DGDecodeNV - How good is the deinterlacing? Exemples?

[Gargalash]

Hello people,
In my work flow, I have to deinterlace DVDs. I was using MvBob for a while, then I switched to Yadif. Yadif was not satisfying me and I recently tried out TGMCmod. I have found the results to be amazing, but then I have found it to be really slow as well, as you all know... Quality comes at a time price!

I have to go through a pile of DVD (over 200). I need a fast workflow. I can't slack on deinterlacing quality as the clips created will be offered over streaming full DVD size.

I have looked around and re-discovered DGDecodeNV, searched and found it was available since the last time I heard it was in creation. I know it uses nvidia PureVideo deinterlacer. I know it's supposed to be really fast for the quality it provides. I currently have an old ATI Radeon x700.

Questions:
- I read in the manual that DGDecodeNV can be used with MPEG2 files, I am assuming that this includes DVD vobs. Am I right?

- The last time I read about DGDecodeNV it was in development and someone told me that it would be even better then MVBob. How good is the PureVideo deinterlacing?

- Can you post examples of this deinterlacer at work?

- There are a lot of GPUs with CUDA and PureVideo capabilities. I have an old Athlon X2 4400+ with 3.25gig of RAM on Win XP. Can you suggest some GPUs that will be consistently speeding things up at about 200-250$ ?

Thank you very much for your time and answers! :)

[Guest]

[bump after original post determined not a rule 6 violation and reopened]

[Sulik]

The PV deinterlacing quality depends on the GPU, so you probably want to stay away from the very low-end. 9600GT and 9800GTs are fairly cheap these days.

[Gargalash]

Hi Sulik,
Thanks for your reply. I will make sure I aim for a higher quality GPU then.

Subjectively, when you get into the higher quality GPU, can the deinterlace compare to filters like TempGaussMC or MVBob? If not, does it get near?

Thanks!

[linyx]

The PV deinterlacing quality depends on the GPU, so you probably want to stay away from the very low-end. 9600GT and 9800GTs are fairly cheap these days.
I thought PureVideo relied on the VPx Engine? And even then I thought the speed would be the only thing affected, not quality; or are the VPx chips only for decoding, and I'm making a fool of myself.:rolleyes:

[Sulik]

Afaik, VPx chips are only used for decoding, and shaders for all advanced post-processing (after all, massively parallel pixel processing is what gpus excel at).

[aegisofrime]

The PV deinterlacing quality depends on the GPU, so you probably want to stay away from the very low-end. 9600GT and 9800GTs are fairly cheap these days.

Really? I have not heard anything to support this, either from users or from neuron2 himself. Of course, any clarification will be helpful, as I'm thinking of getting an nVidia card and a license myself. (Currently using Radeon 4870)

A University friend of mine has a license, and I managed to have a go at DGDecodeNV on his computer. From what I have seen, the deinterlacing quality is very close to TGMC indeed. One quality that TGMC has is that it's very clean, and that dovetails nicely with what Didee said about denoising being a side effect. Additionally, because of it's clean output, file size is smaller of course.

I can think of a few ways to make the PV output come as close as possible to TGMC, but without the obvious speed hits.

1. Denoise after deinterlacing.
2. Sharpen
3. Use a higher quality x264 preset. For example, currently with TGMC and x264@Slower, I get 10-11 FPS. With PV deinterlacing I speculate I could probably do very slow at 25 FPS?

Personally I'm in kind of the same situation as you Gargalash, and short of getting more computing time with my friend's PC, I hope that someone else had done the same research of comparing TGMC vs PV.

[Gargalash]

aegisofrime,
Interesting input, thanks for sharing.

[zilog jones]

The PV deinterlacing quality depends on the GPU, so you probably want to stay away from the very low-end. 9600GT and 9800GTs are fairly cheap these days.

I don't know what US prices are like but a GTS 250 can be found for similar money to these (sometimes even cheaper) - it's basically a more efficient 9800GTX+ and is more easily available with 1GB memory.

[hydra3333]

Will an 8800GT do the trick ? I understood from the specs it's basically got more bits and has more throughput than some of the newer models apart from the newer VPx support ?

[aegisofrime]

Can we have more input on whether the GPU actually affects the output quality? It will influence my decision on which card to buy.

[zilog jones]

Some differences between PureVideo implementations are listed here: http://en.wikipedia.org/wiki/PureVideo#PureVideo_HD
I don't know how these relate to the deinterlacer quality though. Note the resolution limitations with VP3, which could be a problem if you're cropping before deinterlacing or doing something else weird. I'm sure differences in decoding speed will be significant between graphics cards.

I would also be interested to see how it compares to the deinterlacers Gargalash mentioned. I tried PureVideo decoder when I first got my 7600GT using Nvidia's own software (seemed to only work with WMP at the time) and was not impressed, but that was a VP1 card and nearly 4 years ago now.

[Guest]

Will an 8800GT do the trick ? I understood from the specs it's basically got more bits and has more throughput than some of the newer models apart from the newer VPx support ? Be careful as not all the 8800 cards have a VP2.

[Sulik]

I'm pretty sure all the 8800GT have VP2 (G92-based). The ones to avoid are the GTS (GTS320 and GTS512 are G80-based = VP1)

Last time I checked, HQV scores were different for different GPUs, so this would point to different processing being applied (most likely scaled according to gpu processing power)