Lines on scene changes - NTSC LD upscaling

[adhouse13972]

Hi

I'm using a Pioneer CLD -D925 to upscale some LDs. I have it going through my Panasonic DMR-EX99V and then to ELegato HD60 capture card at 1080p.
PAL is fine but on some NTSC discs I have a problem with lines on some scene changes. See pic. Any idea how to get rid of these??

Thanks!!!

https://imgur.com/a/qAdHR

[johnmeyer]

Do you see this when you watch the video? It looks like nothing more than what you often get at scene changes on interlaced material. What happens is that the capture card combines one field with the other and is off by one field. This makes 0.0% difference when watching because there really isn't such a thing as "field matching," meaning that fields just get shown one after another, and it is only when you freeze a frame that the two fields, each from a different moment in time, are combined together. When this is done in the middle of a scene, and when there is lots of motion, you see this same sort of frame, with interlace combining. People who are new to video freak out about this, but is doesn't matter one iota when the video is displayed. The same thing is true at scene transitions.

Bottom line: nothing needs to be done.

[adhouse13972]

Thanks for the reply. Yes, it is noticeable when watching the video but only on the scene changes. Some NTSC discs don't have this problem....

[LMotlow]

Upscaled from SD? I don't get it. Can't your player or TV upscale during play?

Capped at 1080p? Why is there interlaced combing with progressive video?

[adhouse13972]

Upscaled from SD? I don't get it. Can't your player or TV upscale during play?

Capped at 1080p? Why are there interlacing artifacts with progressive video?

I'm trying to transfer my laserdiscs to 1080p, not watch them on the TV

[LMotlow]

Same question. Progressive video doesn't have "fields". And my first question wasn't answered: can't your players and graphics card upscale?

[adhouse13972]

Unfortunately, I don't have a video input on my PC so have to use a USB capture card to record

[LMotlow]

Most people do use capture cards to record analog source (LD is analog).
Again:
1080p is progressive video. There are no interlaced fields. I'm betting some of those NTSC videos are telecined, not interlaced.
Why upscale for capture and get upscaling glitches and quality issues? Can't your PC players and graphic cards upscale during play?

[johnmeyer]

NTSC laserdiscs are all 29.97 fps, with 24 fps material telecined.

[adhouse13972]

NTSC laserdiscs are all 29.97 fps, with 24 fps material telecined.

Thanks. I don't really understand. Is there any way for me to capture this?

[johnmeyer]

All NTSC analog video, whether broadcast, videotape, or laserdisc is hard-wired to 29.97 fps. In order to show movies that were recorded on sound film at 24 fps, the film was re-timed to 23.976 fps and then every third field (field, not frame) was duplicated. This produced video that, when played at 29.97 fps, had the correct speed, but could be played on television sets that were also hard-wired to display video at the single frame rate of 29.97 fps.

You can find out as much as you want about this subject by Googling either "video telecine" or "video pulldown." The word pulldown was used to describe the duplicate fields that were added to the film.

There is absolutely nothing special or different you have to do to capture 24 fps material that has been telecined to 29.97 because, as far as the capture hardware is concerned, it is just another analog 29.97 fps stream. However, if you want to recover the pristine 24 fps stream, without the added pulldown fields ((highly recommended) then you need to remove that pulldown using inverse telecine software. Most people use AVISynth with a plugin called TIVTC. The AVISynth script need to perform the job is simply:

TFM
TDecimate

[jagabo]

Standard definition NTSC video is always 59.94 interlaced fields per second (it takes 2 fields to cover the entire face of the screen so this is often called 29.97 frames per second). To get 1920x1080 59.94 frames per second video from that something has to deinterlace. Comb artifacts like you are seeing indicates something isn't deinterlacing properly.

Try this: Force the DVD recorder to output 1080p at 59.94 fps. Then set the capture device to capture 1080p at 59.94 fps. They may refer to it as 60 fps.

