AVCHD video at higher bitrates

[Arkady Bolotin]

Oy vey, David, give me a break…

You discard my statements as baseless and instead offer yours equally deniable and unsubstantiated…

Think about it. The problem here is not the logic or wordplay, we are arguing about something that can be only proved or disproved by experiments and observations.

Neither you nor I possess any real knowledge about what AVCHD coder Sony or any other manufacturer uses in their cameras, what algorithms they apply and what the profile of H.264 they employ, how lossless or lossy the encoded video could be in various situations.

Anyway, thank you for answering. I am glad you are still here.

[David Heath]

Originally Posted by Arkady Bolotin

You discard my statements as baseless and instead offer yours equally deniable and unsubstantiated…

Think about it. The problem here is not the logic or wordplay, we are arguing about something that can be only proved or disproved by experiments and observations.

OK - it's an observation that a computer with a low power won't playback AVC-HD at all. Increase the processing power, and at a certain threshold it will start to play with stuttering, then with a little more power play cleanly. (I think Dave Blackhurst confirms this?) No matter how much more powerful a processor you then put on the task, it won't make any better job of the decoding quality wise - the coding quality is a function of bitrate for any given coder.

Try that as an experiment for yourself. Play your samples on a range of computers with different powers and see if you notice any quality difference.

That's why I feel my statements are not unsubstantiated, and ARE based on observations. (As well as theory.)

I'll let everybody else make up their own mind.

[Arkady Bolotin]

You said “No matter how much more powerful a processor you then put on the task, it won't make any better job of the decoding quality wise - the coding quality is a function of bitrate for any given coder”.

But this is an exact confirmation of what I did: I made two clips in the different video formats – AVCHD (M2T) and VC-1 (WMV) – at the different bitrates – 24 MBps and 80 MBps respectively – and showed that “the coding quality is a function of bitrate”, i.e. that the quality of the 80 MBps WMV clip is higher.

Look, David, we can do it eternally: your argument – my argument and so on, and so on…

I think we should stop right here. I am sure: everyone has made up his mind long time ago.

[Dave Blackhurst]

There are TWO variables here - as we are speaking of, the horsepower and hardware, AND the encoding bitrate. If something is recorded at a lower bitrate, it'll play back on lower grades of hardware, but it will not look as good as a higher bitrate stream, and the likelyhood of ANY extra detail being available to extract is pretty small.

Based on what I understand of compression, it should, under some circumstances be possible to pull additional detail out of a decompressed stream, and in theory use a higher bitrate to "keep" that detail in a more intact form.

I think the main problem with the theory is that one needs to know how well movement in the frame is dealt with before you could even know whether the endeavor is worth the hassle, my suspicion is that the specifications for the standard are set up in such a way as to be as "gracefull" as possible handling detail and movement with "average" hardware, under "average" shooting conditions, and under those "average" conditions, the possible improvement would be minimal.

[Arkady Bolotin]

To decipher the “mystery” of more-detailed-appearance of the high-bitrate clips rendered from original AVCHD video two words have a very important meaning. Those words are transcoding and interpolation.

Transcoding is the conversion between different encoding formats. Naturally, transcoding between two lossy formats – be it decoding and re-encoding to a different codec or to a different bitrate – causes generation loss (i.e. loss of the video quality).

However, if lossy-compressed video is transcoded to lossless or uncompressed one, then it can be technically qualified as a lossless conversion because no information would be lost.

In practical terms, it means that transcoding video from the AVCHD@24 MBps format to the VC-1@80 MBps format shouldn’t produce (perceptible) video loss. In other words, the original video clip and the transcoded one must have the same video quality.

So, why the VC-1 clips do appear as more detailed? It is so due to the bicubic interpolation utilized in the VC-1 video codec. By constructing new sub-pixels, the bicubic interpolation brings about the appearance of fractionally filled pixels (i.e. smoothness) in the picture. In other words, the bicubic algorithm (which is lacking in the AVCHD format) increases apparent sharpness (i.e. detail level).

The downside of the bicubic interpolation is that it reduces contrast: that is why the rendered clips look less contrast.

Thus, transcoding video from the AVCHD@24 MBps format to the (more advanced) VC-1@80 MBps format is merely a video enchantment. That’s it, nothing more, nothing less.


POST SCRIPTUM:

Robert and David, I think I owe you both a great deal of apologies.

David, I feel very sorry. You were right, I was wrong. Instead of paying attention to your arguments I drowned them deep into empty demagogy.

My sincere apologies go also to everyone who felt mystified by those high-bitrate clips. I was bewildered by them as well. But this is called being human.

[Robert Young]

Arkady
One of the very useful purposes of these fora is to stimulate interesting discussion.
I can't help but notice that your original post has generated 4 pages of discussion and more than 1,000 viewers- a sure sign you hit a sweet spot!
When one notices something interesting, intriguing, or unusual, you can hardly go wrong by presenting it here.
It's not so important to be right or wrong- it's the challenge of examining an interesting topic that makes it fun and thought provoking for everyone :)

[Ron Evans]

Arkady
I echo Robert's post. It has been a very interesting discussion and that is why I read the forum. Keeps my retired brain active !!!

I look forward to your next discussion.

Ron Evans

[Phil Lee]

Hi

I was just reading this with interest.

I think what might be overlooked is the conversion to higher bit-rates seem to also swap to a different codec. What is being observed is probably differences with the types of de-interlacing taking place depending on what decoder is being loaded and used and how much CPU might be left over. Certainly I always find MPEG2 at Blu-Ray bit-rates compared to H264 at the same bit-rate, has the MPEG2 looking more watchable on the computer, even though H264 is more efficient and can do even more with the high bit-rate.

Maybe a better test would be to use progressive video to rule that out, but you also can't rule out a different decoder adding a bit of sharpening or other processing, or the placebo effect, which might also be a big part of this.

The acid test is to export frames with lots of detail from an editor (you can pick the exact same frames then) and flick between them as static images. Using Alt-Tab you can do this very easily on a PC, and I've taken snapshots from the original AVCHD footage, from footage exported as uncompressed video, and from the footage converted to high bitrate MPEG2 and H264, and they have looked exactly the same!

Regards

Phil