[Theora-dev] Question about Huffman Tables in Setup Header

VP3HoSwiYO vp3 at go8.enjoy.ne.jp
Sat Feb 12 19:15:32 PST 2005

  I think that this mechanism is not desirable for both theora
developers and users.

It is important to optimize Huffman code. According to some
Magazine, H.264 can decrease 10% bitrate by using Arithmetic
compression than Huffman code. And, Once I show you, Comparison
VP3 original Huffman code with RangeCoder under special condition.
In this comparison, Huffman code can't reach to RangeCoder.

However, Huffman code is not inferior to Arithmetic compression.
Good Huffman codes like a VP3's are always match for Arithmetic
But we have to know that VP3's Huffman codes are going too well.
We can imitate its performance easily?

To optimize codebooks, Developer will need a lot of time and a
lot of sample. And, It is difficlut to evaluate his new idea
using best Huffman code immediately.
Or, If developer can not produce optimized Huffman code, Users
will be forced to use wrong codebooks reluctantly or 2-pass
encode to minimize the size(bit-rate).
It is easy to guess that Creating such high performance Huffman
code is too hard. and I am also apprehensive that this mechanism
will be obstacle to many developers.

original japanese sentence are:




More information about the Theora-dev mailing list