What's the difference between DV and lossless Huffyuv in a PAL contest?

[elmuz]

Hi all,
I was discussing with friends about this... I think there isn't much difference between a DV-PAL backup rather than a Huffyuv AVI encoding. Here's the task:
We have some mpeg2 music videoclip, coming in VOB, ripped from major label Press DVD. They come in standard PAL, 720x576@50i. We drag them into Virtual Dub and we apply some picture filter (like resize/crop and other few things), we cut head and tail and we export a backup copy in AVI.
If we use DV (we are in a proper PAL situation) we just resample colorspace in 4:2:0 (actually it's already in that way) and we don't touch anything else. As far as I knew DV codec save bits only by sampling colorspace, right? Every field is stored indipendently from others...
What benefits (in terms of quality) should I take by choosing Huffyuv codec?
I always knew Huffyuv is a good choice when resolution isn't PAL or NTSC compliant, right?
Thank you in advance

elmuz

[poisondeathray]

As far as I knew DV codec save bits only by sampling colorspace, right? Every field is stored indipendently from others...

What benefits (in terms of quality) should I take by choosing Huffyuv codec?

DV is intraframe only, but still uses DCT compression . You can see differences in quality from original source if you look closely

Huffyuv's decoded image is bit for bit identical - it's lossless

[edDV]

Hi all,
I was discussing with friends about this... I think there isn't much difference between a DV-PAL backup rather than a Huffyuv AVI encoding. Here's the task:
We have some mpeg2 music videoclip, coming in VOB, ripped from major label Press DVD. They come in standard PAL, 720x576@50i. We drag them into Virtual Dub and we apply some picture filter (like resize/crop and other few things), we cut head and tail and we export a backup copy in AVI.
If we use DV (we are in a proper PAL situation) we just resample colorspace in 4:2:0 (actually it's already in that way) and we don't touch anything else. As far as I knew DV codec save bits only by sampling colorspace, right? Every field is stored indipendently from others...
What benefits (in terms of quality) should I take by choosing Huffyuv codec?
I always knew Huffyuv is a good choice when resolution isn't PAL or NTSC compliant, right?
Thank you in advance

elmuz

PAL DV applies 5x DCT intraframe compression with 4:2:0 chroma sampling @ 25Mb/s (plus audio) bit rate.

HuffYUV uses a "lossless" compression so decoded output equals input, but the bit rate is much higher and the decode is more CPU intensive.

Now to your work flow. I'm not clear where you are encoding to DV or huffyuv? On export?

You have DVD MPeg2 source which will be converted to uncompressed RGB for resize and clipping.

Then you export to DV or huffy? What is the next step? How will these clips be used?

[elmuz]

Thanks you!

1) What do you mean by 5x DCT? Well, I know what DCT is... but why 5 times? Once you are in the frequency domain what do you do?

2) Going back to the problem, from VDub we export clip into DV AVI, we store it in case we need for some edit stuff (NLE jobs, like Premiere compositions in a DV/HDV timeline)...
Let's say we don't have problems of free space for storage (even if actually we have), or it just a temporary file, we should work with Huffyuv codec to achieve from Premiere better quality?

[jagabo]

1) What do you mean by 5x DCT?

Uncompressed PAL YV12 video is about 125 Mb/s. DV is about 1/5 of that.

[poisondeathray]

2) Going back to the problem, from VDub we export clip into DV AVI, we store it in case we need for some edit stuff (NLE jobs, like Premiere compositions in a DV/HDV timeline)...
Let's say we don't have problems of free space for storage (even if actually we have), or it just a temporary file, we should work with Huffyuv codec to achieve from Premiere better quality?

Actually not necessarily better quality. Depending on what type of editing you are doing , and your source characteristics , it may result in even worse quality. It's because of how premiere handles huffyuv .

First, the quality difference is going to be very tiny from a DVD source between using huffyuv and DV-AVI. You will be hard pressed to see the differences, only if you "pixel peep" or do FX composites or keying where you really need as few compression artifacts as possible. With a higher quality source, the differences are more easily detected, but not from a DVD source.

If you use vdub filters, you will undergo an RGB conversion (you mentioned crop/resize). If you convert to RGB, you should stay in RGB colorspace. Colorspace conversions are lossy.

Premiere converts huffyuv to RGB colorspace anyway (even if you use YUY2) ; So even though huffyuv is lossless, premiere doesn't treat it as lossless. It converts to RGB using rec601 , so you can actually get some clipping if you have superbrights or superdarks in your DVD source (unlikely to have that much if it's a commercial DVD). But when using DV-AVI on a DV sequence , premiere works in YUV colorspace. It smart renders DV-AVI (only segments that have effects like transitions get rendered, everything else is passed through), but has to re-render the entire huffyuv sequence even with cuts only editing (but the only quality loss from re-rendering huffyuv would be from colorspace conversions, not compression losses - since huffyuv is lossless)

In your particular case, I would use DV-AVI . If you did your crop/resize in avisynth you can stay in the same colorspace (e.g. use fast recompress mode in vdub, also beware of proper interlaced resizing). If you had a higher quality source (other than DVD) , or were doing FX work, I would consider using huffyuv or some other lossless format

[edDV]

Thanks you!

1) What do you mean by 5x DCT? Well, I know what DCT is... but why 5 times? Once you are in the frequency domain what do you do?

DV is a lite compression format. Y (luminance) is captured full D1 at 720x480i. Chroma is maintained or converted to 4:2:0. Then each frame is compressed 5 times using DCT CBR. That gets you a 25 Mb/s bit rate for video. Audio is handled separately as uncompressed PCM 16bit/48Khz stereo(or 12bit/32KHz for 4 channels)

2) Going back to the problem, from VDub we export clip into DV AVI, we store it in case we need for some edit stuff (NLE jobs, like Premiere compositions in a DV/HDV timeline)...
Let's say we don't have problems of free space for storage (even if actually we have), or it just a temporary file, we should work with Huffyuv codec to achieve from Premiere better quality?

I agree with above but you didn't say what you did with the clips after Virtualdub.

Alternate would be to do all of this in Premiere. In that case you would import DVD MPeg2, render the timeline to RGB, apply clips and filters, then output to selected format. Similar process.

True Premiere can be configured for Y, Cb, Cr with some YCbCr filters but this is mostly done within an uncompressed Premiere project. Most will work in RGB space for "workspace render" and filters.

Bottom line, I don't see adavantage to huffyuv into Premiere for your workflow.

[elmuz]

thank you both...