MATLAB: How can i compute probabalies in haffmaneco

MATLAB

how i can i compute probabalties to compress all letters small capital and numbers

Best Answer

[unique_symbols, ~, bin_number] = unique(YourText(:));
symbol_counts = accumarray(bin_number, 1);
symbol_probs = symbol_counts ./ numel(YourText);

Alternately,

sparse_count = sparse(1, double(YourText(:)), 1);
[~, unique_symbols, symbol_counts] = find(sparse_count);
unique_symbols = char(unique_symbols.');
symbol_probs = symbol_counts.' ./ numel(YourText);

In both cases, after the code, unique_symbols will be a column vector of characters that exist in the text, and symbol_probs will be the corresponding probability.

The sparse() version has a minor flaw that it will fail if any characters in the text are binary 0.

Related Solutions

MATLAB: How to calculate the probability of each element in string

Although Huffman encode works on "strings", you can apply it to numbers also, of course. histcounts will be usual for your purpose.

The code can be simplified. E.g.:

flag=ismember(symbol,string(i));      %symbols
if sum(flag)==0

can be written as:

if ~any(string(i) == symbol)

This is at least confusing:

for i=1:1:size((count)');

Note that size replies a vector and 1:size(x) need not be, what you expect. Transposing the array only to get the wanted dimension as first output of size is a waste of time. Better:

for i = 1:size(count, 2)

or because count is a vector:

for i = 1:numel(count)

But this loop is not required at all:

prob = count / total;
ent  = -sum(prob .* log2(prob));

This is very inefficient also:

for i = 1:length(deco)
  str = strcat(str, deco(i));
end

The iterative growing of an array is a very bad programming pattern, because the costs grow exponentially. Better without a loop:

str = char(deco);

Finally: num2str(rgb_image) is a very bad idea. Imagine the first pixel has the red channel value 0.2543. Now you convert this by num2str to '0.2543'. Then the Huffman encoding will consider the probability of digits, but you want to encode the probability of colors. This is something completely different.

So start from scratch. 1. improve the code for the Huffman encoding. 2. modify it such, that it accepts numbers from 0 to 255 as inputs. 3. Convert your RGB image to the uint8 format. Afterwards you can apply the Huffman encoding to your image.

MATLAB: Arithmetic coding for image compression

reshape() the image into a vector. Apply arithmetic encoding.

Best Answer

Related Solutions

MATLAB: How to calculate the probability of each element in string

MATLAB: Arithmetic coding for image compression

Related Question