Talk:Entropy coding

Is "entropy coding" the same thing as "lossless compression"?

Remark: This article contains some problems that appear not worth correcting because the article seems approximately fully-redundant with the article on lossless compression. Also "entropy coding" would be better a better subject title than "entropy encoding". -- comments by 4.65.146.242 moved from article

Entropy encoding is only a subset of lossless compression. LZ77 compression, for instance, is an important compression technique that isn't any form of entropy encoding. -- Antaeus Feldspar 20:43, 4 Jun 2005 (UTC)

Actually, I don't know where Feldspar gets his definitions, but I mostly agree with the prior comment by 4.65.146.242. In the usage I have experienced, the terms "entropy coding" and "lossless compression" are synonymous, and both terms apply to such things as Lempel-Ziv coding. I have never previously heard anyone assert that LZ coding or other such things are not entropy coding. The content of this page also seems rather overly simplistic. For example, I don't think what it describes is a very accurate description of what happens in arithmetic coding. I also agree that "entropy coding" is better than "entropy encoding". Pawnbroker 05:28, 5 Jun 2005 (UTC)

Then I don't know where you get your definitions, Pawn. People might use "entropy coding" as if it was a synonym of "lossless compression" (just as there are some major RFCs that incorrectly use "Huffman code" as if it was a synonym of "prefix code", instead of denoting only to those prefix codes created by a Huffman algorithm) but that's definitely not correct usage. Entropy encoding is encoding where each symbol is assigned a pattern whose length/cost corresponds to its entropy (hence the name). While entropy encoding is quite often used with LZ77 compression, as the two techniques complement each other, LZ77 is not an example of entropy encoding. -- Antaeus Feldspar 17:03, 5 Jun 2005 (UTC)

Can someone point to any textbook on information theory or any similar such authoritative source that draws a distinction between the two terms in this manner? Pawnbroker 05:46, 12 Jun 2005 (UTC)

Repetition of content

Much of this content seems to be saying that the length of the codes are assigned proportionally to the inverse logarithm of the probability. Couldn't this just be said once?