More about Video Compression & Video Formats


DXNMedia

Moderator
Staff member
Dec 7, 2006
1,024
4
38
Singapore
www.dxnmedia.net
I find it very annoying when many people, especially even industry professionals, get very misinformed about the video recording technology developed by manufacturers or by any means of technology hype.

I have a buddy who is an experienced video cameraman, and I think he blindly spent close to $6k on a highly compressed video format when he could have spent a bit more to get a camera with lesser compression that will yield better quality in the footage acquisition.

It's been a while since I posted any info-sharing thread, so here goes my contribution.
I'm sure there are many more info on the web, but the gist is this:

What is Video Compression?
It is to save storage space in your video image, much akin to how JPEG file vs. a TGA/RAW photo file.

What does it do?
Video compression discards data (Luminance, Chrominance &/or motion) based on an algorithm that predicts and thinks what the human eye will not see or notice. Different compression technology have it's own algorithm to discard these video data, but compression can be classified generally into two categories. Color Space recording capability, and Interframe Compression vs. Intraframe Compression.


Color Space
Color space determines the amount of color info being recorded in the video. It is expressed in ratios seen like 4:4:4, 4:2:0 vs. 4:2:2. Video is generally split into 3 channels, the Y (Luminance), R-Y & B-Y. In the true 4:4:4 color space recording, every data in a 4-pixel block is recorded pixel for pixel without any interpolation like the other type does.

Further Reading:
http://en.wikipedia.org/wiki/Chroma_subsampling
http://en.wikipedia.org/wiki/YUV
http://www.usa.canon.com/dlc/controller?act=GetArticleAct&articleID=3389


Intraframe vs. Interframe Compression
Intraframe compression is compressing every video frame like what JPEG does to a still photo. Benefits of this type of compression is that it preserves motion data very well, and is very well suited for chroma keying applications. Downside is that the video file sizes are bigger.

Interframe compression take motion into consideration and in a group of frames, the algorithm involves taking a key reference frame (the i-Frame, aka Intra-frame), and it predicts the motion and discards information along the way. Interframe compression can produce very efficient filesizes, but may not be good for certain applications, eg. sports, fast action & chroma-keying. In the not-so-earlier days of computing, it is also more intensive for the CPU to process Interframe based video files for decoding & re-encoding.


Examples of Interframe based compression

MPEG, MPEG2, H.264, MXF as file formats.
MPEG-IMX, BetacamSX, XDCAM, HDV, XDCAM HD, XDCAM 422, XDCAM EX, Canon XF, and AVCHD as camera recording formats.



Examples of Intraframe based compression

Uncompressed 10bit/8bit 422 Quicktime, ProRES422, DVCPro50, DVCProHD as file formats.
miniDV, DVCAM, DVCPro25, DVCPro50, DVCproHD, AVC-Intra50, and AVC-Intra100 as camera formats.

The holy-grail of high quality video is still Digital Betacam for Standard Definition, and HDCAM for High Definition at 1440x1080 or even HDCAM SR for Full HD at 1920x1080.
These camera formats can be considered as a reference for almost no compression at 4:2:2 color space, although in technical specification, there are still some compression involved. (DCT compression)

High end formats like RED ONE, Arri Alexa, etc.. are in the league of raw 4:4:4 color space and higher than 2K recording capabilities.



So in summary for the layman....
The smaller your file size, the more compression you have in your video.
The more info you discard, the lousier your video quality is going to be in terms of color, image sharpness, motion artifacts & compression noise / image blockiness.

