AVCHD video at higher bitrates

[Arkady Bolotin]

AVCHD (Advanced Video Coding High Definition) utilizes MPEG-4 AVC/H.264 (AVC) video coding for video compression. The maximum allowed video bitrate for AVCHD, as we know, is 24 Mbit/s (at the level 5.1).

So, the question is if AVCHD-based camera recorded footage at maximum-allowed bitrate, can we go beyond this limit? In other words, can the increase of video bitrate in postproduction produce new information (i.e. better quality) in AVCHD video?

Theoretically speaking, the answer to this question should be possibly yes, or might be.

Given such features of AVC as multi-picture-inter-picture prediction, lossless macro block coding and loss resilience, the improved quantization design, it is possible that decompressing AVCHD video and then recompressing it again but at the higher bitrate can produce higher quality.

Without starting technical dispute here, I just want to show some results I got. Please observe the following two fragments: the first fragment has been cut from the original footage recorded at approx. 21 MBps, and the second one was from the postproduction movie rendered at approx. 80 MBps.

As you can see, the central element of the ornament is more evident in the high bitrate fragment.

[Robin Davies-Rollinson]

I don't see how you put detail in that wasn't there in the first place...

[Robert Young]

I agree- you can't produce information that was not there to begin with.
However, decompressing your original footage and recoding into higher bitrate, higher level (10bit, 12bit, 4:2:2) digital intermediate codec for editing gives you more redundancy and "digital headroom" for applying extensive effects, color correction/grading, etc. and can result in a better quality image in the final delivery format. Otherwise, I doubt that there is any serious advantage.

[Arkady Bolotin]

Robin:

The (additional) information does not appear out of nothing; rather it was unseen in the first place and revealed in the process of rendering at higher bit rate.

Let me take an example: imagine that we have number 25.63452985032(7)… If the brain power of a PC processor is limited to certain extent then instead of that number we read just 26. However, if the processor would be enforced somehow, we would get more accurate approximation of the original number, say, 25.635.

Something analogical goes with AVCHD video. It is severely compressed so to be played back it has to be decompressed first. Where this decompressing is taking place depends on the video bitrate. It could be decompressed completely by the processor on-the-fly i.e. during replay. Or it could be partially decompressed by rendering process at a higher bitrate and then playing back with lesser burden on the processor (i.e. more accurate = with smaller number of artifacts).

Robert:

I believe I already answered to your first argument.

I agree that besides of what I just said, decompressing could also increase redundancy and doing so facilitate the application of extensive video editing.

Regarding your doubt in serious advantage of all this, let me respectfully refuse to respond. The reason is plain: I do not like vague or fuzzy words or concepts behind such words. Please clarify what you mean by serious advantage, and then I will eagerly respond. To me it’s simple: if I see what I couldn’t see before it’s advantage.

[John Abbey]

I have been using a firmware hacked panny GH1 and now get somewhere in the 40mbps range, and to my eye it looks fantastic.

[Robert Young]

Originally Posted by John Abbey

I have been using a firmware hacked panny GH1 and now get somewhere in the 40mbps range, and to my eye it looks fantastic.

Are you referring to the recording datarate??

[Robert Young]

Originally Posted by Arkady Bolotin

Regarding your doubt in serious advantage of all this, let me respectfully refuse to respond. The reason is plain: I do not like vague or fuzzy words or concepts behind such words. Please clarify what you mean by serious advantage, and then I will eagerly respond. To me it’s simple: if I see what I couldn’t see before it’s advantage.

All theory aside, the way I usually compare various workflow ideas is to carry them out to the final delivery format to see if there is any difference, and if so, what really looks best. At the end of the day it is the image quality of the final delivery product that really matters.
So you can take the original AVCHD clip, a higher bitrate version, and a DI version, like Cineform, render them all out to (for example) BluRay, put it on a BD & watch on a high quality large screen HDTV. If you can see a difference, you've got your answer. If you can't see the difference, then it doesn't make any difference.

[Bruce Dempsey]

Hey Arkady I've just received a Blackmagic Intensity Shuttle to capture from the HDMI port on my AVCHD based camera (cx550)
Should be fantastic looking video of my figureskating subjects. The bitrate of uncompressed video is the astronomically high figure of 155.52 Mbps (standard SI-units) but I plan on capturing as mjpeg so as to realistically store the footage long enough to trim and encode to BluRay disc and the figureskating routines usually do not exceed 4 minutes in length so it's manageable I hope
I must say I enjoy reading your comments

[Arkady Bolotin]

It’s four o’clock in the morning here, so I’ll be brief.

Robert, it is simple indeed: if I could not see any difference, there would be no post and I would not have the pleasure of this talk with you.

Bruce, thank you! I think you understand what I’m trying to do… And my congratulations on your new Blackmagic capture board!

[Dave Blackhurst]

If I should be so bold, I think Arkady may be on to something... if we were talking a single "still" frame, there are ways to enhance the amount of "data" through interpolation (there was an interesting thread at one time here on DVi about an algorithm that could extract virtually infinite detail from a small digital sampling)

With video, if my understanding is correct, the "goal" of compression is to keep as much data as possible, using methods of retaining the pixels that remain the same as much as possible within a frame, and only generating the "different" information. for the sake of argument, I'd think it's safe to postulate that within all teh "same" information is a certain level of "change" which is in fact recorded in the intermediate frames, and if one were to, after decompressing the data stream, utilize the additional data across frames to enhance the ramaining frames, it in theory should be quite possible to enhance the overall individual frames to a substantial extent.

