Isochroma
5th April 2006, 06:27
I've been spending some time researching LCD and plasma displays for my near-future home theatre setup (with PC input feed, of course!).
There are many advantages and disadvantages to both LCD and plasma displays for this application. At this point in time I'm leaning toward LCD for its superior lifetime, higher reliability, and higher resolution.
Despite its black level problems (solved soon by dual-panel sandwitching or other neat new ideas!), the advantages of LCD are many, so I've decided to give it a try first.
Besides black level, one other item stands out as a certain disadvantage of this display technology: motion smear. Due to an LCD's sample-and-hold characteristic, video frames persist until overwritten by the next. By contrast, a CRT's phosphor fades to black between frames, 'cleaning' the retina of image persistence.
Some manufacturers are beginning to address this problem. LG.Philips showed an LCD with 120 Hz. backlight (http://ultimateavmag.com/features/106ces2/index2.html) at CES 2006, for example.
The problem with a strobed backlight is that its oscillation rate is arbitrary and is unlikely to match the video's native framerate - ie. 23.976/24/25 Hz. vs. 60/120 Hz. There is not yet any panel on the market that can detect native framerate from image analysis.
So pondering upon this problem, and also realizing that LCD monitors/TVs will become the majority of the market very soon, I realized that a simple software solution, implemented in ffdshow for example - could be used.
The proposal is to add a frame-doubler to ffdshow. It just duplicates frames by an integral number, and inserts black frames in the user-specified number of extra frames. Say I had a 24-fps file and played it on a 72 Hz. LCD. In this case, the framerate would be increased 3x to 72 FPS by duplication, except that the last frame of every 3 would be black. The LCD would strobe at 24 Hz., but since only 1/3 of the frames would be black it wouldn't be too visible (hopefully). If flicker was too apparent, a luma-reduced frame could be used instead of black.
Such a filter shouldn't use much CPU, and since ffdshow can accept raw formats it can be used as a postprocessor for any other decoder.
This would allow LCD owners to completely eliminate motion smear without having to wait for manufacturers to slowly implement their inflexible and cumbersome 'solutions'.
Perhaps even a simple avisynth script plugged into ffdshow could do the trick? I'll think about it and add some more tomorrow...
There are many advantages and disadvantages to both LCD and plasma displays for this application. At this point in time I'm leaning toward LCD for its superior lifetime, higher reliability, and higher resolution.
Despite its black level problems (solved soon by dual-panel sandwitching or other neat new ideas!), the advantages of LCD are many, so I've decided to give it a try first.
Besides black level, one other item stands out as a certain disadvantage of this display technology: motion smear. Due to an LCD's sample-and-hold characteristic, video frames persist until overwritten by the next. By contrast, a CRT's phosphor fades to black between frames, 'cleaning' the retina of image persistence.
Some manufacturers are beginning to address this problem. LG.Philips showed an LCD with 120 Hz. backlight (http://ultimateavmag.com/features/106ces2/index2.html) at CES 2006, for example.
The problem with a strobed backlight is that its oscillation rate is arbitrary and is unlikely to match the video's native framerate - ie. 23.976/24/25 Hz. vs. 60/120 Hz. There is not yet any panel on the market that can detect native framerate from image analysis.
So pondering upon this problem, and also realizing that LCD monitors/TVs will become the majority of the market very soon, I realized that a simple software solution, implemented in ffdshow for example - could be used.
The proposal is to add a frame-doubler to ffdshow. It just duplicates frames by an integral number, and inserts black frames in the user-specified number of extra frames. Say I had a 24-fps file and played it on a 72 Hz. LCD. In this case, the framerate would be increased 3x to 72 FPS by duplication, except that the last frame of every 3 would be black. The LCD would strobe at 24 Hz., but since only 1/3 of the frames would be black it wouldn't be too visible (hopefully). If flicker was too apparent, a luma-reduced frame could be used instead of black.
Such a filter shouldn't use much CPU, and since ffdshow can accept raw formats it can be used as a postprocessor for any other decoder.
This would allow LCD owners to completely eliminate motion smear without having to wait for manufacturers to slowly implement their inflexible and cumbersome 'solutions'.
Perhaps even a simple avisynth script plugged into ffdshow could do the trick? I'll think about it and add some more tomorrow...