Here's the files size comparison:
- 60mins of HDCAM uncompressed 10bit 4:2:2 1080i25 video file is taking up almost 500GB of space
- 60mins of ProRES HQ 1080i25 video eats up 100GB
- 60mins of DVCProHD / AVC-Intra100 1080i25 video gobbles up 60GB
- 60mins of Canon XF 50mbps 1080i25 video uses up about 30GB of space
- 60mins of XDCAM HD 35mbps 1080i25 video takes up about 20GB of space
- 60mins of HDV 1080i25 video takes about 13GB
- 60mins of AVCHD 1080i25 video takes about 5GB

Here's a rough price comparison of pro video cameras:
- HDCAM cameras above $80k range
- ProRES recorders about $7k range (no cameras other than Arri Alexa)
- DVCProHD/AVC-Intra cameras $8k & above
- XF cameras $10k and above
- XDCAM cameras $7k & above
- HDV cameras $6k range
- AVCHD cameras $4k range



Conclusion

Like MP3 vs. uncompressed CD audio, JPEG vs. RAW files, some people are just happy to have smaller & manageable files at a compromise over quality.

Compression is not a bad thing, and as compression technology gets more efficient, we're able to attain the best of both worlds in our video recording. There's always an application for every technology and as people who loves dabbling with video as a hobby or as a pro, we'll need to understand the limitations and advantages of our tools & equipment.

Don't be hyped into buying some video equipment just because it promises larger sensors, and bigger lenses because ultimately the limitation and bottleneck is still the recording format.

I believe MPEG2 & AVCHD have been the recent popular video acquisition format mainly because broadcasters have adopted native MPEG2 & H264 files for tapeless & server-based transmission. It is no doubt that these formats are good delivery format, but not necessarily a good acquisition format.

:thumbsup:


2 cents worth of info that I hope is useful for CS videographers.
Feel free to add more info to this thread pertaining to Video compression and type of video formats.
________
ROMULUS ENGINE
 

Last edited:
excellent post, thanks for sharin
________
Cx-5
 

Last edited:
bro..thank you for sharing, i read and am trying to understand and have to read few time to understand better....have to learn to remember all these format things...iam from the VHS ERA and then to DV format..using 3ccd cameras..mainly canon camcorders xm2...which were priced afforadle 6-7 yrs ago.....now i want to upgrade tand there the confusion comes in with all sorts of cameras and recording format....

many ogf us are in the mid-range group, doing weddings and output to dvd....and now or later all of us have to move into TAPELESS era...

bro if u know or have some forums or training..pls do let me know..i am at email rajsingh90@hotmail.com or phone 012-5887719 ipoh.

thanks and i know GOD will bless you for lifting up others in these thecnology....

rajsingh:)
________
Honda cbr900rr specifications
 

Last edited:
Compiled from other threads:

What kind of Video format is Canon 7D produces?

Hmm..dun know how to explain..it is for sure not HDV right? The reason I asked is because I am looking at getting a Final Cut Express HD and not sure if can be used to edit the video taken from a 7D?

Anyone can advice?




Canon 7D uses a H.264 codec wrapped in a Quicktime container, thus giving you the .mov filename.

H.264 is a MPEG4 based compression, essentially a compression & decompression (CODEC) algorithm that is what AVCHD compliant camera uses too. It is no doubt one of the most efficient video compression in today's video technology, yielding the smallest file size while maintaining a very decent picture quality.

http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC

The coding & decoding of the video file is pretty intensive on the CPU, thus it's becomes a very unfriendly format to work with in post production.
Best workaround is to transcode the files to a format that is friendly & native to the editing system.

Although AVCHD, MPEG4 & H.264 are formats widely accepted for web delivery & by some broadcast TV stations, these however are only meant to be a delivery format or as a 'consumer' format. Although manufacturers have been pushing the AVCHD format to be part of the professional product line, I seriously think that is it not intended to be an acquisition (recording) format for critical work.

Even my professional camera have this disclaimer:
"This Product is licensed under the AVC Patent Portfolio License for Personal & non-commercial use of a consumer....no license is granted for any use other than the personal uses detailed below:
- To encode video in compliance to AVC
- To decode AVC video encoded by consumer engaged in personal & non commercial activity
- To decode AVC video that was obtained from a video provider licensed to provide AVC Video."

To me, this simply means that even the manufacturers refuse to take responsibility for professional use of the AVC codec mainly because recording is too compressed that it may distort the accuracy of image acquisition, and it may add unnecessary compressional noise & artifacts into the video image.
And when this happens to critical imaging applications like medical video documentation where every pixel data is important for post surgery analysis, etc... this format is a sure no-go for such use.


