Adaptive Huffman coding: Difference between revisions

Revision as of 13:08, 16 March 2016

Latest revision as of 23:50, 5 December 2024

(48 intermediate revisions by 34 users not shown)

Line 1:

{{short description|Data compression technique}}

'''Adaptive Huffman coding''' (also called '''Dynamic Huffman coding''') is an [[adaptive coding]] technique based on [[Huffman coding]]. It permits building the code as the symbols are being transmitted, having no initial knowledge of source distribution, that allows one-pass encoding and adaptation to changing conditions in data.<ref name="LiDrew2014">{{cite book|author1=Ze-Nian Li|author2=Mark S. Drew|author3=Jiangchuan Liu|title=Fundamentals of Multimedia|url=https://books.google.com/books?id=R6vBBAAAQBAJ|date=9 April 2014|publisher=Springer Science & Business Media|isbn=978-3-319-05290-8}}</ref>

The benefit of one-pass procedure is that the source can be encoded in real time, though it becomes more sensitive to transmission errors, since just a single loss ruins the whole code, requiring [[error detection and correction]].

==Algorithms==

Line 7 ⟶ 8:

===FGK Algorithm ===

It is an online coding technique based on Huffman coding. Having no initial knowledge of occurrence frequencies, it permits dynamically adjusting the Huffman's tree as data are being transmitted. In a FGK Huffman tree, a special external node, called ''0-node'', is used to identify a newly- coming character. That is, whenever a new data is encountered, output the path to the 0-node followed by the data. For a past-coming character, just output the path of the data in the current Huffman's tree. Most importantly, we have to adjust the FGK Huffman tree if necessary, and finally update the frequency of related nodes. As the frequency of a datum is increased, the'' sibling property'' of the Huffman's tree may be broken. The adjustment is triggered for this reason. It is accomplished by consecutive swappings of nodes, subtrees, or both. The data node is swapped with the highest-ordered node of the same frequency in the Huffman's tree, (or the subtree rooted at the highest-ordered node). All ancestor nodes of the node should also be processed in the same manner.

Since the FGK Algorithm has some drawbacks about the node-or-subtree swapping, Vitter proposed another algorithm to improve it.

===Vitter algorithm===

Some important terminologies & constraints :-

Code is represented as a tree structure in which every node has a corresponding weight and a unique number.

* '''Implicit Numbering''' : It simply means that nodes are numbered in increasing order by level and from left to right. i.e. nodes at bottom level will have low implicit number as compared to upper level nodes and nodes on same level are numbered in increasing order from left to right. In other terms, when we have built the Huffman tree, when merging two nodes into a parent node, we have set the one with the lower value as the left child, and the one with the higher value as the right child.

* '''Invariant''' : For each weight w, all leaves of weight w precede all internal nodes having weight w. In other terms, when we have built the Huffman tree, if several nodes had the same value, we prioritized merging the leaves over the internal nodes.

* '''Blocks''' : Nodes of same weight and same type (i.e. either leaf node or internal node) form a Block.

* '''Leader''' : Highest numbered node in a block.

Blocks are interlinked by increasing order of their weights.

Numbers go down, and from right to left.

A leaf block always precedes internal block of same weight, thus maintaining the invariant.

Weights must satisfy the sibling property, which states that nodes must be listed in the order of decreasing weight with each node adjacent to its sibling. Thus if A is the parent node of B and C is a child of B, then <math>W(A) > W(B) > W(C)</math>.

'''NYT (Not Yet Transferred)''' is a special node used to represent symbols which are ''<nowiki>'not yet transferred'</nowiki>''.

The weight is merely the count of symbols transmitted which codes are associated with children of that node.

[[File:Leaf step one.png|thumb|Slide_And_Increment(leaf node) sliding starts. P is a leaf node.]]

A set of nodes with same weights make a '''block'''.

[[File:Leaf step two.png|thumb|Slide_And_Increment(leaf node) sliding step 2. As P is leaf node, it slides in front of next block nodes of equal weight.]]

[[File:Leaf step three.png|thumb|Slide_And_Increment(leaf node) sliding step 3. Here we increase the current weight by 1.]]

[[File:Leaf step four.png|thumb|Slide_And_Increment(leaf node) sliding step 4. Method comes to an end. P is the new parent.]]

[[File:Internal one.png|thumb|Slide_And_Increment(internal node) sliding starts. P is an internal node.]]

[[File:Internal two.png|thumb|Slide_And_Increment(internal node) sliding step 2. Node P slides in front of next block of leaves nodes, with weight wt+1.]]

[[File:Internal three.png|thumb|Slide_And_Increment(internal node) sliding step 3. Now we increase the weight to 9. Thus the '''''invariant is maintained''''' as the current node is an internal node and should occur in front of leaf nodes of equal weight as we have increased the weight.]]

