[opendtv] Re: Microsoft Exec: 1080p HDTV Is Meaningless

At 2:38 PM +0200 8/18/06, Hoffmann, Hans wrote:
......so what do you think: how much bit-rate would be required for a
1080p/60 emission format using e.g. MPEG-4 AVC?

1: equal to 1080i/30 (average critical material)
2: less than 1080i/30 (how much with average critical materail - percentage)
3: more than 1080i/30 (""""""""""""""""""""""""""""""""""""""" - percentage)


Regards, Hans


Good question Hans.

The answer is a bit tricky, as it is dependent on the actual information content of the images that are encoded, and their relative quality in terms of entropy (i.e. the accuracy of the samples).

Keep in mind that the average frequencies of most of what we and a camera see, are lower than the high frequency details that may exist in any imagery. This is the major reason that TV has worked so well for decades, and the reason why cinematographers go out of their way to prevent the acquisition of high levels of detail (e.g through the use of filters and depth of field to eliminate detail outside of the critical focus plane).

As we add resolution to any image two things typically happen:

1. The accuracy of the samples that represent lower frequency information improve - this helps entropy encoders.

2. We acquire more detail - some of this information will be very accurate, and some will not. This stresses an entropy encoder in two ways: you need more bits to represent the extra information and you need more bits to deal with the entropy (distorted information).

So the answer is that you will need more bits in complex scenes, while in average scenes you might actually need the same or even less bits. It will all depend on the quality of the images produce by the camera. I would also add that the higher frame rate, and elimination of interlace will help with the motion compensated prediction routines in the entropy coder, which can produce major gains in coding efficiency. But this ALSO requires good sample accuracy, which is more difficult to achieve as the frame rates increase.

I would note that most recording systems for 1080 line HD cameras today, downsample to 1440 or 1280 samples per line prior to compression. The main reason for this is the high levels of entropy in the full 1920 x 1080 raster. Since these cameras do not oversample, they cannot do a good job of capturing information at the highest frequencies possible for the 1920 x 1080 raster. Typically there is a fairly high level of noise at the highest frequencies, due to the sensitivity limitations of the sensors (i.e. the number of photons hitting each sensor site).

Large CMOS sensors (higher sample counts and larger physical size) may help with the sensitivity issue, and the resulting images can be resampled to help eliminate sampling errors. But we are still not there with this technology.

I hope this helps.

Regards
Craig



----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: