A interlacing image kell factor, interlace factor, what is 1080i dynamic resolution?
-
LIVE: AVSForum Tech Talk Podcast with Scott Wilkinson, Episode 5 Click here for details.
-
Enter to be our Home Theater of the Month Click here for details.
- Status
1 - 20 of 34 Posts
Joined
·
8,104 Posts
HD-DVDMAN,
Not quite. I believe most 1080i content is film content. If you have a poor BOB/WEAVE deinterlacer, you can reach between 540-800 lines.
With a good film deinterlacer (e.g., Terranex) you should be able to retrieve very close to 1080 lines, depending on the strength of the vertical filter (if any) that was applied to the source material.
Not quite. I believe most 1080i content is film content. If you have a poor BOB/WEAVE deinterlacer, you can reach between 540-800 lines.
With a good film deinterlacer (e.g., Terranex) you should be able to retrieve very close to 1080 lines, depending on the strength of the vertical filter (if any) that was applied to the source material.
Hi Oferlaor, I am thinking that the scan lines are drawn this way on a fast motion image.
--_-- _ -- _
And during static scenes, due to the offset 540 line frames and the phosphor refresh characteristics (the first frame will still be "glowing" on the raster when the 2nd is drawn) you see, they are drawn this way.
= = =
I would think it is worth noting that because of the Kell factor, the effective static vertical resolution of 1080i video is not 1080 lines, but rather somewhere between 65% and 80% of that number (depending on who you believe). On moving images, the resolution is of course 540 lines because, The apparent drop in resolution is due to the interplay between the shifting motion of the scanlines and the motion in the image. There are peaks and valleys in the apparent vertical resolution depending on the vertical component of the motion at any given point in the image and in time. The "valleys" (1920x540) would occur when the vertical motion was an odd integer multiple of one pixel (out of 1080) per field. The "peaks" (1920x1080) would be at even integer multiples of a pixel per field (including 0, which is still or sideways only). When the vertical component of the motion is accelerating or decelerating, you may see a shimmer as the apparent resolution cycles between the peaks and valleys. For fast or jerky vertical motion this may break down to an apparent resolution that stays at 1920x540, but motion blur will likely come into play, making it not matter so much. These things mainly apply to full rate 60Hz video. Interlacing artifacts aren't as apparent with a film source and 3:2 pulldown. They are overshadowed by temporal effects such as judder and apparent multiple images during motion.
Where would I be going wrong here?
--_-- _ -- _
And during static scenes, due to the offset 540 line frames and the phosphor refresh characteristics (the first frame will still be "glowing" on the raster when the 2nd is drawn) you see, they are drawn this way.
= = =
I would think it is worth noting that because of the Kell factor, the effective static vertical resolution of 1080i video is not 1080 lines, but rather somewhere between 65% and 80% of that number (depending on who you believe). On moving images, the resolution is of course 540 lines because, The apparent drop in resolution is due to the interplay between the shifting motion of the scanlines and the motion in the image. There are peaks and valleys in the apparent vertical resolution depending on the vertical component of the motion at any given point in the image and in time. The "valleys" (1920x540) would occur when the vertical motion was an odd integer multiple of one pixel (out of 1080) per field. The "peaks" (1920x1080) would be at even integer multiples of a pixel per field (including 0, which is still or sideways only). When the vertical component of the motion is accelerating or decelerating, you may see a shimmer as the apparent resolution cycles between the peaks and valleys. For fast or jerky vertical motion this may break down to an apparent resolution that stays at 1920x540, but motion blur will likely come into play, making it not matter so much. These things mainly apply to full rate 60Hz video. Interlacing artifacts aren't as apparent with a film source and 3:2 pulldown. They are overshadowed by temporal effects such as judder and apparent multiple images during motion.
Where would I be going wrong here?
Joined
·
8,104 Posts
With a good film mode deinterlacer, the original frame is fully retrieved (look up 2:3 pulldown in "secrets of home theater hifi" for a good explanation). This means that even during fast motion you get 1080 lines. Any motion blur you get is in the source.
In video mode and using kell factor, you are 100% correct - the kell factor varies between 80% and 50% (50% for consistant fast motion, around 80% for completely static images). When reconstructing the frame using bob/weave you get between 65% and 100% retrieval of the original frame.
