BetacamSP no chrominance signal in top field

[johnkins]

Hello all,

Have stumbled across an issue that I haven't seen before and hoping someone has knowledge of/fixed the same problem.

I have been capturing some BetacamSP this week through a PVW-2800 (component output). I was confused when my image was looking very undersaturated even though analysis with vectorscope was showing good levels. After monitoring the image with side-by-side fields I realised that my top field has no chrominance signal, while the bottom field looks to be fine.

This issue hasn't been happening with all the tapes I have so I'm hoping it's not a major issue with the VTR. Has anybody else had this issue or could give advice on whether it can be fixed.

See an example in the image I've attached.

[pandy]

Weird indeed... i think you can safely interpolate chroma from second field (perhaps deinterlace luma first then interpolate chroma then reinterlace chroma+luma back).

[johnkins]

Weird indeed... i think you can safely interpolate chroma from second field (perhaps deinterlace luma first then interpolate chroma then reinterlace chroma+luma back).

Thanks for the reply. Is this a process that would happen in post? Not sure what software I would use to do this.

Have also confirmed that it appears to be the tapes themselves and not the VTR. Have tested several known-good tapes and these all correctly display chroma in both fields.

[jagabo]

Weird indeed... i think you can safely interpolate chroma from second field (perhaps deinterlace luma first then interpolate chroma then reinterlace chroma+luma back).

Thanks for the reply. Is this a process that would happen in post?

Yes.

Not sure what software I would use to do this.

AviSynth or VapourSynth. Some NLE's might let you do it with a few tricks.

A simple AviSynth script to copy the chroma from the bottom half of the image to the top half:

ImageSource("Field-view.JPG") 
ConvertToYV24()

top = Crop(0,0,0,height/2)
bot = Crop(0,height/2,-0,-0)

top = MergeChroma(top, bot) # merge luma from top with chroma from bot

StackVertical(top,bot)

[Attachment 61708 - Click to enlarge]

[johnkins]

A simple AviSynth script to copy the chroma from the bottom half of the image to the top half:

ImageSource("Field-view.JPG") 
ConvertToYV24()

top = Crop(0,0,0,height/2)
bot = Crop(0,height/2,-0,-0)

top = MergeChroma(top, bot) # merge luma from top with chroma from bot

StackVertical(top,bot)

Very simple. Thanks so much for that. I of course have sent the example with the fields separated. Is it easy enough to alter that script to copy the chroma from the bottom field to top field in an interleaved video? example pic attached.

[poisondeathray]

The interleaved fields png image does not display the same characteristics as the image in the 1st post in terms of chroma signal. Both fields have chroma signal, as if the signal has been comingled between scanlines. It might be from the way you converted to RGB for the png image


Post #5 Top field

[Attachment 61711 - Click to enlarge]

Post #5 Bottom field

[Attachment 61712 - Click to enlarge]

But it would look like this, for the interleaved version

#your source
T=AssumeTFF().SeparateFields().SelectEven()
B=AssumeTFF().SeparateFields().SelectOdd()
Tfix = MergeChroma(T, B) 
Interleave(Tfix, B)
Weave()

[johnkins]

The interleaved fields png image does not display the same characteristics as the image in the 1st post in terms of chroma signal. Both fields have chroma signal, as if the signal has been comingled between scanlines. It might be from the way you converted to RGB for the png image

Interesting, the original PNG file definitely has the top field chroma issue. Perhaps something to do with compression upon uploading here? Regardless, thank you for the script. Will give this a try and report back.

[poisondeathray]

subsampled YUV has to be converted to RGB in a different manner , when the fields are interleaved

When fields are separated (such as in the 1st post top and bottom, or if they arranged were side by side), they can be treated as "progressive" in terms of resampling

[jagabo]

Perhaps something to do with compression upon uploading here?

Your source was probably interlaced YV12 and it was converted to RGB as if it was progressive YV12. That would have caused the chroma blending seen in the image.

Simply copying the chroma will result in misalignment of the luma and chroma when there is significant motion. If you can provide a video sample we can give you a script that uses motion interpolation to replace the missing chroma (as Pandy suggested earlier). That may give better alignment.