What bit rate? at DVinfo.net

Ben Longden · June 1st, 2008, 07:53 AM

A major news client of mine has specified to send news vision encoded to the H.264 (AVC or MPEG4 part 10) format for SD 16:9 material.

Comparing vision quality to bit rates, a file encoded AVC at 4Mbps is equivalent roughly to the same quality as MPEG2 encoded at 8Mbps. The file sizes are different, the MP4 being 15 Mb while the MP2 is 60Mb.

While I can easily deal with this, my question is, at what bit rate is considered an acceptable compromise between transmission time (ie; looming deadline) and vision quality?

Anyone else here string for news?

Cheers,
Ben

Lorenzo Asso · June 3rd, 2008, 03:06 AM

if other conditions are equals, h264/AVC hits the same mpeg2 quality with, at least, 30% of lower bitrate.
This is true, if we are speaking about "very high" bitrate level.

In lower bitrate conditions, mpeg2, generally, loses efficiency...since to arrive in streaming condition where AVC can do "miracles" and where mpeg2 is completely unuseful (for example at 250-300kbps).

4mbit can be good or not. It depends.
Compressibility test help us to understand the video "nature" and to catch, by a very good approximation, the bitrate value you need.
Let's do some examples. Fixed a resolution, a video clip very static and dark could need 4mbit at 1920x1080 to get a "good quality".
But, at the same resolution, for a clip of the same lenght but with very complex motion, could not be enough getting 10mbit to hit a good video quality (motion artifacts free).

Compressibility test are based on the hypothesis that the best possible quality in a video content is influenced by quantizer level (quantizers are the level of compression that are assigned to each compressed frame).
The lower the quantizer is, the higher quality you can hit.

In mpeg4 ASP (xvid, divx...) the lower possible quantizer is Q=2 (saturation level). In AVC the equivalent of Q=2 is Q=18.
Each quantizer level, in each different video, will require a certain level of bitrate.

Empirically is possible to verify that encoding between 70%-80% of the saturation level, will give you the best settlement between quality and size.

So, finding the saturation level of certain video, if the resolution and other parameters are fixed, you can decide what bitrate is the "best choice" for that video.

To find resolution level off course you need to to an encode at Q=2 or Q=18 (for AVC) to find the best bitrate....But, off course, it would be a too slow way.
Here it is the compression test help us.
Compression test (possibile using avisynth or some automatic tool) permits you to encode, for example, 1 frame each 14, or something similar, doing a 3-5% encoding of the total lenght...In this way you can do a compression test in a few minutes and you can find the saturation level with a very very good approximation. (usually 2-3% error).

Practical consequences of the compressibility test are multiple...the most important is that, knowing saturation level, you will give always the "good" bitrate level to your video. Not too much and not to low...

ciao!!

---
N.B. Quantizers in mpeg4 ASP and AVC can be under that value (respectively under 2 and under 18) but in 99% of cases is unuseful and not efficient to go under that value...

June 1st, 2008, 07:53 AM	#1
Ben Longden Regular Crew Join Date: Apr 2008 Location: Echuca, Victoria, Australiamate Posts: 179	What bit rate? A major news client of mine has specified to send news vision encoded to the H.264 (AVC or MPEG4 part 10) format for SD 16:9 material. Comparing vision quality to bit rates, a file encoded AVC at 4Mbps is equivalent roughly to the same quality as MPEG2 encoded at 8Mbps. The file sizes are different, the MP4 being 15 Mb while the MP2 is 60Mb. While I can easily deal with this, my question is, at what bit rate is considered an acceptable compromise between transmission time (ie; looming deadline) and vision quality? Anyone else here string for news? Cheers, Ben Last edited by Ben Longden; June 1st, 2008 at 07:54 AM. Reason: Shpellink

June 3rd, 2008, 03:06 AM	#2
Lorenzo Asso Regular Crew Join Date: Apr 2008 Location: Italy Posts: 47	if other conditions are equals, h264/AVC hits the same mpeg2 quality with, at least, 30% of lower bitrate. This is true, if we are speaking about "very high" bitrate level. In lower bitrate conditions, mpeg2, generally, loses efficiency...since to arrive in streaming condition where AVC can do "miracles" and where mpeg2 is completely unuseful (for example at 250-300kbps). 4mbit can be good or not. It depends. Compressibility test help us to understand the video "nature" and to catch, by a very good approximation, the bitrate value you need. Let's do some examples. Fixed a resolution, a video clip very static and dark could need 4mbit at 1920x1080 to get a "good quality". But, at the same resolution, for a clip of the same lenght but with very complex motion, could not be enough getting 10mbit to hit a good video quality (motion artifacts free). Compressibility test are based on the hypothesis that the best possible quality in a video content is influenced by quantizer level (quantizers are the level of compression that are assigned to each compressed frame). The lower the quantizer is, the higher quality you can hit. In mpeg4 ASP (xvid, divx...) the lower possible quantizer is Q=2 (saturation level). In AVC the equivalent of Q=2 is Q=18. Each quantizer level, in each different video, will require a certain level of bitrate. Empirically is possible to verify that encoding between 70%-80% of the saturation level, will give you the best settlement between quality and size. So, finding the saturation level of certain video, if the resolution and other parameters are fixed, you can decide what bitrate is the "best choice" for that video. To find resolution level off course you need to to an encode at Q=2 or Q=18 (for AVC) to find the best bitrate....But, off course, it would be a too slow way. Here it is the compression test help us. Compression test (possibile using avisynth or some automatic tool) permits you to encode, for example, 1 frame each 14, or something similar, doing a 3-5% encoding of the total lenght...In this way you can do a compression test in a few minutes and you can find the saturation level with a very very good approximation. (usually 2-3% error). Practical consequences of the compressibility test are multiple...the most important is that, knowing saturation level, you will give always the "good" bitrate level to your video. Not too much and not to low... ciao!! --- N.B. Quantizers in mpeg4 ASP and AVC can be under that value (respectively under 2 and under 18) but in 99% of cases is unuseful and not efficient to go under that value...