In video mode and using kell factor, you are 100% correct - the kell factor varies between 80% and 50% (50% for consistant fast motion, around 80% for completely static images). When reconstructing the frame using bob/weave you get between 65% and 100% retrieval of the original frame.
Joined
·
771 Posts
Now, wouldn't this only apply to frame rates faster than 30 frames per second? Faster than that of course you could only get two fields per frame with a stationary picture. But if the source is 30 frames per second, the resolution should stay constant.
Joined
·
8,104 Posts
HD-DVDMAN,
From what I know, most native HD sources are film mode. I think both the cameras are inherently progressive + all converted (telecine) material is definitely film mode. I suspect video mode on 1080i will primarily be important during scene transitions and bad edits.
From what I know, most native HD sources are film mode. I think both the cameras are inherently progressive + all converted (telecine) material is definitely film mode. I suspect video mode on 1080i will primarily be important during scene transitions and bad edits.
I think maybe the kell factor is also lowered for telecined fillm material if you can't be 100% sure it is really progressive source. This happens when you can't trust, or don't have, the video flags in the material. Then you have to constantly be making a decision about whether to deinterlace and thus be careful of anything that would give too much vertical detail, since it could really be comb artifacts.
And most HDTV material that I can capture seems NOT to have the repeat flags set at all. I'm not sure why.
This means they are not compressing as effectively as they could and it also means that a lot more work must be done to properly reclaim progressive film source by any scaler.
- Tom
And most HDTV material that I can capture seems NOT to have the repeat flags set at all. I'm not sure why.
This means they are not compressing as effectively as they could and it also means that a lot more work must be done to properly reclaim progressive film source by any scaler.
- Tom
Yes, kell factor is not as big of a deal with a film mode 1080i source but it is still there. With 1080i video mode you are looking at about 50% loss with fast motion. With film mode it's more like 20 to 30% maybe 15% with fast motion. In some cases all detail can be displayed 100% but, most of the time that can't be done even with a Teranex.
It's not such a big deal really with film mode, only with video mode. There is a huge difference from 50% loss to 15% loss.
It's not such a big deal really with film mode, only with video mode. There is a huge difference from 50% loss to 15% loss.
Quote:
On the other hand, the eye's resolution is better for static images than for moving ones as well.
| Originally posted by ZAGBY It's not such a big deal really with film mode, only with video mode. There is a huge difference from 50% loss to 15% loss. |
Quote:
Sorry, really can't agree! From what I know, most native HD sources are video mode.
There are 6 HD channels running 7x24 on the Japanese BS system. Except a few movie programs (film transfer to HD, can't really say those are "native HD") and maybe a few documentary shows originally shot on film (again not "native" HD), all else are native HD video sources I believe.
Unless you say all those handheld HD cameras and those large floor-stand cameras inside the studios are all film based.
regards,
Li On
| Originally posted by oferlaor From what I know, most native HD sources are film mode. |
There are 6 HD channels running 7x24 on the Japanese BS system. Except a few movie programs (film transfer to HD, can't really say those are "native HD") and maybe a few documentary shows originally shot on film (again not "native" HD), all else are native HD video sources I believe.
Unless you say all those handheld HD cameras and those large floor-stand cameras inside the studios are all film based.
regards,
Li On
Joined
·
8,104 Posts
MDRiggs,
You don't need anymore than 24p. After all, it works fine for movies.
Li On,
Not sure about Japanese HD, it's a completely different system. However, most shows in the US are now shot either on Film (and so are 24fps) or using Sony's 24fps HD camera.
Of course, there are still quite a few cameras shooting at 60fps (e.g., Panasonic), but AFAIK they are not in widespread use in the US.
You don't need anymore than 24p. After all, it works fine for movies.
Li On,
Not sure about Japanese HD, it's a completely different system. However, most shows in the US are now shot either on Film (and so are 24fps) or using Sony's 24fps HD camera.
Of course, there are still quite a few cameras shooting at 60fps (e.g., Panasonic), but AFAIK they are not in widespread use in the US.