Similar analogy can be pulled from why still photographer like to shoot RAW compared to JPEG. Because when you shoot RAW, you'll have more pixel info to manipulate...when you shoot in JPEG, you lose a lot of critical data.....same goes for AVCHD & H.264 video recordings.

But having said all these, most of our applications are for web delivery, basic television work & short film productions for entertainment purposes, I guess many people won't mind the downside of using such a heavily compressed format for video recording purposes. :thumbsup:
________
DODGE 400
 

Last edited:
Compiled from other threads:

To me, this simply means that even the manufacturers refuse to take responsibility for professional use of the AVC codec mainly because recording is too compressed that it may distort the accuracy of image acquisition, and it may add unnecessary compressional noise & artifacts into the video image.
And when this happens to critical imaging applications like medical video documentation where every pixel data is important for post surgery analysis, etc... this format is a sure no-go for such use.

I think it's more of a license issue than quality issue. As for critical work with H.264, given high enough bitrates, it is able to offer near-lossless quality. By the way, H.264 is the same as MPEG4-AVC, for it was developed under ITU-T and subsequently adopted as Part 10 of MPEG4. AVCHD is a commercial name for a certain profile under H.264.
________
No2 Vaporizer
 

Last edited:
I think it's more of a license issue than quality issue. As for critical work with H.264, given high enough bitrates, it is able to offer near-lossless quality. By the way, H.264 is the same as MPEG4-AVC, for it was developed under ITU-T and subsequently adopted as Part 10 of MPEG4. AVCHD is a commercial name for a certain profile under H.264.

Agreed with your point about the high bitrate recordings, but ultimately, compression is still compression....you lose data. Even within the AVC profiles, in particular AVCHD which is interframe based, it'll still not yield better quality then it's own AVC-Intra based profile, especially when recording fast motion images.

Then question falls back to how high you want your bitrate to go, and how big the file size eventually becomes. When the file sizes gets bigger due to higher bitrate recordings, will it eventually be more efficient to use other formats that will probably yield better quality?


I hope that this discussion will not end up like how many audiophiles who like to argue which burr-brown DACs are better, expensive = good, etc....but in my tests, and this was put thru to many of my clients, I did perform native footage recording of DVCProHD and AVC-Intra100. Both codecs yield about the same level of compression & bitrate, but in blind selections, many people always preferred the images from DVCProHD footage over the AVC-Intra100.

In the case of my test, many of my peers and colleagues in the video post industry who have dealt with Uncompressed video on a day-in and day-out basis also agreed about how DVCProHD compression still yield a better quality than AVC-intra based compression.

In my point of view, H.264 is still compression and no matter how lossless or near-lossless it is, raw data is being lost due to compression and these data is essential for post, especially if you need to do compositing & color grading work. It is no doubt a good & efficient codec, and for me, I always like it as a delivery format...not an acquisition format. :D
________
SUBARU WORLD RALLY TEAM HISTORY
 

Last edited:
And just in case you're prepping a video file for audio post, do take note that we prefer intraframe compressions and not interframe. If the file size is a problem, then break your film up into 'spools'. If your film is that long anyway, you should be doing everything professionally for the love of the effort in the film.
________
Buy vaporizers
 

Last edited:
I think it's more of a license issue than quality issue. As for critical work with H.264, given high enough bitrates, it is able to offer near-lossless quality. By the way, H.264 is the same as MPEG4-AVC, for it was developed under ITU-T and subsequently adopted as Part 10 of MPEG4. AVCHD is a commercial name for a certain profile under H.264.

If the compression is great for the image, then audio must be the one suffering. No one really cares if audio is compressed greatly do they? ;(
________
TOYOTA K ENGINE SPECIFICATIONS
 

Last edited:
If the compression is great for the image, then audio must be the one suffering. No one really cares if audio is compressed greatly do they? ;(

What do you mean? Can clarify please? The audio is compressed separately from the video and has its own set of parameters. You can compress the audio less or more depending on your allowable bitrate/file size and your audio quality requirements.
________
Vaporizer
 

Last edited:
What do you mean? Can clarify please? The audio is compressed separately from the video and has its own set of parameters. You can compress the audio less or more depending on your allowable bitrate/file size and your audio quality requirements.

