Huffman Coding with Non-Sorted Frequencies

1
Huffman Coding with Non-Sorted Frequencies

• Shmuel Tomi Klein
• Dana Shapira
• Bar Ilan University, Ashkelon Academic
  College, ISRAEL

2
Outline
Background and motivation
Using non-sorted frequencies
Dynamic compression of data packets
Relevance to other compression
Conclusions
3
(No Transcript)
4
Applications
5
Optimal Trees
6
Optimal Trees
7 5 3 2 3 2
22
7 5 5 3 2
12
10
7 5 5 5
5
5
7
5
10 7 5
12 10
3
2
2
3
22
Non-Huffman Tree
7
Optimal Trees
7 5 3 3 2 2
22
7 5 4 3 3
12
10
7 5 6 4
6
4
7
5
10 7 5
12 10
3
2
3
2
22
Non-Huffman Tree
8
(No Transcript)
9
(No Transcript)
10
Start with any order Then
use 2 queues
11
Partial Sort parameter K
4-grams 3-grams 2-grams 1-grams Tests
21886 6026 808 52 English
56078 18864 2965 131 French
12
Average bits/char vs partition blocks
1-grams
13
Average bits/char vs partition blocks
French
7
6
5
4
3
2
8
32
128
512
2048
14
Dynamic compression of data packets
15
Bub-For-5 Bubble Blocked Block size Bigrams Bigrams
5.06 5.06 3.81 2000 Comprs
5.06 5.06 3.81 20000 Comprs
11.6 7.3 30.1 2000 Time
15.2 11.1 37.4 20000 Time
6.35 6.34 4.11 2000 Comprs
6.34 6.34 4.11 20000 Comprs
14.0 9.9 286.2 2000 Time
17.6 13.4 290.4 20000 Time
English
French
16
Relevance to other compression schemes
17
Conclusion
Compresion / Time tradeoff
18
Thank you !