search

oyvindln / vhs-decode

Software defined VHS decoder - Fork (maybe temporary) of the ld-decode Laserdisc rf decoder
GNU General Public License v3.0
1.12k stars 43 forks source link

Betamax PAL: Chroma artefacts in bright scene #92

Closed davidvankemenade closed 1 year ago

davidvankemenade commented 1 year ago

Betamax PAL seems to work well generally with vhs-decode, but I came across a specific scene with chroma artefacts in the tbc files and the mkv from gen_chroma_vid.

RF capture through DdD: https://drive.google.com/file/d/1MrCZNKYiI4R1SLcS8kQNH37Qh7fZUKRy/view?usp=drive_link Video showing gen_chroma_vid output: https://youtu.be/BGLGds2iA8Q Video showing analog video capture: https://youtu.be/g2VQmMFTGH4

Other scenes (e.g. the first and the last bit of the capture) are fine, only the middle song exhibits the chroma artefacts. They seem to coincide with wide shots with a lot of backlight.

Is there anything I can do to avoid these artefacts, or do these scenes show up any known limitations of the current software?

vhs-decode command line: ./vhs-decode -s 3085 -l 150 --no_resample --level_detect_divisor 1 --fallback_vsync --chroma_trap --recheck_phase --pal --threads 1 --tape_format Betamax /home/user/capture.flac /home/user/BetamaxDecode

I've also tried with --cafc, omitting --chroma_trap, with -nocomb and with --clamp, but none of these seem to have an impact on the chroma issue.

Sample frame from ld-analyse: DollyDotsFrame5vhs-decode Sample frame from video capture: DollyDotsFrame5vidcap

atsampson commented 1 year ago

With the options you've listed above, the TBC struggles all the way through that clip. With just vhs-decode --no_resample --pal --clamp --tape_format Betamax, it starts off OK, then the middle shot has some noise overlaid on it (which looks like it's very near the subcarrier frequency - you can see it beating with the chroma burst), then the noise fades out as they fade back to the other camera...

Assuming it's a fault in the original recording, I wonder if we might benefit from a sharper filter in front of ld-chroma-decoder's burst phase detector?

oyvindln commented 1 year ago

Will have to investigate the sample but I'm spotting some black streaks/"fishes" as well suggesting that rf filters might need some more tweaking so that may also be causing problems.

davidvankemenade commented 1 year ago

Thanks for looking at this @atsampson.

With the options you've listed above, the TBC struggles all the way through that clip. With just vhs-decode --no_resample --pal --clamp --tape_format Betamax, it starts off OK, then the middle shot has some noise overlaid on it (which looks like it's very near the subcarrier frequency - you can see it beating with the chroma burst), then the noise fades out as they fade back to the other camera...

I've tried your shorter vhs-decode and that indeed runs fine.

Assuming it's a fault in the original recording, I wonder if we might benefit from a sharper filter in front of ld-chroma-decoder's burst phase detector?

How would this be a fault in the original recording? This scene plays back fine when using normal video capture.

A sharper filter might indeed be beneficial?

oyvindln commented 1 year ago

I think this should be mostly fixed by 844ea5d, at least it looks fine when testing here. (The portion in the uploaded sample does have the typical artifacts from starting a recording in the middle of an existing recording without a flying erase head but that's visible on the conventional output too.)

davidvankemenade commented 1 year ago

Hi @oyvindln, many thanks for this fix! I've also tested it and everything seems perfectly fine now. I'll close this issue.