DV AVI vs. Uncompressed

Bill Ravens · November 21st, 2007, 02:29 PM

There are also a number of lossless compressed formats available, altho they don't provide much reduction in size over uncompressed. Formats like PicVideo or HuffyUV will provide lossless reductions in total file size at the expense of CPU processing cycles. Personally, I think Cineform is the best thing going for intermediate file format.

Glenn Chan · November 21st, 2007, 02:54 PM

Some notes...

1- If you do chroma subsampling correctly, you will get generation loss (this is pretty noticeable and way more than the generation loss you get from repeated DCT transform + quantization).

If you use one of the "passthrough" chroma schemes (my term; vegas always does box reconstruction on chroma), then you won't get generation loss from chroma subsampling/resampling the chroma over and over. However, 1st generation material will look a lot worse than doing chroma subsampling the right way. (Though eventually with the right way of doing things, with enough generations you'll accumulate enough generation loss that it looks really bad.)

Right now none of the major NLEs handle chroma subsampling correctly. They all do the passthrough schemes.

One situation where doing the chroma the wrong way will bite you in the butt is if you have a red title and you render that to DV. If you did it the right way then it wouldn't look nearly as bad. (You can sort of see this if you change the Vegas prefs to use Microsoft's DV codec. You will need to restart Vegas. But don't do that for real projects, you will likely get worse results than using Vegas' codec.)

2- I remember the Vegas DV codec holding up very/extremely well over 50 generations (looks very similar to the original).

John Cline · November 21st, 2007, 05:13 PM

Technically, it would be best to render straight from the timeline, applying all transitions, filters and color corrections, directly to MPEG2 and bypass the AVI step altogether.

NTSC DV is sampled in the 4:1:1 colorspace, but MPEG2 is sampled at 4:2:0, and to complicate things, in each line only one color difference channel is stored with half the horizontal resolution. The channel which is stored flips each line, so the ratio is 4:2:0 for one line, 4:0:2 in the next, then 4:2:0 again, and so on. This leads to half the horizontal as well as half the vertical resolution, giving a quarter of the color resolution overall. 4:1:1 DV only has 1/4 the color resolution relative to the luminance resolution to begin with, and 4:2:0 is also 1/4 chroma resolution, but when going from 4:1:1 to 4:2:0 some of the chroma samples don't line up in the conversion and they get lost, consequently, a DVD MPEG2 file encoded from a DV AVI file only has 1/8th or 12.5% of the original color information left.

Transitions, color correction, filters, generated media and titling in Vegas is all done in the RGB colorspace, which is 1:1:1, or more accurately, 4:4:4. So any of these things wouldn't take as big a chroma sampling hit if you rendered directly to MPEG2 from the timeline. The DV footage would still take to the 4:1:1 to 4:2:0 hit and there is nothing you can do to avoid this, but anything that Vegas has generated would go directly from 4:4:4: to 4:2:0 and look much better chroma-wise.

John

Glenn Chan · November 21st, 2007, 10:36 PM

Rendering to a 4:4:4 intermediate like Cineform and then from there going to MPEG2 might be helpful for some workflows. You avoid problems with 4:1:1 DV (as an intermediate) that way.

Anyways, there are lots of different ways to go about it.

2-
4:2:0 applied to interlaced material is screwed up.
Usually, the effective performance is kind of 2X reduction in chroma resolution horizontally, and 4X reduction vertically (not 2X).
This is un-ituitive and took me some time to digest.

The 4:2:0 information here explains why you can't apply 4:2:0 subsampling in a straightforward manner (like you would on progressive material):
http://en.wikipedia.org/wiki/Chroma_subsampling

Then read this:
http://www.hometheaterhifi.com/volum...ug-4-2001.html

For progressive footage, you can avoid that nastiness by encoding your footage correctly.

Annie Haycock · December 13th, 2007, 12:55 PM

Hi

I'm new here, and this thread seems as close as I've found to answering a question, but it's getting a bit too technical.

My normal way of working in SD was to download the whole tape into Premiere Elements (I'm not shelling out for a more expensive program until I have time to make good use of it). Then I go through and cut out a clip, put it on the timeline, then export movie to an AVI file.

Now I've got a Canon XH A1, will there be any noticeable loss of quality by doing the same thing with HDV as it involves two lots of compression? And then there will obviously be a further lot of compression when I make a complete video.

Premiere elements doesn't chop HDV into scenes, and anyway, a lot of my stuff (wildlife) involves leaving the camera running for long periods. I don't have a separate deck (I'm broke after buying the camcorder).

Thanks

PS sorry if I'm in the wrong place, I got here by doing a search on compression.

