Title of article :
Maximal codeword lengths in Huffman codes
Author/Authors :
Y. S. Abu-Mostafa، نويسنده , , R. J. McEliece، نويسنده ,
Issue Information :
دوهفته نامه با شماره پیاپی سال 2000
Pages :
6
From page :
129
To page :
134
Abstract :
In this paper, we consider the following question about Huffman coding, which is an important technique for compressing data from a discrete source. If p is the smallest source probability, how long, in terms of p, can the longest Huffman codeword be? We show that if p is in the range 0 < p ≤ 1/2, and if K is the unique index such that 1/FK+3 < p ≤ 1/FK+2, where FK denotes the Kth Fibonacci number, then the longest Huffman codeword for a source whose least probability is p is at most K, and no better bound is possible. Asymptotically, this implies the surprising fact that for small values of p, a Huffman codeʹs longest codeword can be as much as 44% larger than that of the corresponding Shannon code.
Keywords :
Data compression , Huffman codes , Fibonacci numbers
Journal title :
Computers and Mathematics with Applications
Serial Year :
2000
Journal title :
Computers and Mathematics with Applications
Record number :
919009
Link To Document :
بازگشت