My bad, was just winching because the DSLRs use the same codec. Was thinking about the audio in the capture stage. If it's for audio post, it don't matter so much because the accompanied audio on the video file is only used as a reference for sync. The actual audio used is from the OMF file.
________
MARIJUANA INDICA
 

Last edited:
To contribute to audio...i find no one complains about the audio being "weak" and "muddled" when using PCM. less dvd time but the superior sound tradeoff i think in most cases work quite well since most of the events i cover at most is 8mins~15mins max. Even the long winded holiday videos i receive feedback sounds better when played back on their HT system.

Video on the other hand...if not doing broadcast work, i find mp4 h.264 is the most flexible to use with either on windows/mac platform. AVCHD is too much pain in my opinion if need to bring here and there or require onsite work.
________
Suzuki Gsx-R1000
 

Last edited:
Agreed with your point about the high bitrate recordings, but ultimately, compression is still compression....

...

:D

The thing is, given the same bit rate, H.264 offers the best flexibility - in other words, trade-off range between complexity and compression ratio. You can simply do it from all-intra, all the way up to blended multiple references at quarter-pixel motion precision. It also offers the best quality given the same bitrate. As for loss in quality - H.264 encoder can be strictly lossless if one choose so.

However, for the best compression ratio, one needs very accurate motion estimation and long group-of-pictures. These two requirements imply very long processing time and large memory usage, and are not suitable for video editing. Besides, video editing systems have to work on decoded (online or offline) videos frames. Aiming for compression efficiency doesn't make much sense. So I completely agree with you that H.264 (especially high profile) is more for content delivery.
________
Chevrolet bel air specifications
 

Last edited:
I just did a AVC-Intra100 side-by-side footage comparison with AVCHD @ 24Mbps.

Test scenario as such:

HPX372-09d.jpg

Camera source - AG-HPX372, records straight into P2 AVC-Intra100 at 1920x1080, 50i

HMR10Unit.jpg

Using my AG-HMR10 AVCHD recorder, I use the camera's HD-SDI output and record into PH mode which yields a 1920x1080 recording with variable bitrate of up to 24mbps.


In my post workflow using Final Cut Pro 7, I ingested my AVC-Intra100 footage as native P2 format, and AVCHD footage transcoded to the best possible resolution using Apple ProRES HQ.

Pipeline01.jpg
JVCmon02.jpg


Both footage were cut into an uncompressed 10 bit 4:2:2 HD full raster timeline, and image was output using AJA IoHD via HD-SDI to a proper JVC broadcast monitor.



Conclusion:

AVCHD footage looks a bit softer than AVC-Intra100.
Compressional artifacts are most prominent during fast moving objects and highly detailed objects. They are also most observable during a dissolve or a defocus to focus pull focusing.
Dark, low-contrast & black areas are observed to be very noisy, with lots of artifacts floating around.

Comparing these two codecs, AVC-Intra100 definitely holds the image better, but somehow, sub-consciously I can see a bit of noise with micro blocks of compression artifacts floating around....too small to be observable if using normal displays.


Probably because I'm very used to looking at HDCAM, HDCAM SR & uncompressed HD footage, the compression traits of AVCHD footage does irks me & caused some discomfort to my viewing. For most non-critical & consumer based viewing, I'd say that it's good enough.


I've got the uncompressed videos which is pointless to upload since it does nothing after a round of youtube compression, but anyways, here's a youtube preview:
http://www.youtube.com/watch?v=XVvSQChcJ3U

I've also done a DVCProHD vs AVCHD@24mbps comparison, and another DVCProHD vs AVC-Intra50 test will be on my list of work once I'm free. :thumbsup:
________
Cloudy trichomes
 

Last edited: