Deinterlacing 1080i HDTV for 1080p displays

Started by jjkilleen Oct 31, 2007 13 posts
Read-only archive
#1
Is there a standard way TVs display interlaced DTV? Do they collect the entire frame before refreshing or refresh half a frame at a time?
#2
There is no standard unfortunately... Ultimately like all things consumer it is function of how many complain and that will be very few since what ever method is used it works good enough for the masses.
#3
Depends on the display technology. A 1080i CRT [is already interlaced by nature - that is how it displays an image and no conversion or scaling is required]. [All other displays are] inherently progressive and must collect and display the entire frame at once [so some form of processing has to take place].

Why do you want to know?

Edited by Richard 10/31/07
#4
Thanks for the answers. I'm an engineer, so I'mcursed with an endless need to understand how things work. This was an aspect of TV processing about which I had seen conflicting statements, and I suspected the reason was, as you've said, there's no standard way of doing it.
#5
The improper way is to filter out the vertical resolution and smash the two fields together. Even consumer CRT RP does this filtering to prevent aliasing artifacts from appearing. Performance folks expected more from 1080p products but nearly all the first and many of the second gen displays still do it. Appears external scaling has found yet another purpose to exist along with anamorphic CinemaScope.

How this appears is if you feed a display a a grid pattern that is one pixel for the white lines you will note the vertical lines are correct but the horizontal lines, vertical display response, is smeared and when you go up to the screen you can see three active pixels for these lines instead of the single pixel it should be - that is vertical filtering in play for 1080p displays.

It doesn't have to be that way, we have the technology via adaptive motion or 2/3 pull down, but many chose the cheap and dirty route and for good reason; most viewers are at 5 screen heights or more and with a 1080p pixel matrix this artifact is impossible to see. What it does create is a softened detail response and at 5 or more screen heights you are yet again in the realm of difficult to perceive. In a mas consumer market that is racing towards the bottom for price, price, price this is to be expected.
#6
Engineering is all about tradeoffs! Then I assume the same methodology is used in processing 1080i for a set with 768 vertical resolution?
#7
768 is not native to anything HD. 720 is and not only is filtering taking place but some of the response has to be thrown out depending on the image - you lost the other 360 pixels...
#8
Richard-the reason I said 768, is that most 720P sets I've seen have 768 vertical resolution. Won't the 1080i be rescaled to fill the 768? I thought that both 720P and 1080i would be rescaled to match the sets 768 resolution.
#9
Richard-the reason I said 768, is that most 720P sets I've seen have 768 vertical resolution. Won't the 1080i be rescaled to fill the 768? I thought that both 720P and 1080i would be rescaled to match the sets 768 resolution.

Yes. I think Richard's point was that nothing is native 768 so everything has to be scaled whereas a 720p set would only have to scale 1080i/p and not 720p.
#10
Ya, another CEA issue of definitions that skips the performance attribute of 1:1 pixel mapping for the sharpest presentation. You only get that by feeding 720p to a 1280X720 display or feeding 1080p to a 1920X1080 display. I left 1080i out due to the topic of this thread - takes correct scaling/conversion to get the most out of a 1080i source. HD disc is native 1080p24 for the film content.

You are correct - difficult if not impossible to find a 1280X720 pixel matrix in a flat panel.
#11 (edited Nov 3, 2007)
Jjkilleen,

I believe your original question was not responded as fully as you illustrated on your statement.

1080i has two fields of 540 lines each appearing every 60th of second to make one frame. In the progressive display world we live today each 1080 frame is displayed at once (not per field as CRT).

If the set is a 720p, or 768 panel, etc then there is scaling that take place in addition to the processing statements I include below, so to keep it simple, in a world of more 1080p sets that we imagined, I will cover 1080p displays upconverting from 1080i.

There are (MANY) TVs that take each incoming field of 540 lines and complete the remaining 540 lines of the 1080 frame to be shown with video processing, and shoot the 1080 lines-frame in 1/60 (the speed of the field).

In this case half of the pixels of the frame are calculated and interpolated within the image, there were not present in the original image. This type of interpolation is an art, there are good painters, but there are also many bad painters around.

Imagine using Photoshop to blow up a photo, finding that it is now too grainy because is missing pixels, and you go with the touch tool and start inventing pixels, half of the whole picture. Even if that picture frame could have been done to perfection, imagine now another 59 pictures right behind that one, all within one second.

Perfection has a limit, and also is rare, and most manufacturers are not looking for perfection, they like revenue, the sooner the better, so another model can be sold again 3 months from now, even to the same person.

There are other TVs that actually interleave both 540 fields into the same 1080 frame, then shoot that whole frame at 1/60, and display again the exact same (repeated) frame in the next 1/60. The following 1/60 will contain a frame made with the next two new incoming fields, and repeat that one again at the next 1/60, and so on.

In that case, one has to remember that the 540 lines of each field record an instant in time of pixels that come from a different position than the next field. Objects in the image in front the camera are recorded in halves, and they could have moved during the short interval of each field recording, putting those two together without any adaptive video processing could be perceived as lower resolution and blurriness.

There are TVs that combine both techniques depending on the area of the image, if the image is not moving much, like a fixed picture of the blue sky or the grass of a golf course with a camera shot that did not move, then such area could afford calculated pixels. The area of the image recording the golfer
#12
Thanks Rodolfo for your time and effort! :D
#13
Thank you all for your most informative posts. Clearly, this is a very complex and interesting subject.