MPEG-2 motion blur

[Rick Llewellyn]

I find that with a HV10, when I do a slow pan, the image blurs. This effect is about 4 times bigger than the interlace "blur". I presume this is caused by the image processing chip giving up on tracking motion of the scene and basically dropping to DV resolution or less. I-frame, long GOP and all that.

If I pan extremely slowly, this blurring effect does not happen.

1. Is my assumption correct or is something else going on?
2. Do all HDV cameras do this? Do some cameras have more processing and lessen the effect?
3. Would the HV20 or XHA1 act the same way?

Rick

[Robert Ducon]

Rick, I own an HV20. I only noticed blurs when the shutter speed was to low for movement above 1/60th sharpened things up quite a bit. Could it be this?

[Pat Reddy]

Maybe it's my imagination, but I notice significantly less panning blur on the HV20 compared to the HC1. I deinterlaced a lot of footage from my HC1 to correct this, and it seems that I will not need to do this as often with the HV20.

[Robert Ducon]

On the built in LCD monitor ON the HV20, I noticed weird effects.. blurs (LCD lag) on pans, and massive oversharpening on high contrast items - I was scared my footage was being messed with.

But upon import and then watching on my external TV monitor, everything is fine!

So are you both talking about watching the footage on the built in HV20 LCD? I wouldn't trust it for detail - just for framing really. It also overscans, so more of the image is being captured to tape than it shows.

Good stuff to know.

[Dennis Vogel]

It could simply be an artifact of the HDV MPEG-2 on-the-fly compression. Could you post some video for us to see?

Good luck.

Dennis

[Rick Llewellyn]

Well, I think we can put this one to bed.

I went back and shot a bunch more video under controlled conditions at various pan rates. The bottom line is that the mpeg-2 compression is not causing any motion blur! That is good news. I don't know what caused the problem in my previous footage.

Here are some details that might be interesting:
I took a number of shots at pan rates of 1 second per horizontal FOV up to about 20 seconds. I shot at 1/500 with a HV10, so it is 1080i. In all the shots there was the predictable interlace. Even at a 1 second pan across the FOV the image was fairly sharp. At 1 second per HOV, and 1/500 you would expect about 4 pixels of blur and I saw no more than that in a deinterlaced image.

Further, I was playing the same video back on a 24 inch iMac and it clearly blurs the imagery much more. The iMac bluring could be the LCD lag time or a mpeg-2 bandwidth limit.

I suspect that my original problem was the 1/60 shutter speed. At 1/60 and a 5 second pan you are looking at about a 7 pixel horizontal blur. That and the iMac display were probably my problems.

Thanks for your suggestions and comments, especially about the shutter speed. It is good to know that a lowly HDV camera's mpeg-2 compressor can keep up with that kind of motion.

Rick

[Kevin Samborn]

Well, I have to say I noticed it too, when I first got the camera, and I haven't seen it since.

Exactly the same thing - pans on a tripod, looking at the sea.

A couple of times in the footage there were some weird glitches.

But, I never saw it again!

[Robert Ducon]

I had said that CMOS was due to lags, not HDV compression. My earlier post was edited by a moderator and 2/3 of what I wrote was removed. I've got a lot of experience with other CCD HDV cameras, so I can tell. It's not an issue though, just something to be aware of when shooting (try to keep away whip pans/tilts).

[Pieter Jongerius]

I think typically, when MPEG2 is overchallenged, you will see blocks instead of blurs. When I get my HV20 (any day now :)) I plan on shooting water splashes, high DOF long grass in strong wind, maybe chase an impala on the serengeti (gotta stop these day dreams) and shoot other fast moving complex images that require a lot of detail to be encoded, surpassing MPEG2 motion detection. Just to see where 25Mbps ends...
(Ironically as well as completely beside the point, MPEG2 has a breeze when it has to encode motion blur: the softer gradients and general decrease in detail is fairly easy to 'describe').
regs
pieter

[Mike Dulay]

Reviewing all my m2t files for the first time I noticed this horizontal banding during panning movement. I'm not too familiar with the correct lingo, are these motion artifacts? interlace "blur"? Looking at other raw footage I see this is common with HV20/HV10 and appears to me like interlace lines. Even in HDV24PF mode (tried cine and TV-48) these are present. Then I remember that the camera is a 1080i device so all m2t must get encoded interlace, does it not?

Anyway, the artifact/interlacing/blur(?) is most apparent on overexposed areas and edges. Recompressing to a true 24fps output makes the video true progressive. It took me a while to get my head around it as my head was stuck on my old Elura (first model) which had a separate progressive mode. So this means, to have pretty playback on a progressive display (computer/DVD) you must remember to deinterlace/make progressive.

