Camcorder compression

[junglebeast]

I'm trying to find the best cheap solution to get high spatial resolution with high spatial precision.

I tried using a Sony Handycam with miniDV, but after importing with iMovie I found that the video had SEVERE artifacts -- on almost every frame, random rectangles of the image had been horizontally offset by up to 20 pixels sometimes. I see similar artifacts on my webcam and on some online movies, but not this bad. Was this a fault of the Handycam, or the miniDV format, or iMovie?

At first I thought this was a result of image compression (eg, MPEG-2) but after further research, it looks like miniDV uses only intraframe image compression (no temporal compression). What could cause this?

After reading about video formats, it seems that all the digital formats do lossy spatial compression, and most of them also do MPEG-2 temporal compression. My intended use is to do high precision 3D reconstructions, which requires accurate spatial position...this is why I cannot have things like random blocks of the image being transposed around, or temporal aliasing. Spatial compression is not as bad, but if the compression artifacts become visible, then it too causes problems.

I was thinking about using S-VHS because it is analog, so I can avoid the compression issue. Then I could buy a TV Card and use that to create a digital file that has lossless compression.

In looking at S-VHS, I noticed something interesting: S-VHS stores uncompressed data equivalent to 560×480. After converting to digital, at 30 fps this would be the equivalent of 560*480*3*8*30*10e-9 Mbps = 193.54 Mbps...which apparently is written into S-VHS. That would be something like Panasonic AG450.

I was confused when I compared this to some newer miniDV cameras, that only write at 13 Mbps. Why would they do so much compression to reduce bandwidth to 14 Mbps if there is other media that can sustain 193 Mbps?

Is my best bet really to buy an old Panasonic AG450, or are there newer (low end) cameras that I can be assured will not have these kinds of issues?

[edDV]

Did you capture the MiniDV over a Firewire connection? Something isn't right. DV format lightly intraframe compresses (~5x) with no temporal compression. Audio is uncompressed.

[junglebeast]

I used a Sony Handycam DCR-HC 62. I used a brand new MiniDV. I loaded it into a tapedeck which was in the media lab, so I don't know what type of port the tapedeck used to connect to the computer.

I should have saved screenshots to show how it was messed up...but the quality was such crap that I didn't even bother to save the digital import. However, the artifacts were exactly the same as this type which came from my webcam,



There was stuff like this on every single frame

[edDV]

I used a Sony Handycam DCR-HC 62. I used a brand new MiniDV. I loaded it into a tapedeck which was in the media lab, so I don't know what type of port the tapedeck used to connect to the computer.
...

There was stuff like this on every single frame

You must use Firewire to get the full bitrate 720x480 DV. The USB port will get something like CIF 352x240 intended for webcam use. That looks like CIF that has been upscaled.

The DCR-HC 62 is a low end consumer model with a single 1/6" CCD sensor. It will be noisy. You can do much better within the DV format up to broadcast cams.

[edDV]

I was thinking about using S-VHS because it is analog, so I can avoid the compression issue. Then I could buy a TV Card and use that to create a digital file that has lossless compression.

In looking at S-VHS, I noticed something interesting: S-VHS stores uncompressed data equivalent to 560×480. After converting to digital, at 30 fps this would be the equivalent of 560*480*3*8*30*10e-9 Mbps = 193.54 Mbps...which apparently is written into S-VHS. That would be something like Panasonic AG450.

S-VHS isn't that good. You are calculating a capture as if it was uncompressed RGB. The U and V chroma components have only 0.5MHz bandwidth (equiv ~ 50x480 or ~12 to 16 luma pixels wide). S-VHS is also full of cross color and cross luma artifacts and unprocessed S/N is below 40.

Luma capture would actually be at 720x480, 640x480 or 352x480 with typical capture hardware.352x480 would seriously subsample.

You will find 4:1:1 MiniDV has much better S/N (60's possible) and better chroma bandwidth. Better yet would be a 4:2:2 broadcast camera. Used 4:3 models are available at reasonable cost as TV stations dump them for wide screen SD or HD.

[junglebeast]

You must use Firewire to get the full bitrate 720x480 DV. The USB port will get something like CIF 352x240 intended for webcam use. That looks like CIF that has been upscaled.

I will go back to the lab and check to see if they use firewire or not. However I don't see how that would be responsible for the problems I was having. The resolution and color quality were good enough -- what I cannot tolerate is large chunks of the image shifting around on every frame like a photo "collage". Do you have any idea what stage in the process might be causing this? It would be silly if I bought a new video camera that was higher quality, and still had these issues due to the problem being something else.

The DCR-HC 62 is a low end consumer model with a single 1/6" CCD sensor. It will be noisy. You can do much better within the DV format up to broadcast cams.
...
You will find 4:1:1 MiniDV has much better S/N (60's possible) and better chroma bandwidth. Better yet would be a 4:2:2 broadcast camera. Used 4:3 models are available at reasonable cost as TV stations dump them for wide screen SD or HD.

You must be talking about chroma subsampling,
http://en.wikipedia.org/wiki/Chroma_subsampling

but what is 4:3? Specifically what types of cameras are you actually suggesting?

[edDV]

what I cannot tolerate is large chunks of the image shifting around on every frame like a photo "collage". Do you have any idea what stage in the process might be causing this?

Dirty heads would cause pixelation. What you see is not normal.

You could rent a Sony VX-2000/2100 if you want to see good DV format.

Broascast models would start with the Sony DSR's or the Panasonic DVC-Pro.


Sony DSR-300A


Panasonic DVC-Pro

4:3 is aspect ratio. Most TV stations want 16:9 aspect camcorders now. This assumes a high quality need. The lens alone originally sold in the thousands.

[edDV]

This is DV format recorded off SD cable