Working with ProRes 422HQ 1920 x 1080 30P and Final Cut

questions about practical use of Neat Video, examples of use
Post Reply
picree
Posts: 7
Joined: Wed May 20, 2015 2:19 am

Working with ProRes 422HQ 1920 x 1080 30P and Final Cut

Post by picree »

Using NeatVideo plugin v4. I am using Final Cut Express on Mountain Lion and am trying to process some professionally rendered ProRes 422HQ material at 1080 30P. It was rendered from 24 fps 35mm video. I am interested in removing the film grain but ALSO I need to remove the original dust and scratches.

Unfortunately, FCE must use the Apple Intermediate Codec which "sees" the input material as 1080i. So NV only gets handed a 1920 x 540 field to work with. It seems like the unintended deinterlacing would mess up the ability of NV to remove dust and scratches as it's only seeing a synthetic field from an original frame.

Would it be better if I invested $300 in Final Cut Pro so I could use ProRes 422 as the intermediate codec? Would that give NV a better shot at removing Dust and Scratches? As it is, in FCE removing dust and scratches works "OK" but not super. They mostly are still there but just less intense.

If I understand Final Cut Pro correctly my workflow would then be to import and process ProRes 422HQ at 30P and it would then directly hand off the 30p frames which were rendered from the original 24 pfs video directly. Would that be less confusing for NV? Right now my workaround that makes FCE hand NV a full frame is to change the Field Dominance from Lower (odd) to None (but I think this has resulted in some memory issues and crashes with FCE which has made this workaround basically unworkable...but it does force FCE to hand a full frame to NV).
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

The way FCE sees the clip depends on FCE project settings, not on codec. FCE project settings determine whether it works with interlaced or progressive data. If the project is progressive then the data from input clips are deinterlaced by FCE to match the properties of the project.

I am not sure using full frames is necessary though. It should work with interlaced data too.

You mention that the original 24 fps video is converted into a 30 fps clip. That most likely means that some frames are repeated in that 30 fps clip. Which suggests using the Slow Shutter mode in Neat Video 4. Please check the user guide to see how it should be used. That should also help the Dust and Scratches filter reduce those scratches.

Remember that when using Slow Shutter it is generally better to increase the temporal filter radius too.

Hope this helps,
Vlad
picree
Posts: 7
Joined: Wed May 20, 2015 2:19 am

Post by picree »

Thank you for the quick and thoughtful reply.

But this is the problem. The source material really is progressive. So the dust and scratches were originally on progressive frames (with a few new frames thrown in to get from 24fps to 30fps). But since I am using FCE it does not interpret it as such. It won't let me set the project/sequence at 1080p. The Apple Intermediate codec converts it to 1080i.

My fundamental concern centers around the unnecessary deinterlacing done by FCE which can only use the Apple Intermediate Codec. The AIC was never designed for 1080p. I have read that, when faced with 1080p, AIC will see and process it as 1080i. Even so, FCE can import, process, and export 1080p just fine (or it appears to). The NV preview window gives a clue that all is not well since it displays 1920 x 540 because, I guess, AIC/FCE said it’s interlaced.

If FCE de-interlaced something progressive that didn’t need de-interlacing and then handed only one field (half the dust or half the scratch) to NV would that confuse NV? The easy answer here may be, “Well, I’m not sure so just buy FCP”, but of course that is another $300 (I’m assuming FCP can process internally with native ProRes 422HQ and solve this problem if it is a problem). I was hoping the Neat Video team or someone might know enough about NV coding and processing to confirm. What would NV do if it was analyzing what it thought was interlaced video and one half of the scratch was in one field and the other half was in the next field?

There may not be any way to know since FCE/AIC is modifying the 1080p source. I was very excited to see the dust/scratch removal tools in v4. Sadly my first impression was, “Well, it sort of works. On the most aggressive settings it is at least “less bad””. When comparing my exports to the originals there are some OBVIOUS one-frame hairs and blobs that, I would think, NV could have (and should have) easily removed. But if FCE/AIC is only giving it a 1080 x 540 field then maybe not?