Raw M2T file, HDV24PF -49MB Cinemode -- pan on a manfrotto 506B -- watch the shiny door knob
http://www.megaupload.com/?d=BPXIDP2Y

Reencoded as MP4, 24fps progressive, 1.333 aspect, no stretch, 1920x1080 - 1.2MB
http://www.megaupload.com/?d=GIVUFQC6

[Colin Gould]

Originally Posted by Rick Llewellyn

Well, I think we can put this one to bed.

I went back and shot a bunch more video under controlled conditions at various pan rates. The bottom line is that the mpeg-2 compression is not causing any motion blur! That is good news. I don't know what caused the problem in my previous footage.

Here are some details that might be interesting:
I took a number of shots at pan rates of 1 second per horizontal FOV up to about 20 seconds. I shot at 1/500 with a HV10, so it is 1080i. In all the shots there was the predictable interlace. Even at a 1 second pan across the FOV the image was fairly sharp. At 1 second per HOV, and 1/500 you would expect about 4 pixels of blur and I saw no more than that in a deinterlaced image.

Further, I was playing the same video back on a 24 inch iMac and it clearly blurs the imagery much more. The iMac bluring could be the LCD lag time or a mpeg-2 bandwidth limit.

I suspect that my original problem was the 1/60 shutter speed. At 1/60 and a 5 second pan you are looking at about a 7 pixel horizontal blur. That and the iMac display were probably my problems.

Thanks for your suggestions and comments, especially about the shutter speed. It is good to know that a lowly HDV camera's mpeg-2 compressor can keep up with that kind of motion.

Rick

Thanks for some good scientific testing.
I especially perked my ears up when you mention the onscreen/LCD blurring-
is this what is being talked about? on computer monitor, or on actual HDTV display?

I have a Viewsonic 5-8ms response 20"ws LCD, very good quality, and it blurs/judders a decent amount for fast motion (even on smooth VLC playback given my CPU).
I see NO such blur etc when connected via component HD to my RP-CRT HDTV, it's MUCH smoother. LCD can't touch CRT :)
So, just like w/ SD, never trust your computer display for "real" monitoring, but use a real display :) (of course if you have an LCD HDTV, maybe it won't be as different, but...)

[Rick Llewellyn]

Collin
I was watching on a iMac 24". The bluring seemed like a lot more than something on the order to 10-8 msec. At 60Hz field rate, that is still 16 msec per field. Unless the tail on the lag time is really long, then maybe there is something in bandwidth limitations of the display system, like the video cord.

In any case, your point is very well taken- computer monitors and the stuff behind it should not be assumed to display video well.

Thanks
Rick

[Rick Llewellyn]

Mike-
I looked at your video. The m2t clearly is progressive with a pull down. Not to quibble over terms, my understanding is that the HV20 sensor in 24p gives you true progressive images. The the processing and recording system adds a pull down so the 24p imagery meets the format requirements of 60i.
It looks like you removed the pulldown from the 60i stream rather than just recompressing to mp4- is that correct? Also how did you deinterlace? Although it is hard to tell with your subject matter, it didn't look like you just threw away all the even fields.

I found in my pans that the motion blur showed up at a bit fast pans. Yours looks nice and clean to me.

Rick

[David Chien]

1) CMOS sensors blur when motion is too quick. That's just life - read out of the individual pixels takes time, so if the motion is too fast, it'll result in oddly bent shapes.
2) LCD monitors blur. They haven't gotten it down to perfect ala CRT yet, so don't expect to be able to judge the final video quality w/o a CRT, esp. with fast motion. Most panels state 8+ms pixel response time, with some going down to 2ms, but realistically, that's for gray-to-gray pixel response, not black to white or vice versa response. So, don't expect silky smooth video on anything but a CRT (or plasma). Forget most LCDs.
3) Playback of interlaced material on progressive panels.
Ah, yet another hiccup if your PC's video playback software isn't so good at decoding. It'll often result in visible, horrible banding (eg. here, I can definitely say that playing my interlaced DVDs from HV10 on iMac 22" is just bad! The apple DVD player is horrible vs. my PC WinDVD software!! You can easily see banding, interlace artifacts, etc.), esp. if you scale the image up to a full-screen high-res monitor.
Again, only real way to judge is to play it on a real CRT from the HV camcorder (or other source - HD DVD, BluRay, or DVD made from your video).
4) Bad settings.
eg. in Vegas Video, if you accidentally don't pick 'interpolate' when downsizing to a smaller resolution, you'll pick up all sorts of interlacing problems outputting to DVD video! Same goes with top-field vs. bottom-field. You really must test the video on a TV/CRT playing back from that DVD first to make sure you've got the settings right -- anything else will result in blurring/interlacing problems that you can see right away.

[Mikko Lopponen]

Originally Posted by David Chien

1) CMOS sensors blur when motion is too quick. That's just life - read out of the individual pixels takes time, so if the motion is too fast, it'll result in oddly bent shapes.

They don´t blur... They just scan the image from top to bottom, there´s no more blurring than in ccd devices.

And there are cmos cameras without any rolling shutter effects.