Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 21 views

    SF PDF

    Uploaded by

    Duppalapudi Sainadh
    Shannon-Fano coding is one of the first lossless compression algorithms that assumes a probabilistic model of the data source. It works by recursively splitting the data into halves based on probability distributions until each subset contains a single element. An example demonstrates splitting a probability distribution into subsets and assigning binary codewords starting with 0 or 1 to each subset. The Shannon-Fano coding is shown to produce a non-optimal encoding in this example, as a slightly different encoding achieves a shorter average code length.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    SF PDF

    Uploaded by

    Duppalapudi Sainadh
    21 views2 pages
    Shannon-Fano coding is one of the first lossless compression algorithms that assumes a probabilistic model of the data source. It works by recursively splitting the data into halves based on probability distributions until each subset contains a single element. An example demonstrates splitting a probability distribution into subsets and assigning binary codewords starting with 0 or 1 to each subset. The Shannon-Fano coding is shown to produce a non-optimal encoding in this example, as a slightly different encoding achieves a shorter average code length.

    Original Title

    sf.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Shannon-Fano coding is one of the first lossless compression algorithms that assumes a probabilistic model of the data source. It works by recursively splitting the data into halves based on probability distributions until each subset contains a single element. An example demonstrates splitting a probability distribution into subsets and assigning binary codewords starting with 0 or 1 to each subset. The Shannon-Fano coding is shown to produce a non-optimal encoding in this example, as a slightly different encoding achieves a shorter average code length.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    21 views2 pages

    SF PDF

    Uploaded by

    Duppalapudi Sainadh
    Shannon-Fano coding is one of the first lossless compression algorithms that assumes a probabilistic model of the data source. It works by recursively splitting the data into halves based on probability distributions until each subset contains a single element. An example demonstrates splitting a probability distribution into subsets and assigning binary codewords starting with 0 or 1 to each subset. The Shannon-Fano coding is shown to produce a non-optimal encoding in this example, as a slightly different encoding achieves a shorter average code length.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 2

    Shannon-Fano Coding

    September 18, 2017

    One of the first attempts to attain optimal lossless compression assuming


    a probabilistic model of the data source was the Shannon-Fano code. It is
    possible to show that the coding is non-optimal, however, it is a starting
    point for the discussion of the optimal algorithms to follow.
    Assume that each letter in the alphabet occurs according to a fixed
    probability distribution. In particular, assume that the probability of a
    letter is the same regardless of where it appears in the input string.
    The algorithm is as follows. Let the probability distribution of the letters
    be (p1 , . . . , pn ). If there is only one letter, then we are done, and we encode
    the letter using 0 bits. Otherwise, we divide the probability vector into two
    halves of roughly equal probabilities. If any of the halves contains exactly
    one probability component, then we encode that using 1 bit. Otherwise, we
    set the Most Significant Bit of one half to 1, and the other half to 0. We
    repeat the algorithm on each of the halves.
    The second stage of the algorithm decides the second bit in each of the
    output encodings. Thus the leading bits in the four quarters have bits set
    to 00, 01, 10 and 11.
    The process repeats until each of the subsets are singletons. We illustrate
    the process with an example.
    Example 1. If the probabilities of the letters are

    {0.35, 0.17, 0.17, 0.16, 0.15},

    then the Shannon Fano coding proceeds as follows.


    The first stage divides the set into {0.35, 0.17} and {0.17, 0.16, 0.15}.
    Every string in the first set has an encoding starting with 0, and those in
    the second, with 1.
    In the second stage, the subsets are {0.35}, {0.17}, {0.17}, and {0.16, 0.15}.
    The first subset is encoded as 00, the second as 01 and the third as 11. The
    elements of the final subset have their leading bits set to 10.

    1
    In the final stage, the two-element set in the previous stage is split as
    {0.16} and {0.15}, encoded 100 and 101 respectively.
    The expected code length is

    0.35 ∗ 2 + 0.17 ∗ 2 + 0.17 ∗ 2 + (0.16 + 0.15) ∗ 3 = 2.31.

    This allocation is not optimal.


    The following encoding produces a shorter code on average.
    0.35 0
    0.17 100
    0.17 101 This code has expected length 2.3. ♦
    0.16 110
    0.15 111

    You might also like