Help me solve this camcorder mystery once and for all

[okc_smoker]

I wasn't sure of the best room to post this in so I figured I'd post it here. I've had a Canon ZR500 for a couple years now but I just started getting a little deeper into post processing to clean up the footage a bit. I had always assumed that the camera shot 480i since the manual doesn't say anything about progressive, but now that I'm looking closer I'm not so sure.

When I examine the video closely I see no interlacing lines or combing. Deinterlacing the video doesn't seem to result in any change whatsoever. Using AVISynth, if I step through the frames after calling SeparateFields() the video is perfectly smooth whether I assume TFF or BFF.

I even found a few websites that report it as progressive (like here (http://reviews.cnet.com/digital-camcorders/canon-zr500/4507-6500_7-31660615.html)), but the kicker is that Canon support says that it shoots interlaced. I'd be much obliged if anyone could help me get to the bottom of this. Is my BFF/TFF test a definitive way for determining if a video is interlaced or is there a better way?

[neuron2]

Can you post a small unprocessed source sample so that we may verify your analysis?

Separate the fields and step through. Denote a new picture with a new letter. Do it in a motion area. If you see this:

A B C D E F ...

then it is interlaced. If you see pairs of the same picture:

A A B B C C ...

then it is progressive.

[okc_smoker]

Sure, but I'm at work now. I'll post a clip when I get home.

[Gavino]

Is it possible something in your codec chain is deinterlacing before you even see anything?

What format is your source file? Have you checked it with GSpot?

How do you import it to Avisynth? :script:

[okc_smoker]

Is it possible something in your codec chain is deinterlacing before you even see anything?

That's basically what *I'm* asking you guys! :)

The chain looks like this: Camera -> WinDV -> Hard Drive -> AVISynth -> VDub (in Fast Recompress mode)

If I load the file into GSpot immediately after capturing to my hard drive it says the video is "dvsd" (I believe... I'm at work so this is from memory). It also says that it's interlaced, but GSpot is known to report this incorrectly. I'm using the Cedocida DV Codec, FWIW.

For testing purposes, my script looks like this:


AVISource("clip.avi")

AssumeTFF() # or AssumeBFF()
SeparateFields()


I then play the result in VDub to see if I can detect any abnormalities in the movement. Both TFF and BFF appear perfectly smooth.

[neuron2]

I then play the result in VDub to see if I can detect any abnormalities in the movement. Both TFF and BFF appear perfectly smooth. Which of the signatures do you see as described in my post above?

[okc_smoker]

Which of the signatures do you see as described in my post above?

When you say "denote a new picture with a new letter", are you just illustrating how you would watch the frames or is there a way to actually label each picture? If so, how? Sorry, I'm still somewhat new at all this...

[neuron2]

Gosh, I don't know how to say it clearer.

Hmm, maybe like this?

When stepping through the fields, do you see each picture twice (the same except for spatial field offset) or do you see a new picture every field?

[okc_smoker]

Gosh, I don't know how to say it clearer.

Hmm, maybe like this?

When stepping through the fields, do you see each picture twice (the same except for spatial field offset) or do you see a new picture every field?

See, now that was much clearer :) . I thought maybe you were talking about putting some sort of a caption in each frame and then watching how the letters appeared when played back or something like that. I'll give it a shot when I get home and report back.

:thanks:

[okc_smoker]

Here's a link to a raw clip, exactly as it came from the camera. Either click the thumbnail and then click Save or click the link for the zip file.

http://drop.io/camcorderclip

[okc_smoker]

Can you post a small unprocessed source sample so that we may verify your analysis?

Separate the fields and step through. Denote a new picture with a new letter. Do it in a motion area. If you see this:

A B C D E F ...

then it is interlaced. If you see pairs of the same picture:

A A B B C C ...

then it is progressive.

It's definitely A A B B C C, so I guess that settles it. Didee mentioned in another thread that one of my clips shows signs of "bicubic" deinterlacing in some of the motion scenes. The only possibility that I can imagine that fits all the evidence is that the camera does indeed shoot interlaced (as Canon claims) but performs some sort of deinterlacing before recording to the MiniDV tape. Is this a common practice?

[neuron2]

Can't figure out how to download from that site. Why don't you use a well-known storage site like megaupload.com?

Anyway it appears to be a FLV. We need the original source material! Isn't it obvious that giving us something transcoded is totally useless?

The camera is a mini-DV camera. So give us the DV AVI that you captured from it over its firewire interface.

[Didée]

Yeah, drop.io has bit of an unusual layout. But okc_smoker did really upload the raw AVI, as captured by WinDV. (Click thumbnail, the preview-FVL starts playing, click "safe" on right-top of that preview window.)

For convenience, here's the captured AVI on MediaFire: klick (http://www.mediafire.com/?4ngoj3exzmf)

The disappointing fact is: the camera seems to deliver only dumb-deinterlaced material.

