William Patin

Information

This article was written on 28 Mar 2012, and is filled under Blog, English, Video.

Current post is tagged

, , , ,

Canon and the h.264 encoding profile

EOS DSLR users who own a camera prior to the 5D MkIII model are facing the problem of compression artifacts in their footage. I own a Canon 60D and have been annoyed by the poor compression quality in low light situations. Thinking of that, I started blaming the poor encoding settings and the algorithm that Canon must use in their cameras. In fact, when it comes to h.264 encoding, there are huge differences from software to software and there are even more parameters influencing the encoding quality in the background of most GUIs. So what could be the issue for the poor output quality of Canon DSLRs? The answer seems quite simple to me: it’s the h.264 profile.

As seen in Philip Blooms review, the 5D MkIII produces a dramatically better image in low light conditions than its predecessors (jump to 11:35). Most people are going to think that this is a result of a better sensor, I don’t. This is why I ran an unaltered clip from the 5D MkIII through the video analysis software “VideoSpec”  and these are the results:

*** Canon 5D MkIII – Video Track Parameters ***
- Format: H.264/MPEG-4 AVC
- Bitrate: Max.: Undefined / Average: 26.2 Mbps / Min.: Undefined
- Frame rate (fps): Max.: — / Average: 25.000 / Min.: —
- Encoding profile: High@L4.1
- Width (Pixel number): 1920*1080
- Pixel Aspect Ratio: 1:1
- Display Aspect Ratio: 16:9
- Interlacing: Progressive

Pay attention to the High encoding profile at Level 4.1. There are variety of encoding profiles at different levels, but the most common are the Baseline, Main and High profiles (find an easy explanaition here, and some more detailed information on Wikipedia). The High profile at Level 4.1 is mostly being used for high quality h.264 videos content in the broadcast industry and for Blu Rays. The higher the profile, the more complex the decoding process becomes but the picture quality increases.

Now let’s compare this data with different EOS camera models:

-5D MkII, Encoding Profile: Baseline@L5.0
-7D, Encoding Profile: Baseline@L5.0
-60D, Encoding Profile: Baseline@L5.0
-600D/ti3, Encoding Profile: Baseline@L5.0

The 5D MkIII uses the new, faster processor DIGIC 5+ which should explain its ability to record at High profile. According to Wikipedia, the models EOS 1100D/Rebel T3, EOS 500D/Rebel T1i, EOS 550D/Rebel T2i, EOS 600D/Rebel T3i, EOS 50D, EOS 60D and EOS 5D Mark II use the DIGIC 4 processor, while the EOS 7D and EOS-1D Mark IV use a faster Dual DIGIC 4 processor.
Now, whoever has seen footage from a Panasonic GH2 must have been asking themselves why they had opted for Canon? The footage is smooth and there are nearly no compression artifacts visible. The reason is the following, watch the specs:

*** GH2 – Video Track Parameters ***
- Format: H.264/MPEG-4 AVC
- Bitrate: Max.: 16.0 Mbps / Average: 15.9 Mbps / Min.: Undefined
- Frame rate (fps): Max.: Undefined / Average: 25.000 / Min.: Undefined
- Encoding profile: High@L4.0
- Width (Pixel number): 1920*1080
- Pixel Aspect Ratio: 1:1
- Display Aspect Ratio: 16:9
- Interlacing: Interlaced – Top Field First

Now, presuming that the EOS models using the DIGIC 4 sensors would be to slow for encoding h.264 at higher profiles, seeing that the GH2 encodes at High profile at Level 4.0, why shouldn’t the 7D be capable of that? Not even at Main Profile?

Of course there is Magic Lantern, but ML only increases the bitrate, not the recording profile. A clip encoded with the profile High@L4.0 and a bitrate of 50Mbps is going to have a much better quality than one encoded at a Baseline Level 5.0 and a boosted bitrate. If you want to see a direct comparison of a snapshot encoded at different profiles, head over to Martin Fiedler’s site and scroll down to the “Subjective quality” paragraph.

That said, I am not an expert on the field of video compression (I’ll let the geniuses study that), but I have to deal with encoding pretty often and I don’t understand how poor the signal processing power of the DIGIC 4 series must be if it can’t encode a video at Main profile. This leads me to the following questions:

If the DIGIC 4 processors are to slow for Main profile encoding, would it be possible at lower frame rates (25p & 30p)?

Where are the benefits of the Dual DIGIC 4 processor in video encoding for 7D owners?

Is the encoding profile being kept low by Canon in order to push customers to buy their more pricier/professional models?

Don’t get me wrong, I love my 60D and haven’t regretted buying it. In the end, it is clear that these devices remain what they were built for, they are photo cameras. I just hope to find some answers. If anyone is more qualified and can provide information on that, feel free to drop a comment, I’ll appreciate that.

  • Damian

    Very Illuminating. I would be curious to see if there are newer cameras available that may not have the sensor size or the low light capability of a 60d but produce cleaner images based on processor/Codec handling!

  • tt

    Note that the GH2 is cheating: it’s using interlaced frames.

Recent articles

  • Welche Kamera kaufen?20 Sep 2013
  • Wie werde ich gut in dem, was ich tue?13 Sep 2013
  • Wo finde ich gute Lernressourcen?20 Aug 2013
  • Die richtige Ausbildung19 Aug 2013
  • Welche Software ist die richtige?19 Aug 2013

Recent comments