Joined
·
11,139 Posts
Interesting topic, vertical resolution. Don't see much discussion of it online, perhaps with the exception of the OpenDTV forum at Topica.com.
Earlier in this section I posted a simplified table showing measured vertical resolution results obtained during the mid-1990s approval of ATSC HDTV. There was a difference of opinion about how relevant these data were today. But if someone has newer independent data, or can detail what's happening with 720p/1080i broadcasts currently that makes this data not relevant, please post the information. (Seems to me that U.S. broadcasters haven't yet reached the resolutions achieved during these approval tests by the FCC's experts committee. Expressed my 2 cents why I think this is so just yesterday , so no need to repeat it.)
But, going to the original query, the measured results with a 5-rpm test pattern indicated vertical resolution at 1080i was 400 lines and for 720p it was 420 lines. The committee provided some notes about vertical resolution below its table 2.3 . My assumption would be that vertical resolution at 1080i would drop to 400 only within the rotating test pattern, and that vertical resolution would increase (to 800 lines static) as the test pattern rotation slowed.
Would vertical resolution be significantly different with video motion from film originally captured at 24p or video taped at 60i (60 interlaced fields per second)? Wouldn't think it would match a simple rotating test pattern, but regardless of the original source our eyes are blending two 1/60-sec fields into 1/30-sec frames with 1080i. I'd think vertical resolution reductions would be similar to those obtained in the ATSC tests.
This recent post outlines why vertical resolution is deliberately reduced (smeared) to prevent interline flicker. Combined with a Kell factor of ~0.7 (times the scan lines) this accounts for the drop in actual resolution as opposed to scan lines. Oversampling is sometimes mentioned as an aid to boosting resolution (h and v). With oversampling, after standard filtering that gives you greater resolution. But does anyone have any reliable information on how much HBO, NBC, and ABC may be oversampling? -- John
Earlier in this section I posted a simplified table showing measured vertical resolution results obtained during the mid-1990s approval of ATSC HDTV. There was a difference of opinion about how relevant these data were today. But if someone has newer independent data, or can detail what's happening with 720p/1080i broadcasts currently that makes this data not relevant, please post the information. (Seems to me that U.S. broadcasters haven't yet reached the resolutions achieved during these approval tests by the FCC's experts committee. Expressed my 2 cents why I think this is so just yesterday , so no need to repeat it.)
But, going to the original query, the measured results with a 5-rpm test pattern indicated vertical resolution at 1080i was 400 lines and for 720p it was 420 lines. The committee provided some notes about vertical resolution below its table 2.3 . My assumption would be that vertical resolution at 1080i would drop to 400 only within the rotating test pattern, and that vertical resolution would increase (to 800 lines static) as the test pattern rotation slowed.
Would vertical resolution be significantly different with video motion from film originally captured at 24p or video taped at 60i (60 interlaced fields per second)? Wouldn't think it would match a simple rotating test pattern, but regardless of the original source our eyes are blending two 1/60-sec fields into 1/30-sec frames with 1080i. I'd think vertical resolution reductions would be similar to those obtained in the ATSC tests.
This recent post outlines why vertical resolution is deliberately reduced (smeared) to prevent interline flicker. Combined with a Kell factor of ~0.7 (times the scan lines) this accounts for the drop in actual resolution as opposed to scan lines. Oversampling is sometimes mentioned as an aid to boosting resolution (h and v). With oversampling, after standard filtering that gives you greater resolution. But does anyone have any reliable information on how much HBO, NBC, and ABC may be oversampling? -- John
Quote:
Well, it works! Fine I'm not so sure about. A higher frame rate would be less prone to motion artifacts. My main objection, however, is that it forces the use of pulldown in the U.S. and other countries whose TV systems operate at 30 and 60 fps. I think the resulting judder is pretty annoying sometimes. If you're using video cameras, it just seems wrong to use a frame rate that doesn't map directly to the display rate. I wouldn't mind so much if 72 Hz were a commonly available display rate, but it's not.
| Originally posted by oferlaor You don't need anymore than 24p. After all, it works fine for movies. |
Quote:
But at 72 Hz you would get other strange effects when watching 60i video source. I don't know if this is also called judder when it would be compounded with interlace.