In most cases interlaced video is what it is and it brings its own set of challenges. Of course NV is designed to handle this as best it can. But what would it do if it was only seeing half of a progressive frame that it thought was interlaced? It’s easy for me to think it would work best if I could reassemble the two fields and present NV with one original frame that would have the original dust and scratch all there. Then the only issue to deal with is the repeated frames going from 24 fps to 30 fps (Slow Shutter as you say).

So, this is where I really need help. If I spend $300 on FCP so I can process ProRes natively at 1080p all the way through will NV do a noticeably better job at removing dust and scratches? Will it do better at finding dust and scratches if I can present both fields at the same time in one frame OR does NV do some sort of re-interlacing magic when faced with interlaced video to put the fields back together before looking for temporal inter-frame dust and scratches? Sorry, maybe I’m just thinking too hard (but after spending almost $2,000 to have these videos transferred and encoded I want to make sure I edit them the best possible way…even if I have to spend ANOTHER $300 on FCP!).
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

picree wrote:If FCE de-interlaced something progressive that didn’t need de-interlacing and then handed only one field (half the dust or half the scratch) to NV would that confuse NV?
Do you mean "If FCE re-interlaced"?
I think NV should still work correctly. You just need to be careful of those repeated frames inserted to increase the frame rate. That could cause problems and specifically for Dust and Scratches functionality, which is why we have that new Slow Shutter mode in NV4.
picree wrote:I was hoping the Neat Video team or someone might know enough about NV coding and processing to confirm.
We do but without knowing everything about the project and clip you work with it is rather difficult to say anything definite. I guess the best way to check that would be take a look at some small test case (a project with one short (2-3 sec) clip and one effect applied to it). Can you prepare such a test case and send it to support [at] neatvideo.com?
picree wrote:What would NV do if it was analyzing what it thought was interlaced video and one half of the scratch was in one field and the other half was in the next field?
It would reduce them both.
picree wrote:There may not be any way to know since FCE/AIC is modifying the 1080p source. I was very excited to see the dust/scratch removal tools in v4. Sadly my first impression was, “Well, it sort of works. On the most aggressive settings it is at least “less bad””. When comparing my exports to the originals there are some OBVIOUS one-frame hairs and blobs that, I would think, NV could have (and should have) easily removed. But if FCE/AIC is only giving it a 1080 x 540 field then maybe not?
The only reason that I can think of is the presence of those additional frames. That could cause that problem. Perhaps something else too but repeated frames can easily do that, I saw that myself in direct tests before. However, the Slow Shutter mode can help with that.

Anyway, a test project is probably the best way to proceed. We can make a lot of guesses but it is better to take a direct look at the project, clip and how NV4 works with it. Also, on the noise profile you have built for that clip. That may be important too.

Thank you,
Vlad
picree
Posts: 7
Joined: Wed May 20, 2015 2:19 am

Post by picree »

NVTeam wrote:
picree wrote:If FCE de-interlaced something progressive that didn’t need de-interlacing and then handed only one field (half the dust or half the scratch) to NV would that confuse NV?
Do you mean "If FCE re-interlaced"?
I think NV should still work correctly. You just need to be careful of those repeated frames inserted to increase the frame rate. That could cause problems and specifically for Dust and Scratches functionality, which is why we have that new Slow Shutter mode in NV4.
No. I mean de-interlace. I must not be explaining this well or am missing something. Let me try one more time. This has nothing to do with repeated frames or the 3:2 pulldown conversion when going from 24fps to 30fps. This is not the issue and is a diversion to our conversation.