[johnmeyer]

jagabo,

Wouldn't you agree that if the original material is from film, that the OP would be better off capturing at SD resolution (720x480); doing IVTC on the computer; and then upscaling? There is no need to deinterlace material that was film and it is, I think, always the wrong thing to do.

Of course I am still not clear if the OP's laserdiscs are all movies or whether they are videos. I still have an extensive LD collection, and I'd guess that 80% are film and therefore need to be handled the way I just described.

[jagabo]

Wouldn't you agree that if the original material is from film, that the OP would be better off capturing at SD resolution (720x480); doing IVTC on the computer; and then upscaling?

Of course. But not everybody is willing to invest that much time and effort to get the best possible results. Presumably that's why he bought all this equipment.

[JVRaines]

The OP must be using HDMI from the Panasonic because that's all the Elgato has for input. I presume the Panasonic has already auto-selected 1080p, though it can be forced from Connection/HDMI Settings/HDMI Video Format. The manual makes no mention of selecting a frame or field rate; I'm guessing you are stuck with a deinterlaced 29.97 fps. And we see from the sample image that the deinterlacer is stupid about scene boundaries. The best solution, which was offered by johnmeyer, is to use a different converter to create an interlaced 29.97 SD file and then deinterlace/detelecine/upsample in software.

[adhouse13972]

Wouldn't you agree that if the original material is from film, that the OP would be better off capturing at SD resolution (720x480); doing IVTC on the computer; and then upscaling?

Of course. But not everybody is willing to invest that much time and effort to get the best possible results. Presumably that's why he bought all this equipment.

Thanks. Yes I am willing to spend the time for the best possible picture.
I'm transferring music videos

Thanks for both of your recent help. SO I capture in SD and then upscale? Forgive my ignorance but can you poiint me in the direction of how to do the upscaling etc?
Thx

[johnmeyer]

Capture using the S-video or composite output from the LD player. For various reasons, composite will give you results that are just as good as S-video (unlike videotape where you should always use S-video, if available).

Then if, and only if, the LD was from a movie (i.e., it was shot at 24 fps on movie film), then perform inverse telecine (IVTC) using TIVTC, an AVISYnth plugin. All you need to do is install AVISynth, write the following three lines in a Notepad text file, save it as IVTC.AVS (or any other name, as long as it has the AVS extension).

AVISource("my video.avi")
TFM
TDecimate

You then open this AVS file in a video editor, just as if it were a real video file. VirtualDub is often used for this.

If you capture to something other than an AVI file, you'll have to replace the first line with something else. Also, you obviously need to insert the name of your video file in place of "my video.avi".

That AVISynth code will give you a pristine, progressive video file that plays at 23.976 fps. This is the NTSC equivalent of 24 fps, and is what you want. You can then apply all sorts of restoration to the video (gamma, color, dirt removal, grain reduction, etc.), or you can simply upscale it to 1920x1080. Remember, however, that the up-scaling will not improve the video in any way whatsoever, but will greatly increase the file size.

[jagabo]

To upscale use nnedi3 in AviSynth.

AVISource("my video.avi")
AssumeTFF() # or AssumeBFF()

TFM() # match fields
TDecimate() # discard frames duplicated by 3:2 pulldown

ColorMatrix(mode="rec.601->rec.709" # SD colors to HD colors

# picture adjustments and noise reduction go here, depends on source

nnedi3_rpow2(2, cshift="Spline36Resize", fwidth=960, fheight=720) # for 16:9 use 1280x720
# mild sharpening goes here, depends on source
nnedi3_rpow2(2, cshift="Spline36Resize", fwidth=1440,  fheight=1080) # for 16:9 use 1920x1080
# mild sharpening goes here, depends on source

That will give you better upscaling than any player or TV.

For truly interlaced material replace TFM() and TDecimate() with QTGMC(). PAL/NTSC conversions will require different handling, usually something like QTGMC().SRestore().