I think what is being proposed is that "post decompression" it is possible to recover levels of information which upon "recompression" at a higher bitrate, no longer have to be discarded, thus resulting in a "better" picture.

I've always been fascinated with compression algorithms, and how one can basically toss huge portions of the data, yet still manage to re-create the "original" from what would logically seem to be a flawed (due to "loss of information") digital file. The math goes way over my head, but the concept is still fascinating!

[Robin Davies-Rollinson]

Well, never let it be said that I don't have an open mind.
I'm going to look into this today, but I want to try it with a test card to see if there will indeed better linear resolution if I render out at, say, 80Mbs.
Well done Arkady for giving us something to think about!
One thing Arkady, how did you export the stills?

[Robert Young]

Here's the part I have difficulty with- Arkady is decompressing his raw 24mbs AVCHD and transcoding/recompressing to a higher bit rate format, then comparing the transcode to the original footage.
Is this not the same as transcoding 24mbs AVCHD to a DI such as Cineform 100mbs .avi? The transcoded footage looks great, it's 10 bit or even 12 bit, 4:2:2, and so forth, and will be more "lossless" downstream.
But even the companies who develop and sell DI software do not claim that their product will produce an image that is "better" than the original footage. Their only claim is to preserve the original image quality throughout the editing process all the way to the final delivery format.
If Arkady's assertion is correct, then it has somehow been completely missed by all of the engineers, software developers, & video professionals who work with these issues on a daily basis, all with the goal of trying to squeeze every fractional improvement possible out of the image data.

[Bruce Dempsey]

Sony's PMB has yielded the best looking stills from a m2ts file for me extracting action shots from moving subjects (I do not know what Arkady did for his stills)

Something similar occurs in the HDV world. How is it that when a scene is recorded as hdv and downconverted to sd then played back on an upconverting dvd player it looks better than a straight sd recording and play back on the same upconverting dvd player?

[Arkady Bolotin]

Hi guys, thank you everyone who participated in this discussion so far.

Regarding all your queries, allow me to respond to them not individually but principally.

1. How I did this.

For postproduction I used Sony Vegas Pro 9.0e choosing for the rendering template the settings of either Windows Media Video V11 or MainConcept MPEG-2 at the variable bit rate (actually I varied only the peak bitrate). I tried different rates including 25, 40, and 80 Mbit/s.

To take snapshots I used VLC Media player (ver.1.50), and the frame fragments presented in this post were made partially in Windows 7 Paint and partially in Adobe Photoshop.

2. Why it is possible to uncover (recover) additional details by rendering AVCHD footage at higher bitrates.

I think a few factors – not one – might play the role here. One of them, as I said before, is a reduced processor load during the replay of a less-compressed movie.

Another one, as Dave pointed out, might be the utilization of the additional data across frames resulting in some enhancement of the remaining frames to a substantial extent.

Also, it could be that the inter-frame compression algorithm takes its part. In the course of rendering, this algorithm (together with the block matching algorithm) can find a matching block with little prediction error so that, once rendered, the overall size of motion vector plus prediction error is lower than the size of a raw encoding.

3. Why it has been completely missed by all of the engineers, software developers, & video professionals who work with these issues on a daily basis, all with the goal of trying to squeeze every fractional improvement possible out of the image data.

No, it hasn’t. Improving video quality by means of increasing bitrates is well-known and well-studied theoretical approach.

The problem here is practicality. The maximum bitrate of each video standard (24 MBps for AVCHD, 25 MBps for HDV, 40 MBps for Blu-ray Disc) is not determined pure theoretically but in accordance with the corresponding media. Thus, 24 MBps bitrate is the upper limit for flash memory media, 25 Mbit/s is the maximum bitrate for magnetic tape, and 40 Mbit/s is the maximum for optic disk media.

This means that even if we manage to squeeze some additional detail by the increase of bitrate, the resultant video could be played back only on above-the-mainstream computers powered with 3.0 GHz or better processors and equipped with RAID-type hard drives.

[Guy McLoughlin]

Originally Posted by Arkady Bolotin

The problem here is practicality. The maximum bitrate of each video standard (24 MBps for AVCHD, 25 MBps for HDV, 40 MBps for Blu-ray Disc) is not determined pure theoretically but in accordance with the corresponding media. Thus, 24 MBps bitrate is the upper limit for flash memory media, 25 Mbit/s is the maximum bitrate for magnetic tape, and 40 Mbit/s is the maximum for optic disk media.

This means that even if we manage to squeeze some additional detail by the increase of bitrate, the resultant video could be played back only on above-the-mainstream computers powered with 3.0 GHz or better processors and equipped with RAID-type hard drives.

OK, you've lost me here...

40 Mbps is only 40 mega-bits-per-second which translates to 5 MB per second of data. Even my crappiest USB key can handle a 7 MB per second data transfer rate.

Modern SATA hard-drives that are in most computers built in the past 2 years can handle a data transfer rate of 70 - 110 MB per second. ( 560 - 880 mega-bits per second )

The problem with higher AVCHD bit rates is that they are not part of the standard set by Panasonic and Sony, and thus the encoded video files may see very little benefit from this additional data to work with.

I expect the AVCHD standard to evolve, and hope that 4:2:2 color is implemented at some point soon, as I think 4:2:2 color will have a bigger impact on the finished results than just using a higher data rate.