This is about the Apple Intermediate Codec not working correctly. Read this SLOWLY: :) If I import a 1080/30p clip into FCE the codec I must use is AIC. AIC does not understand 1080/30p so it treats it like 1080/30i. For most filters this probably doesn't matter but I am trying to figure out if it matters to NV and its ability to remove dust and scratches. The AIC codec tells NV the material is interlaced. So the NV preview pane displays a 1920 x 540 field. The potential problem is that half of the scratch and dust information is in the other "field" (because they truly did come from the SAME original frame). Will NV work better if it "sees" the entire progressive frame (which has all of the information for the original dust or scratch) or does NV "recombined" the two INCORRECTLY deinterlaced fields into one frame?

Maybe the question is, how does NV handle interlaced fields when performing temporal restoration for dust and scratches? Does it assume dust and scratches must have originally come from the same frame? If so, does it combine the two digital "fields" first before applying the temporal dust and scratches filter?

When faced with interlaced material does NV do "inter-field" or "inter-frame" temporal dust & scratches restoration? :? :?
Last edited by picree on Wed May 20, 2015 11:28 pm, edited 1 time in total.
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

picree wrote:so it treats it like 1080/30i.
So it in fact re-interlaces the data. I don't know how exactly it does that, perhaps that introduces additional problems. Need to see a test case.
picree wrote:Will NV work better if it "sees" the entire progressive frame
I do not think it will work better in this case unless the abovementioned re-interlacing actually lost some useful data present in the original progressive frames.
picree wrote:Maybe the question is, how does NV handle interlaced fields when performing temporal restoration for dust and scratches? Does it assume dust and scratches must have originally come from the same frame? If so, does it combine the two digital "fields" first before applying the temporal dust and scratches filter?
No, it works without combining. That is not necessary to detect and remove dust and scratches in a stream of frames or fields.

Vlad
picree
Posts: 7
Joined: Wed May 20, 2015 2:19 am

Post by picree »

NVTeam wrote:
picree wrote:so it treats it like 1080/30i.
So it in fact re-interlaces the data. I don't know how exactly it does that, perhaps that introduces additional problems. Need to see a test case.
Well-when FCE exports yes it RE-interlaces (or maybe puts the progressive frame back together). But the workflow as best I can understand is thus:

1080/30p source material --> FCE/AIC import --> AIC "de-interlaces" the progressive source material and converts to 1080/30i --> NV sees one of the de-interlaced fields and is asked to remove dust and scratches --> FCE/AIC re-interlaces the result (post filter) for H.264 export back to 1080/30p

So, NV doesn't ever see the material as progressive as the "RE-interlacing" doesn't occur until after the filter and upon export. Thus the question. If a piece of dust is in one FIELD and not the other then I worry NV won't detect it. Interlaced material just seems so problematic for digital editing!

Thanks for your help.

Curt
Zach
Posts: 38
Joined: Sat Jun 01, 2013 12:37 pm

Post by Zach »

Seems to be a good solution would be to select a small frame range where you know something is dust/scratches

run the filter on that and export an output file. See what it does.
picree
Posts: 7
Joined: Wed May 20, 2015 2:19 am

Post by picree »

Zach wrote:Seems to be a good solution would be to select a small frame range where you know something is dust/scratches

run the filter on that and export an output file. See what it does.
Oh I have done that. I have run an entire 2 minute sequence. As long as FCE doesn't crash because of memory problems (I must logoff/logon each day) it renders fine with NV. NV works great with film grain but only "OK" with scratches. The question I'm trying to answer is much deeper than "does it work".

At first I thought Vlad was suggesting I send a clip I have processed through NV for them to look at. That may not tell us much however your comment did cause me to think that maybe Vlad was suggesting I send an original clip to support for them to see what they could do with it and then compare that to the best I can do with FCE. Now that would really help.

I will try that tonight unless Vlad says no that's not what he meant.

Curt
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

I meant a test project with a test clip. So that we could check the project settings, source clip itself, the way you apply Neat Video to it, the noise profile you have built, the filter settings you use.

We would open the project, render or preview it, see the results (bad or not) and then trace them back to the source of the problem.

Vlad
Post Reply