zooms), but I'm not seeing any of the "judder" or "twitter" moire problems you are. (encoded in mpeg4...you may get some stuttering on your machine).

Barry

http://homepage.mac.com/barrygoyette/FileSharing16.html

[Richard Hunter]

Hi Barry. Your shots are very nice, and I don't see any problem with them at all.

Following your last post, I played around with all the settings I could think of to try and get better results from this clip. But first, I took the output of my DV deck straight to a TV monitor and confirmed that the problem exists on the raw video.

After that, I played around with project and clip settings for field order and deinterlacing in vegas, but could not get any improvement. I also tried Premiere 6 and Edius Pro with the same results.

Eventually, I tried applying the antiflicker filter in Edius, and this helped quite a bit. Here's a link to the file.

http://www.jaegercat.com/~richard/OldMercedesAF.wmv

So, at least there is some comfort in the thought that I can make things a bit better if this effect shows up in future.

Thanks again for your valuable advice barry.

Richard

[Barry Goyette]

Yes it does look a lot better, although I'm still trying to figure out why you should need to "do" something to the footage to solve the problem. In fact there is no real way to "do" something about moire once it is in the file which leads me to believe, as I've already guessed, that the problem isn't in the DV file/xl2 image at all, but rather in how some setting is causing it to be rendered on your screen. Showing the footage on a TV doesn't really tell us much about the actual image, as typically the resolution of a tv is lower than the DV source(especially a progressive image)..and thus can introduce this kind of patterning when none exists in the footage. Viewing at 100% on a progressive hi resolution monitor (like a computer) is really the best way I know to evaluate the image. But if you are seeing the problem there (on the native DV footage) then something is wrong. For now, you've found a workaround, but something tells me there is a much simpler answer...unfortunately I live in the world of the Mac, Quicktime, and FCP, so I can't be of much help given your workflow.

Barry

[Richard Hunter]

OK Barry, I think I know what the problem is and it looks like you were right all along. Sorry for being so slow to accept what you were saying, and thanks for persevering.

I always use Procoder 2 for file conversion to WMV, and it seems that this is carrying out some kind of interlacing process, even though my source and target are both progressive. I've just tried the conversion again, this time using Procoder Express, which is a cut-down but more up-to-date version of Procoder. This is the result, with no processing.

http://www.jaegercat.com/~richard/OldMercedesPCE.wmv

Richard

[Joseph Andolina]

wow, what a drastic differance of an improvement!!!!

[Richard Hunter]

Hi Joseph. Yes, it is very drastic, as you say! I've now reported the problem to Canopus, so hopefully they will take care of it before too long.

Richard