1. The Huffman Algorithm
We use Huffman algorithm to encode a long message as a long
bit string
- by assigning a bit string code to each symbol of the alphabet and
- concatenating the individual codes of the symbols making up
the message.
Example: Alphabet consists of the four symbols A, B, C, D.
Symbol Code A B C D
The message ABACD would be encoded as
Such encoding is inefficient.

2. The Huffman Algorithm
If we examine any message, we will see that some letters appear
more frequently than others. If the frequently appeared letters are
assigned shorter bit strings, then the length of the encoded message
will be substantially reduced.
Symbol Code A B C D
The message ABACD would be encoded as
Such encoding is efficient.

3. Huffman Tree The message is ABACCDA Choose two symbols with
smallest frequency ( B and D).
Combine these two symbols
into the single symbol BD
of frequency 2. Next two symbols
with smallest frequency are C and BD
0 is assigned for left branch
1 is assigned for right branch
ACBD, 7 1
CBD, 4
A, 3 1
C, 2
BD, 2 1
Symbol code A B C D
B, 1
D, 1

4. Huffman Algorithm
Generally, codes are not constructed on the basis of the frequency of characters within a single massage alone.
Codes are constructed on the basis of their frequency within a whole set of messages.
The same code set is then used for each message.
For example, if messages consists of English words, the known relative frequency of occurrence of the letters of the alphabet in English language might be used.
The relative frequency of the letters in any single message is not necessarily the same.