RoadBikeReview.com's Forum Archives - General


Archive Home >> General(1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 )


Tech tip: OLN pixelation explained...(10 posts)

Tech tip: OLN pixelation explained...mr_spin
Jul 11, 2002 7:41 AM
Anyone watching OLN off a satellite (and maybe cable) has no doubt seen pixelation of the picture, which is a checkerboard effect appearing in parts of the picture. Anytime they show a spinning wheel it is really obvious.

I asked my brother, who works for a major network's sports division and whose work you have all seen (e.g., the Superbowl), to explain this effect. For anyone who wonders, there is his response.

By the way, God bless OLN!




This is a result of MPEG-2 compression, which works in blocks of 8 X 8 pixels... for a total of 720 X 480 pixels flying by at 29.97 frames per second. The video is compressed somewhere from 8:1 to 30:1, a ratio chosen by the DSS people based on how they feel that day. Actually, it's most likely based in the political importance of the TV channel. HBO's premiere of Madonna's concert gets more bits than a bicycle race in a foreign country that most Americans would have difficulty finding on a map. Here's how it works:

The feed is picked up by the DSS provider and compressed into a paltry stream of lossy images using a two-dimensional discrete cosine transform (DCT) on 8 X 8 blocks of pixels, followed by a quantization scheme on the DCT amplitudes, followed by a coding scheme on the quantized coefficients to create a stream of data at approximately 1 to 3 Mbits/sec. 1.5 Mbits/sec. is said to be VHS quality.

In other words... the picture is broken down into blocks... each block assigned a number of bits, say 50 to 200 per block, based on their perceptual complexity. A complete frame is called an I Frame and is similar to a single frame that has been JPEG compressed.

Now lets add some temporal compression! If the previous image is pretty much the same as the current image, why recompress it all? Using a prediction algorithm you can create difference frames that exploit common characteristics of moving video... like spatial redundancy (the sky is all blue), temporal redundancy (the sky hasn't changed since the last frame), and motion (the camera is panning to the right).

These temporal frames are called P Frames (Predictive) and B Frames (Difference). B frames are like temporal spackle. The sequence typically used is IBBPBBPBBPBBPBB. This can be changed depending on the image complexity. You can even do IIIIII if you want... but that defeats the purpose of compression.

A DVD will typically use a data rate above 6 Mbits/sec. and is encoded by a lengthy supervised process that involved multiple passes and fine tuning by a highly paid operator... not to mention they only have 24 frames per second of noiseless film quality images to deal with.

DSS is noisy interlaced video encoded in real time to a constant bit rate decided by underpaid morons and squeezed into a data stream that is shared with hundreds of other channels. It doesn't help that the video traveled through hundreds of electronic components and thousands of miles before it got there.

Does that answer your question? The complexity of all the image movement involved in a bike race is overwhelming the MPEG-2 encoder at the given bit rate. If they could just all stop and stand still the image would look much better.
re: Tech tip: OLN pixelation explained...novagator
Jul 11, 2002 7:48 AM
just as I thought!! :) Still doesn't explain why, out of all the channels that I have, my cable company had a problem last night with only one - OLN!!!! So, I missed the TTT replay :(
My brain hurts now. thanks for the info NMPhatMatt
Jul 11, 2002 8:02 AM
re: Tech tip: OLN pixelation explained...wolfereeno
Jul 11, 2002 8:47 AM
Becuase somewhere along the line, someone decided that OLN doesn't merit as much bandwidth as home shopping club. On a channel by channel basis, someone sets priorities which boil down to how much the signals are compressed.

Guess we know where we stand.
funnyDougSloan
Jul 11, 2002 8:52 AM
That could be absolute bullsh|t and we'd never know the difference. :-)

Doug
mmm...donuts! (nm)JS Haiku Shop
Jul 11, 2002 9:21 AM
I don't think sofiltersweep
Jul 11, 2002 1:59 PM
I'm not using DSS- but rather straight digital cable- and I won't go into the details of how that should give me a BETTER picture... but it is STILL lossy. It is lossy at the source.

The lossy picture is likely because it is taken from a TON of wireless camera feeds (they don't string cords off those motorcycles). It is relatively low-res and low budget. Yes- there is a ton of compression, but it is at the source (the race) not at the network. OLN gets just as much cable bandwidth as NBS or the shopping network.

It isn't nearly as lossy when they are tossing around the jerseys at the end. It does, however, retain that "local news footage" ambience.
correct description of source of problemJohnG
Jul 11, 2002 3:25 PM
Yup, the VAST majority of the pixelization is due to over-compression by the DSS providers. There may also be some losses introduced at the source but they are probably way down compared to what the A-holes at DTV introduce.

I can hardly wait to see how crappy the sat signals get once the dipshits at Dish takeover. :( Here's another case of the FCC and the SEC not doing their job.

JohnG
then explain this:filtersweep
Jul 11, 2002 4:38 PM
and again, I don't use DSS- so this is the true test:

WHY is that Lincoln Navigator commercial always CLEAR AS A BELL ?
two reasonsmr_spin
Jul 12, 2002 7:13 AM
1. It's not live.

2. It's not processed multiple times and bounced off satellites from Europe.