- Tom
| I wouldn't mind so much if 72 Hz were a commonly available display rate, but it's not. |
- Tom
Joined
·
8,104 Posts
MDRiggs,
24fps is the lowest frame rate that movies can run at without significant judder like artifacts. The slower the frame rate, the cheaper the film becomes to produce and shoot. So, this will continue to be the film industry standard from now until eternity (or until the human visual system becomes more sensitive).
About other TV systems, trouble is - except for the US and parts of the far east, the rest of the world works in 50Hz (not 60). Should the US conform and change? Should the rest of the world?
With digital displays, frame rate will become less of an issue as it is for CRTs (which relies on the public electrical system for its frequency). At that point (in quite a few years), it's quite likely we might get a completly new standard that will be a multiple of 24fps.
Ultimately 300hz is the ideal connection point, because it translates well from 24fps, 25fps and 30fps. Once we can get a good display system that'll conform to that, this will be a good rate to standardize on.
24fps is the lowest frame rate that movies can run at without significant judder like artifacts. The slower the frame rate, the cheaper the film becomes to produce and shoot. So, this will continue to be the film industry standard from now until eternity (or until the human visual system becomes more sensitive).
About other TV systems, trouble is - except for the US and parts of the far east, the rest of the world works in 50Hz (not 60). Should the US conform and change? Should the rest of the world?
With digital displays, frame rate will become less of an issue as it is for CRTs (which relies on the public electrical system for its frequency). At that point (in quite a few years), it's quite likely we might get a completly new standard that will be a multiple of 24fps.
Ultimately 300hz is the ideal connection point, because it translates well from 24fps, 25fps and 30fps. Once we can get a good display system that'll conform to that, this will be a good rate to standardize on.
Joined
·
29 Posts
Hi, interesting thread.
re: desirability of 24p standard
24fps is unwatchable (flicker), which is why cinemas show 48fps – but that causes very noticeable judder due to repeat frame flashing (ie your eye smoothly tracking and anticipating moving objects is distracted by repeated flashing of the frame showing the object still in the same place). This judder is only accentuated by further repeat frame flashing such as 72fps (although other artefacts such as flicker are reduced – which some find an acceptable trade-off). The reason the extremely poor motion resolution of film at 24fps is less apparent than it could be is because cinematographers go out of their way to conceal the problem through artful camera work (panning, limited depth of field to avoid background strobing, etc). As frame-rate standards convertors (like the Snell & Wilcox broadcast standard gear) get better (& cheaper) it would be advantageous to be able to upconvert archive film from 24fps to true 48fps or better.
As Ofer said, there is no particular desirability for the arbitrary 24fps for long term future standards, except to allow compatability with archive material, and easy adaptation to present TV standards (50/60 Hz). The higher the frame/field resolution of the capture device the lower the motion resolution at a given frame rate, so it is highly desirable to have better than 24fps for future HD or better material.
If my understanding is incorrect I’d appreciate a correction!
chris
re: desirability of 24p standard
24fps is unwatchable (flicker), which is why cinemas show 48fps – but that causes very noticeable judder due to repeat frame flashing (ie your eye smoothly tracking and anticipating moving objects is distracted by repeated flashing of the frame showing the object still in the same place). This judder is only accentuated by further repeat frame flashing such as 72fps (although other artefacts such as flicker are reduced – which some find an acceptable trade-off). The reason the extremely poor motion resolution of film at 24fps is less apparent than it could be is because cinematographers go out of their way to conceal the problem through artful camera work (panning, limited depth of field to avoid background strobing, etc). As frame-rate standards convertors (like the Snell & Wilcox broadcast standard gear) get better (& cheaper) it would be advantageous to be able to upconvert archive film from 24fps to true 48fps or better.
As Ofer said, there is no particular desirability for the arbitrary 24fps for long term future standards, except to allow compatability with archive material, and easy adaptation to present TV standards (50/60 Hz). The higher the frame/field resolution of the capture device the lower the motion resolution at a given frame rate, so it is highly desirable to have better than 24fps for future HD or better material.
If my understanding is incorrect I’d appreciate a correction!
chris
1 - 20 of 34 Posts
- Status
Join the discussion