December 13th, 2007, 12:55 PM	#20
Annie Haycock Major Player Join Date: Dec 2007 Location: Pembrokeshire, Wales Posts: 734	Please give me a simple answer Hi I'm new here, and this thread seems as close as I've found to answering a question, but it's getting a bit too technical. My normal way of working in SD was to download the whole tape into Premiere Elements (I'm not shelling out for a more expensive program until I have time to make good use of it). Then I go through and cut out a clip, put it on the timeline, then export movie to an AVI file. Now I've got a Canon XH A1, will there be any noticeable loss of quality by doing the same thing with HDV as it involves two lots of compression? And then there will obviously be a further lot of compression when I make a complete video. Premiere elements doesn't chop HDV into scenes, and anyway, a lot of my stuff (wildlife) involves leaving the camera running for long periods. I don't have a separate deck (I'm broke after buying the camcorder). Thanks PS sorry if I'm in the wrong place, I got here by doing a search on compression. Last edited by Annie Haycock; December 13th, 2007 at 01:52 PM. Reason: correction

November 21st, 2007, 02:29 PM	#16
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	There are also a number of lossless compressed formats available, altho they don't provide much reduction in size over uncompressed. Formats like PicVideo or HuffyUV will provide lossless reductions in total file size at the expense of CPU processing cycles. Personally, I think Cineform is the best thing going for intermediate file format.

November 21st, 2007, 02:54 PM	#17
Glenn Chan Inner Circle Join Date: Jun 2003 Location: Toronto, Canada Posts: 4,750	Some notes... 1- If you do chroma subsampling correctly, you will get generation loss (this is pretty noticeable and way more than the generation loss you get from repeated DCT transform + quantization). If you use one of the "passthrough" chroma schemes (my term; vegas always does box reconstruction on chroma), then you won't get generation loss from chroma subsampling/resampling the chroma over and over. However, 1st generation material will look a lot worse than doing chroma subsampling the right way. (Though eventually with the right way of doing things, with enough generations you'll accumulate enough generation loss that it looks really bad.) Right now none of the major NLEs handle chroma subsampling correctly. They all do the passthrough schemes. One situation where doing the chroma the wrong way will bite you in the butt is if you have a red title and you render that to DV. If you did it the right way then it wouldn't look nearly as bad. (You can sort of see this if you change the Vegas prefs to use Microsoft's DV codec. You will need to restart Vegas. But don't do that for real projects, you will likely get worse results than using Vegas' codec.) 2- I remember the Vegas DV codec holding up very/extremely well over 50 generations (looks very similar to the original).

November 21st, 2007, 05:13 PM	#18
John Cline Major Player Join Date: Oct 2002 Location: Albuquerque, New Mexico USA Posts: 333	Technically, it would be best to render straight from the timeline, applying all transitions, filters and color corrections, directly to MPEG2 and bypass the AVI step altogether. NTSC DV is sampled in the 4:1:1 colorspace, but MPEG2 is sampled at 4:2:0, and to complicate things, in each line only one color difference channel is stored with half the horizontal resolution. The channel which is stored flips each line, so the ratio is 4:2:0 for one line, 4:0:2 in the next, then 4:2:0 again, and so on. This leads to half the horizontal as well as half the vertical resolution, giving a quarter of the color resolution overall. 4:1:1 DV only has 1/4 the color resolution relative to the luminance resolution to begin with, and 4:2:0 is also 1/4 chroma resolution, but when going from 4:1:1 to 4:2:0 some of the chroma samples don't line up in the conversion and they get lost, consequently, a DVD MPEG2 file encoded from a DV AVI file only has 1/8th or 12.5% of the original color information left. Transitions, color correction, filters, generated media and titling in Vegas is all done in the RGB colorspace, which is 1:1:1, or more accurately, 4:4:4. So any of these things wouldn't take as big a chroma sampling hit if you rendered directly to MPEG2 from the timeline. The DV footage would still take to the 4:1:1 to 4:2:0 hit and there is nothing you can do to avoid this, but anything that Vegas has generated would go directly from 4:4:4: to 4:2:0 and look much better chroma-wise. John

November 21st, 2007, 10:36 PM	#19
Glenn Chan Inner Circle Join Date: Jun 2003 Location: Toronto, Canada Posts: 4,750	Rendering to a 4:4:4 intermediate like Cineform and then from there going to MPEG2 might be helpful for some workflows. You avoid problems with 4:1:1 DV (as an intermediate) that way. Anyways, there are lots of different ways to go about it. 2- 4:2:0 applied to interlaced material is screwed up. Usually, the effective performance is kind of 2X reduction in chroma resolution horizontally, and 4X reduction vertically (not 2X). This is un-ituitive and took me some time to digest. The 4:2:0 information here explains why you can't apply 4:2:0 subsampling in a straightforward manner (like you would on progressive material): http://en.wikipedia.org/wiki/Chroma_subsampling Then read this: http://www.hometheaterhifi.com/volum...ug-4-2001.html For progressive footage, you can avoid that nastiness by encoding your footage correctly.