I was thinking about improving the quality of the capture before messing around with avisynth noise filters, sharpening, upscaling et al, and, after reading a lot of posts last night, I find a solution.
Multiple capture is THE answer… but the question is: what technique to use? Average, median, TOOT, overlay, merge… a bit of confusion, but I have an idea that must be the final solution… but before explaining it, some info could be useful…
At the end, I decided to use the PAL laserdiscs, for simple reasons: first I have not any smear-free NTSC LDs to capture - neither the JSC nor the Technidisc SWE - and the DC (japan and US) and “Faces” editions have too many DVNR related problem - if I decided to use them, maybe GOUT will be a better idea. Second, the PAL LDs, also if have some DVNR, it’s quite “light” (compared to GOUT - I can’t say how much compared to JSC or Technidisc SWE), have 20% more vertical resolution (50 lines can’t seems too much, but everything counts, in particular when you upscale), and, most important, have not IVTC issues like NTSC film originated material.
OK, after the explanations, here you are the tools I have:
- <span>at least FIVE different PAL laserdisc players (two only composite, the others also S-video)</span>
- <span>four different PAL LD trilogies editions (german, spanish, french VF, french VOST)</span>
- two copies of the spanish LDs and two of the french VF (plus the VOST)
I considered first to do multiple captures of different editions and average (or median, but this was only the initial thoughts, more about it later) them to have a cleaner original clip to work on… but I decided to not follow this idea, because the colors are quite different between them (look at the picture below to see what I’m talking about), and the resulting clip may have different colors between scene and scene… not talking about the (few) scenes with alien subtitles…
Then I considered to pick up the best edition, and capture it with different players… but in this case, when there are drop outs (and believe me, there are quite many), they will be present also in the resulted clip.
So here you are the solution - which seems the better compromise: I will use three french LD trilogies (one VOST and the other two VF) because I have three of them (needed if I want to use the TOOT filter), the video quality is good (if Moth3r used them, there will be a reason…) and last but not least, they are CAV (yes, like the german LDs, but not like the spanish ones).
The first capture will be the VOST, to retain the english audio and let the subsequent audio dubbing easier (than using the french audio track), and the others will be VF1 and VF2. The three resulting caps will be filtered with the TOOT avisynth filter, which choose the two more similar pixels out of the same pixel in the three capture, and discard the other. With this filter, I’ll obtain a drop-out free clip. (Take a look at the same following picture - <span style=“line-height: normal;”>thanks to Karyudo):</span>
<span style=“line-height: normal;”> </span>
Then, I will repeat the operation three times (needed if I want to use a median filter and useful with the other filters), with three different players - the CLD-D925, the CLD-2950 and the third… maybe the CLD-600 or the LD-V4300D, passed though the DVR-320S Pioneer DVD recorder which has a good comb filter.
At the end there will be three resulting clips from nine captures, drop-out free thanks to the TOOT filter. So now I have to decide which technique to use to blend them together… median, average, overlay, merge… reading the forum, it seems that median procedure is slightly better than average, and I think the same, but opinions are welcome and needed!
Thanks to everyone.