When to clean DV source?

[Karovaldas]

"DV to DVD guide" (thanks, by the way) recommends that we edit the DV.avi material BEFORE we apply cleaning to it and encode.

What are the benefits of doing this as opposed to cleaning the DV.avi and editing as well as encoding the "clean" *.avi file?

Here's what I am thinking: if there's imperfection in the source, wouldn't it get amplified in the process of editing? I am using Premiere and superimposing videos on one another combines the noise from both sources. To my mind, it is much harder to clean-up this aggregated noise than it is to clean the source *.avi files ahead of the time.

But then I don't have a clue about the inner workings of the editing process and maybe Premiere actually cleans as it mixes--sometimes, if you add two negatives together you end up with a positive. Is this the case here?

Someday I'll run a test and compare the two ways, but for now, I am working on a project where I cleaned all of my source DV.avi with Conv3D filter before editing them together in Premiere. Do you think this makes any sense to do?

Thanks.

[bb]

I'm not sure how much you can gain when filtering the video before editing. At least you should denoise before the MPEG-2 encoding step, because the DV encoding itself introduces artefacts an MPEG-2 encoder will "see" as noise. Applying the filters twice costs much time and may destroy too much detail.

I believe that superimposing two noisy videos doesn't increase the noise level. The relative noise, i.e. the "noise percentage", should stay the same. Think of a crossfade, for example. When the first noisy video vanishes, its noise vanishes, too, while the second video's noise is zero in the beginning (because the transparency is 100%) and gets more and more amplified together with the "payload" video itself. So during the crossfade the situation is as follows:
- beginning: 100% noise video A, 0% noise video B
- middle: 50% noise video A, 50% noise video B
- ending: 0% noise video A, 100% noise video B
The noise percentage equals the video opacity, by the way. As you can see, the noise total remains constant.

Noise can get amplified by multiple edits because of the recompression, though. You may consider cleaning noisy scenes, e.g. low-light shots, before editing. But that makes the whole process much more time-consuming.

bb

[jkwarras]

Originally posted by Karovaldas
I am working on a project where I cleaned all of my source DV.avi with Conv3D filter before editing them together in Premiere. Do you think this makes any sense to do?

The fact is that when you work on a NLE system with compressed video data as DV (5:1), you should always avoid any recompression. For ex. if you do your editing, and then you apply some fade in/out, transitions, adjusting brightness/colours, blabla Premiere will have to rewrite again the data. You should always check that Premiere don't recompress (is there under options, uncheck 'always recompress'). IN this case, Premiere after you do your editing and export/render your movie to a new file will *only* recompress the places where you applied some filters.

For cleaning, if you apply the cleaning *before* the editing, make sure you apply it and export it to a losless format (like huffyuv) so, your video after the filtering will look exactly as teh source one but with cleaning. if you don't your video will be recompressed, and you'll have filtering+recompression artifacts (+ originals artifacts from original recompression). If you edit it and mix it with another filtered video, then you'll have twice recompressed video + artifacts.

So, maybe the best way is to edit your video without cleaning. Then you could:

1) Export to a losless format (huge file): Then apply your filtering once (all at the same time) and export to a DV file. (Only one compression)

2) Edit your movie: framserve to an avisynth file (you ca do it with premiere I think), then apply filtering and export to DV file. (Only one compression).

Hope it helps.