[Gavino]

The disappointing fact is: the camera seems to deliver only dumb-deinterlaced material.
Wow, thanks Canon! It's a pity they didn't think to add in some extra digital zoom during the process, then we could have had HD too. :rolleyes:

[smok3]

Here's a link to a raw clip, exactly as it came from the camera. Either click the thumbnail and then click Save or click the link for the zip file.

http://drop.io/camcorderclip

can't be;

Complete name : D:\dl\testvideo_clip.avi
Format : AVI
Format/Info : Audio Video Interleave
File size : 21.4 MiB
Duration : 6s 239ms
Overall bit rate : 28.8 Mbps
Recorded date : 2008-07-04 09:37:03
Writing application : VirtualDubMod 1.5.10.2 (build 2540/release)
Writing library : VirtualDubMod build 2540/release

[2Bdecided]

It's 30p. Some people would pay extra for that (!) but probably not at that low quality.

Cheers,
David.

[dr.schanker]

I found some infos about your camcorder here (http://www.camcorderinfo.com/content/Canon-ZR500-Camcorder-Review/Other-Features.htm).
Some camcorders use a very low shutter speed if there is not enough light.
"Auto Slow Shutter" will then use a shutter speed of 1/30th sec. - that means 30p.

[neuron2]

can't be; It is a DV AVI so I assume he just used VirtualDub to set a range and then direct stream copy.

I think the graininess of the clip supports dr.schanker's low light theory. It can be forced on in the menus as well, so the OP should check that.

This is also enlightening:

http://www.google.com/search?sourceid=navclient&ie=UTF-8&rls=GGLD,GGLD:2005-08,GGLD:en&q=canon+zr500+progressive

[okc_smoker]

It is a DV AVI so I assume he just used VirtualDub to set a range and then direct stream copy.

That's exactly what I did.

Thanks for the info!


This is also enlightening:

http://www.google.com/search?sourceid=navclient&ie=UTF-8&rls=GGLD,GGLD:2005-08,GGLD:en&q=canon+zr500+progressive

LOL! The first result is my post on AVSForum.com asking the same question.

[SeeMoreDigital]

Well it's certainly not "pure" interlaced....

That said, it would be nice to see well lit sample with less zooming.

[okc_smoker]

Not sure if I'll have time, but I'll try to step outside today and take a quick shot in the sunlight.

[neuron2]

Make sure it's not forced on in the menus too.

[okc_smoker]

I'll check again, but I couldn't find any setting for this when I looked previously. There is "portrait mode", "sports mode", etc. but none of those were enabled.

[Didée]

From the info page that dr.schanker had found (see post#17):

Auto Slow Shutter - The auto slow shutter mode can be automatically engaged by entering the administrative menu within the function menu, and selecting the camera setup submenu. Within this menu the user will find the A. SL. Shutter option which can be turned on for recording scenes with insufficient lighting; ...

[okc_smoker]

Ah, I guess I missed that somehow. Thanks. Being that I record a fair amount in what would be considered low light, would disabling this really be beneficial in the long run?

[2Bdecided]

Depends - try both on typical lighting - it's really subjective (I also have a Canon, and change it depending on the shot and the lighting).

btw, my Canon will switch to a fake progressive mode if I manually set the shutter speed to less than 1/50th (e.g. 1/30th or 1/25th etc) even in interlaced mode. It's a well known "feature" of the HV20 at least, though it's single field interpolated progressive, rather than the real thing. (The HV20 also has a real progressive mode which you can select separately).

Cheers,
David.

[dr.schanker]

It's not the worst solution, but motion filmed with 1/30s will not be as fluid as filmed with 1/60s shutter speed.

I assume that you maxed out the aperture size (smallest f-stop-number). So the only options left would be:
- gain(camera) or brightness-postprocessing(editing software) -> this will increase noise too
- use a video light (camcorder accessory) -> costs money and could decrease your mobility

[okc_smoker]

btw, my Canon will switch to a fake progressive mode if I manually set the shutter speed to less than 1/50th (e.g. 1/30th or 1/25th etc) even in interlaced mode.

Hmm, I wonder if that's what's going on. Perhaps I'll shoot the sample clip outdoors without disabling the Auto Slow Shutter and see what happens. If it appears to be interlaced, we'll know that was it.

[okc_smoker]

Okay, I think we found the culprit. I left Auto Slow Shutter (henceforth, ASS... uh, nevermind) ON and shot outside in the bright sunlight. This clip is clearly interlaced. I'll run another test in low light with Auto Slow Shutter OFF to confirm that it's this setting that is responsible for the deinterlacing. Another weird thing is that GSpot reports the deinterlaced clips as I/L and BFF, while it reports this interlaced clip as I/L and TFF.

http://www.mediafire.com/?j2tbwcwmdsc