Capture Compression YUY2 UYVY etc

[TazRulezAll]

My capture software supports different compressions (yuy2, uyvy, rgb-16, rgb-24, y41p, yvu9, yv12 & i420). what is the difference between these and which is best?

[TazRulezAll]

I forgot to mention that i am trying to capture MPEG-2, not avi.

[Sulik]

The best is YUY2 or UYVY (native 4:2:2 YUV formats) if you're going to encode as MPEG.
YV12 and I420 are virtually identical and are 4:2:0 YUV format, only recommended if you're capturing single-field (x240 or x288).

Avoid all the others.

[trevlac]

My capture software supports different compressions (yuy2, uyvy, rgb-16, rgb-24, y41p, yvu9, yv12 & i420). what is the difference between these and which is best?

Forgive me if this sounds like a lecture.

First analog. The signal comming into your card (via s-video or rca) is YUV (PAL) or YIQ (NTSC). The black and white info = Y. UV/IQ are color difference numbers. These have a smaller range/bandwidth. A camera captures an image in RGB, where each color is full bandwidth. The signal is converted to YUV because we are less sensitive to color resolution, so the color can be transmitted at 1/2 the bandwidth.

All the Y types are different subsampled versions of YCbCr. Most people call this YUV, but YUV is more appropriately an analog colorspace. If you are in digital, you have YCbCr. Y=Luma and Cb, Cr are color difference information. The different 'fourcc' codes your mention above are different by how the YCbCr data is sub-sampled and /or how it is stored in a file. Sub-sampled is important, because you lose info here. YUY2 is 4:2:2. YV12 is 4:2:0. The latter has less color info. As far as the file storage, this is only important if your codec choice excepts the given type.

The RGB types are a different color space from YCbCr. RGB24 has the full bandwidth for all 3 colors. RGB16 does not.

My answer
You capture YUV/YIQ which gets turned into YCbCr 4:2:2 by your card. If you ask for YUY2, that is what you get and little conversion/loss took place.

The Rub: Your mpeg encoder may only take RGB. So when you specifiy YUY2 (4:2:2), if may get converted to RGB for input, and then come out as YCbCr 4:2:0, which is the color sample rate for mpeg.

So picking YUY2 seems the best, but it's hard to say. I doubt your encoder accepts YV12. That would be the best, because your card probably does the best job at down sampling the color.

[dwiesel]

I hope someone can shed some light here...please!

I am using Pinnacle AVDV Studio 9 to capture to the highest quality they allow, which is YUV2 AVI. I am also using the PCI capture card that came with Pinnacle.

I need to edit AVI files and save as a new AVI files -- my question is, which is my best choice for my target AVI format to avoid ANY quality loss? I've tried converting from YUV2 to YUY2, and from YUV2 to RGB24 (two formats that are available using Pinnacle). I see little difference, but I want to know what the QUALITY benefits/drawbacks are when converting from YUV2 to YUY2, and from YUV2 to RGB24 (not file size benefits, since I know RGB24 uses the most disk space).

I've been saving as RGB24, for the simple reason that it's compressed the least and makes me think this will result in no quality loss. But I know there are likely chroma rules and conversions going on. I wish it were as simple as just cutting frames from the captured AVI and saving as the same file without any conversion (I think Virtual Dub does this, but am not sure if quality loss happens here either, or is any better than Pinnacle's method).

Also, since I'm using TMPGenc to render this AVI to MPG, do I need to consider which AVI format I convert to?

Many thanks....the more info the better.

[donpedro]

Use same codec that you used for capture. Smartrender should kick in = no re-encoding of parts that were not modified. I can't guarantee it with your codec but DV AVI works that way. You can tell by speed if it re-encodes or not.

[SatStorm]

My small contribution on this matter...
Our Argentian friend, the one who writes those amazing btwincap drivers for our beloved bt8xx cards, in this page : http://btwincap.sourceforge.net/features.html mentions something really interest....

"There is plenty of capture video formats (all supported by DirectShow and the hardware). For example, use YV12 for MPEG capture - Performance Boost!)"

Well I did follow his advice and suprice: I was able to capture with mainconcept 1.4.1 reatime to mpeg 2 using all my bt8xx(x) cards at 704 x 576 framesize with a duron 1200 with no framedrops. Before, I was using YUV or YUV2 and that was impossible...

About quality now, with nvidia cards, it is suggested to capture at the UYVY colourspace. This is the reason that some cards, like the older asus, don't allowed any other way of capturing. Something to do with the nvidia cards.

[dwiesel]

Thank you...I want/need to capture at YUV2, and want/need to save as either RGB24 or YUY2...the annoying thing with Pinnacle is, if you can capture in YUV2, why doesn't it have the codec to save it as YUV2? And if it has a codec to save as RGB24, why doesn't it have a corresponding capture codec of RGB24? I asked them and they never answered me.

Thanks...and keep the feedback coming please.

After more research, I'm still not much more in the know...it seems any conversion between RGB24 and YUV2 should be avoided when possible, since quality loss is likely.

I did read that RGB24 is used by TMPGenc, so since I use it, I wish I had the capability in Pinnacle to capture in RGB24 instead of TMPGenc converting my YUV2 to RGB24....and ideas what software would be best to capture with? Pinnacle keeps giving me capture problems, I need something better...thanks.

**********************************

I tried capturing to RGB in ULEAD; using s-video, it was fine....using composite, it was choppy and dropped frames...the thing is, I need to use composite for most of my projects...why does the dropped frames happen only for composite capture? Thanks again.