[[File:Internal four.png|thumb|Slide_And_Increment(internal node) sliding step 4. Now the 'P' points to the former parent ( as in the case of internal node according to algorithm)]]

'''algorithm''' for adding a symbol '''is'''

To get the code for every node, in case of binary tree we could just traverse all the path from the root to the node, writing down (for example) "1" if we go to the right and "0" if we go to the left.

leaf_to_increment := NULL

p := pointer to the leaf node containing the next symbol

'''if''' (p is NYT) '''then'''

Extend p by adding two children

Left child becomes new NYT and right child is the new symbol leaf node

''p'' := parent of new symbol leaf node

leaf_to_increment := Right Child of p

'''else'''

Swap p with leader of its block

'''if''' (new p is sibling to NYT) '''then'''

leaf_to_increment := p

''p'' := parent of p

'''while''' (p ≠ NULL) '''do'''

Slide_And_Increment(p)

'''if''' (leaf_to_increment != NULL) '''then'''

Slide_And_Increment(leaf_to_increment)

'''function''' Slide_And_Increment(p) '''is'''

We need some general and straightforward method to transmit symbols that are "not yet transmitted" (NYT). We could use, for example, transmission of binary numbers for every symbol in alphabet.

previous_p := parent of ''p''

'''if''' (p is an internal node) '''then'''

Slide p in the tree higher than the leaf nodes of weight wt + 1

increase weight of ''p'' by 1

''p'' := previous_p

'''else'''

Slide p in the tree higher than the internal nodes of weight wt

increase weight of ''p'' by 1

''p'' := new parent of ''p''.

Encoder and decoder start with only the root node, which has the maximum number. In the beginning it is our initial NYT node.

Line 32 ⟶ 73:

For every symbol that is already in the tree, we only have to transmit code for its leaf node.

====Example====

For every symbol transmitted both the transmitter and receiver execute the update procedure:

[[File:Adaptive Huffman Vitter.jpg|800px]]

# If current symbol is NYT, add two child nodes to NYT node. One will be a new NYT node, the other is a leaf node for our symbol. Increase weight for the new leaf node and the old NYT and go to step 4. If current symbol is not NYT, go to symbol's leaf node.

# If this node does not have the highest number in a block, swap it with the node having the highest number, except if that node is its parent<ref name=":0">{{cite web|url=http://www.cs.duke.edu/csed/curious/compression/adaptivehuff.html#tree |title=Adaptive Huffman Coding |publisher=Cs.duke.edu |date= |accessdate=2012-02-26}}</ref>

# Increase weight for current node

# If this is not the root node go to parent node then go to step 2. If this is the root, end.

Note: swapping nodes means swapping weights and corresponding symbols, but not the numbers. Notice that although the above description is excellent in capturing the principle of adaptive Huffman coding, it doesn't maintain Vitter's invariant that all leaves of weight w precede (in the implicit numbering) all internal nodes of weight w. See the example below for a more accurate explanation for Vittter's algorithm<ref name=":0">{{cite web|url=http://www.cs.duke.edu/csed/curious/compression/adaptivehuff.html#tree |title=Adaptive Huffman Coding |publisher=Cs.duke.edu |date= |accessdate=2012-02-26}}</ref>.

====Example====

[[File:Adaptive Huffman Vitter.jpg]]Encoding "abb" gives 01100001 001100010 11.

Step 1:

Line 59 ⟶ 93:

Step 3:

NYT spawns two child nodes: 252 for NYT and 253 for leaf node, both with weight 0. Increase weights for 253, 254, and root. To maintain Vitter’sVitter's invariant that all leaves of weight w precede (in the implicit numbering) all internal nodes of weight w, the branch starting with node 254 should be swopswapped (in terms of symbols and weights, but not number ordering) with node 255. Code for "b" is 11.

For the second "b" transmit 11.

It should be noted that forFor the convenience of explanation this step doesn't exactly follow Vitter's algorithm,<ref name=":0">{{cite web|url=http://www.cs.duke.edu/csed/curious/compression/adaptivehuff.html#tree |title=Adaptive Huffman Coding |publisher=Cs.duke.edu |access-date= |accessdate=2012-02-26}}</ref>, but the effects are equivalent.

Step 4:

Go to leaf node 253. Notice we have two blocks with weight 1. Node 253 and 254 is one block (consisting of leaves), node 255 is another block (consisting of internal nodes). For node 253, the biggest number in its block is 254, so swap the weights and symbols of nodes 253 and 254. Now node 254 and the branch starting from node 255 satisfy the SlideAndIncrement condition<ref name=":0">{{cite web|url=http://www.cs.duke.edu/csed/curious/compression/adaptivehuff.html#tree |title=Adaptive Huffman Coding |publisher=Cs.duke.edu |date= |accessdate=2012-02-26}}</ref> and hence must be swopswapped. At last increase node 255 and 256’s256's weight.

Future code for "b" is 1, and for "a" is now 01, which reflects their frequency.

Line 75 ⟶ 109:

{{reflist}}

* Vitter's original paper: J. S. Vitter, "[http://www.csittc.dukeku.edu/~jsv/Papers/Vit87.jacmACMversion.pdf Design and Analysis of Dynamic Huffman Codes]", Journal of the ACM, 34(4), October 1987, pp 825–845.

* J. S. Vitter, "ALGORITHM 673 Dynamic Huffman Coding", ACM Transactions on Mathematical Software, 15(2), June 1989, pp 158–167. Also appears in Collected Algorithms of ACM.

* Donald E. Knuth, "Dynamic Huffman Coding", Journal of Algorithm, 6(2), 1985, pp 163–180.

Line 83 ⟶ 117:

* [http://www.ics.uci.edu/~dan/pubs/DC-Sec4.html University of California Dan Hirschberg site]

* [http://www.cs.cf.ac.uk/Dave/Multimedia/node212.html Cardiff University Dr. David Marshall site]

* [httphttps://code.google.com/p/compression-code/downloads/list C implementation of Vitter algorithm]

* [http://www.cs.duke.edu/csed/curious/compression/adaptivehuff.html Excellent description from Duke University]

Line 89 ⟶ 123:

[[Category:Lossless compression algorithms]]

[[Category:Data compression]]