Impact of Intra VLC?

[qyot27]

A few months ago I began using HCenc 0.23 almost exclusively for my MPEG-2 encoding, after years of using TMPGEnc to do it. To be honest, I've found nearly everything I could do before (and then some) with HC. The only thing that's missing is the option to use the Intra VLC tables option, something which I noticed that ffmpeg could do when I was testing it out (my reason for not using ffmpeg was due to artifacts on the first few frames that I didn't get when using either TMPG or HC, and pulldown issues - and the only way to solve the artifacts issue was to set an abnormally high quantizer setting that would overinflate the bitrate of the rest of the scenes and come close to the 8000kbps ceiling).

Anyway, this strikes to the heart of the question: how much of an impact does using Intra VLC actually make on the encoded files? Considering I know that measures dealing with the data coding can make a big difference (like CABAC in H.264), I was curious about the role such measures play in MPEG-2.

[akupenguin]

I don't use mpeg2 so I don't have any numbers to give you, but you can find out yourself: ivlc affects only entropy coding, so if you use constant qp its effect will show up entirely as a change in bitrate, which makes it easy to measure.

[hank315]

In MPEG2 differences between intravlc 0 and intravlc 1 are small, usually < 0.5%.
Intravlc 0 is a bit more effective on low bitrates, intravlc 1 should be better for high bitrates.

ATM HCenc always uses table 14, that's intravlc 0.
I will add as an option (*INTRAVLC 0/1) in the next release.

[qyot27]

Thanks for the explanations. Is there a certain boundary above which intravlc 1 starts becoming more effective, or is it more dependent on how the bitrate is distributed in the video being encoded?