h.264 Baseline Profile vs High Profile

[delanejenkins]

I am playing my videos on a networked media tank. To be compliant to the specs of the unit, standard definition needs to be encoded at level 3.0 with the baseline profile. I am aware of the difference in the max bitrate(which my encodes never come close to at CQ 18) and inability to use some of the x.264 encoding settings such as cabac, bframes etc. with baseline vs high. My question is if I am using the same constant quality factor of 18, same number of reference frames, and same max bitrate with both profiles, will I see a quality loss using baseline instead of high or will I keep the same quality but just end up with a larger file size with the baseline profile?

[Chengbin]

No. You'll end up with a FAR larger file with baseline (about 4 times)

Why use CQ? Use CRF

[LoRd_MuldeR]

Baseline profile means that you cannot use CABAC

And CABAC gives you ~10-30% (or even more) of extra compression at the same quantizer, compared to CAVLC:
http://akuvian.org/src/x264/entropy.png

[Chengbin]

I'm just curious, what media tank do you have that ONLY supports baseline profile? Even the AppleTV isn't that bad!

[benwaggoner]

Originally Posted by LoRd_MuldeR

Baseline profile means that you cannot use CABAC

And CABAC gives you ~10-30% (or even more) of extra compression at the same quantizer, compared to CAVLC:
http://akuvian.org/src/x264/entropy.png

And you lose B-frames, which are a pretty big hit as well.

[delanejenkins]

I have an HDX 1000. All networked media tanks use the same syabas processor and have the same specs which is baseline level 3.0 for standard definition and up to level 4.1 high profile for high definition. Nmt's will play almost anything you throw at them but will sometimes have some stuttering if the videos are not dxva compliant and close to the spec profiles. I am not too worried if the baseline profile makes my files larger than that of high profile. I just want to make sure I'm not losing quality.

[LoRd_MuldeR]

If the decoder chip can do High profile for HD content, it should be able to do High for SD content easily...

[delanejenkins]

Lord mulder I was thinking the same thing. I definitely do run into more problems with standard definition than high definition encodes tho. Do you think it's the use of more reference frames with standard definition that is the problem and not the profile or settings

[LoRd_MuldeR]

The number of reference frames is limited by the Level! More specifically by the maximum decoded picture buffer size (MaxDPB) defined by the individual Level.

And if the decoder chip can do Level 4.1 for HD content, then I see no reason why it suddenly should be limited to Level 3.0 for SD material.

Since the MaxDPB size remains the same for HD and SD, but SD material has a lower resolution (less data per frame), you can actually have more ref frames for SD at the same Level.

[delanejenkins]

I do see how the reference frames are limited by the levels according to the h.264 wiki. I have been using 8 reference frames for standard definition as compared to 4 reference frames for high def. This is why I question whether more reference frames can be causing the issues with standard def

[LoRd_MuldeR]

Originally Posted by delanejenkins

I do see how the reference frames are limited by the levels according to the h.264 wiki. I have been using 8 reference frames for standard definition as compared to 4 reference frames for high def. This is why I question whether more reference frames can be causing the issues with standard def

As long as the number of ref frames doesn't exceed the MaxDPB size, there should be no problem. And MaxDPB is defined by Level.

So if your device can do Level 4.1, then it has a maximum buffer size of 12.288 * 1024 bytes. Depending on the resolution, you can calculate how many ref frames can be used.

That means for 720x576 or 720x480 you could use up 16 ref frames. For 720p you can still use 8 ref frames. And so on...