Can consumer-level deinterlacers cope with film and video in the same frame?

NicolasB

Distinguished Member
Joined
Oct 3, 2002
Messages
6,862
Reaction score
1,247
Points
1,390
Location
Emily's Shop
Imagine we have a film image onto which some helpful TV channel has superimposed a video image - for example, a report shot on film, with a video "news ticker" running across the bottom of the screen.

Most deinterlacers have to begin by deciding whether they're in "film mode" or "video mode". If it's film, they weave the entire frame. If it's video, they either bob or weave depending on whether there is motion since the last equivalent field (with more sophisticated deinterlacers testing for motion between frames on a per-pixel basis).

But this approach isn't optimal for mixed film and video images. If you treat the whole frame as video, you lose resolution in the film region. If you treat it as film you get combing in the video region. Ideally a deinterlacer ought to be able to make the initial film/video decision on a pixel-by-pixel (or at least region-by-region) basis rather than only frame-by-frame, and then apply film-mode deinterlacing to some parts of the frame, and per-pixel motion-adaptive to other parts.

Are any consumer-level VPs capable of doing this?
 
Yes, pretty sure the Radiance & vp50 & vp50Pro can all do it. So can the abt102 addon board for the vp30/vp20 which might mean that the abt1020 (is that the right product code?) chipset can as well.

Not sure about the EDGE

edit: just saw your post in the SkyHD judder thread & understand where you're coming from now. Let's say that the deinterlacer did have to decide it was dealing with video on a frame basis first (not sure that's true by the way) ... if it has support for mixed mode content, then there's no reason it can't still weave pieces of the image which it later decides on a pixel or region basis should be handled as film as it will still have both original fields available.
 
Last edited:
Imagine we have a film image onto which some helpful TV channel has superimposed a video image - for example, a report shot on film, with a video "news ticker" running across the bottom of the screen.

Would Bloomberg News be a real world exmaple ?

AVI
 
edit: just saw your post in the SkyHD judder thread & understand where you're coming from now. Let's say that the deinterlacer did have to decide it was dealing with video on a frame basis first (not sure that's true by the way) ... if it has support for mixed mode content, then there's no reason it can't still weave pieces of the image which it later decides on a pixel or region basis should be handled as film as it will still have both original fields available.
The deinterlacing decision process (barring non-consumer methods like motion compensation) goes basically like this:

Are we dealing with film or video? If we're dealing with film, weave everything, regardless of whether there is motion in that part of the image or not. If we're dealing with video then we bob in regions of the picture where there is movement, and we weave in regions where there isn't movement. More sophisticated deinterlacers can do the motion-detection on a per-pixel basis, but there's still the basic difference: with film you always weave, with video you only weave when there's no motion.

To appreciate the difference, think about what happens when the camera pans. If the frame is identified as film, every pixel not directly available from the current field is weaved (woven?). If the frame is identified as video, every pixel is bobbed. If, therefore, the presence of the ticker causes the deinterlacer to go into video mode and the camera pans, you lose half of the available resolution in the background film image.

The only way to avoid this (again, barring approaches like motion-compensation) is if you can classify part of the frame as film and part as video. This goes beyond bobbing in some areas and weaving in others, it's about making the decision to bob or weave (on the basis of motion) in some areas and not making the decision at all in others. My question is if any consumer-level device can do this. (I'd be surprised if the VP50 can).
 
Imagine we have a film image onto which some helpful TV channel has superimposed a video image - for example, a report shot on film, with a video "news ticker" running across the bottom of the screen.

Nic

Using Bloomberg as an example. Yes some consumer VP's can make the ticker motion fluid, avoid stutter (even occasional) or combing etc in any part of the image irrespective of what is being displayed. Some can achieve this using Prep on 576p output of the Thomson box over HDMI whilst other can only achieve this over analogue with the SkyHD Thomson box.

AVI
 
Silicon Optix are quite clear about how HQV processing does this. The Realta and Reon do per-pixel film/video mode selection.

I always thought this was peculiar, as it's rather similar to per-pixel motion-adaptive DI. You make a decision on whether to copy ech pixel from the previous field or previous line on a per-pixel basis. Sounds like much the same thing.

I seem to recall that VRS processing does the same thing, but DVDO don't seem to be quite so uneqiuivocal about it.

Nick
 
Silicon Optix are quite clear about how HQV processing does this. The Realta and Reon do per-pixel film/video mode selection.

I always thought this was peculiar, as it's rather similar to per-pixel motion-adaptive DI. You make a decision on whether to copy ech pixel from the previous field or previous line on a per-pixel basis. Sounds like much the same thing.
Well, as I say, it's a different decision. You aren't deciding whether to bob or weave on the basis of motion, you're deciding whether or not you need to check for motion, or just weave regardless of whether there's motion or not.

I seem to recall that VRS processing does the same thing, but DVDO don't seem to be quite so uneqiuivocal about it.
For devices that claim to be able to do film in some parts of the frame and video in others, does it work on 50Hz sources, or only for 60Hz? (Obviously film/video mode detection is much easier if you can rely on 3:2 cadence to do it).
 
For devices that claim to be able to do film in some parts of the frame and video in others, does it work on 50Hz sources, or only for 60Hz? (Obviously film/video mode detection is much easier if you can rely on 3:2 cadence to do it).

Yes it appears to work at 50hz based on certain Sky channel observations (but over HDMI prep is required with the Thomson box to resolve the poor native 576p as 576i is screwed).

AVI
 
Well, as I say, it's a different decision. You aren't deciding whether to bob or weave on the basis of motion, you're deciding whether or not you need to check for motion, or just weave regardless of whether there's motion or not.

For devices that claim to be able to do film in some parts of the frame and video in others, does it work on 50Hz sources, or only for 60Hz? (Obviously film/video mode detection is much easier if you can rely on 3:2 cadence to do it).

There's a test on the HQV DVD which shows it well with a film of a sailing scene with a video text overlay. It's 60Hz but I'm 99.99% certain it works at 50Hz too.

Just thinking; I don't know for certain that it isn't just using a video deinterlace for the whole image. What would be really nice would be test material that has the excellent cafe scene on that disk combined with a ticker. That would prove it (or disprove it) quite nicely.
 
Last edited:

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom