Problems: Averaging two analog captures for noise reduction

[goemon]

Problem 1: layer() generates green/blue shift due to rounding errors
Solution: none?

Problem 2: consecutive broadcasts offset by 1 field
Solution: SeparateFields, Trim(1,0), Weave

Problem 3: consecutive broadcasts shifted positions
Solution: Crop + Addborders (how to work out alignment though?)

[stickboy]

Originally posted by goemon
Problem 3: consecutive broadcasts shifted positions
Solution: Crop + Addborders (how to work out alignment though?)Why do you need to use Crop and AddBorders at all? Layer allows you to set the position of the overlaid video.

As for figuring out the alignment, copy a frame from one video and paste it into Photoshop (or Paint Shop Pro or GIMP). Find the corresponding frame in the other video, copy it, and then paste it into Photoshop as another layer. Reduce the opacity of the top layer (to say 50%), and nudge it around. From there, it should be really easy to align the layers and count how much nudging you need to do.

Another alternative would be to use Crop + AddBorders + Subtract (+ Levels to boost the constrast).

[wotef]

i never seen no colour problems with 2 layers, and i'm very sensitive to green tints

i see lots of "vomit" saturation problems with NTSC material in general (PAL myself), layer or no layer

and if i'm not happy with colour, i always make tweaks or rgbadjusts to deal with it, if an rgbadjust is necessary, then one or two colourspace conversions just have to occur

[sh0dan]

ColorYUY2 (external plugin) for AviSynth 2.0 and ColorYUV (built-in) can do the same as RGBadjust, since you can boost U/V separately.

If you decrease both U/V -> more green pixture, and if you increase only U or V, you get a more red or blue picture. Use:colorbars(16,16)
coloryuv(showyuv=true) in AviSynth 2.5 to see how the colors are mapped.

[wotef]

thanks for that tip, shodan, as always, udaman

[goemon]

Originally posted by stickboy
Layer allows you to set the position of the overlaid video.

Duh! Slap mah fro!

As for figuring out the alignment, copy a frame from one video and paste it into Photoshop (or Paint Shop Pro or GIMP). Find the corresponding frame in the other video, copy it, and then paste it into Photoshop as another layer. Reduce the opacity of the top layer (to say 50%), and nudge it around. From there, it should be really easy to align the layers and count how much nudging you need to do.

Yes its "easy" but its also very time consuming...

Another alternative would be to use Crop + AddBorders + Subtract (+ Levels to boost the constrast).

Yeah I tried this as well to find the alignment, turns out the consecutive broadcasts were shifted in a way that I couldnt figure out how to compensate for...

[sh0dan]

I guess a the problem is, that the material is shifted by less-than a pixel. This should be able to happend along the x-axis, since the sync signal from your source could also be a little shaky.

The only solution I can see is to scale up the image, crop/addborders and scale it down again, which is slow, and very tricky to get adjusted right. :rolleyes:

[goemon]

Originally posted by sh0dan
The only solution I can see is to scale up the image, crop/addborders and scale it down again, which is slow, and very tricky to get adjusted right. :rolleyes:

it would be nice to have gui for this... tmpgenc has VERY nice interface for adjusting deghosting filter, for example...

[WarpEnterprises]

This is all fun but I fear the noise reduction effect will be gone with those operations...