A collection of lossless data compression techniques.
A dictionary coder.
Supports any file format, .txt, .jpg, .pdf, .gif, .png, .avi, .zip, etc.
Let's compress a pdf file, Jeff's and Sanjay's paper on MapReduce.
./LZW/LZW_encoder samples/mapreduce.pdf
This produces LZW_encoded.txt, the compressed text and LZW_alphabet.txt, the set of symbols used in the original input.
Note, that the compressed LZW_encoded.txt is stored as a text file using only the characters '0' and '1'. In practice, the output would be stored using all 256 symbols, so the effective size of the compressed output is actually one eight of LZW_encoded.txt. Fast Lempel-Ziv-Welch (below) stores it using all 256 symbols.
To decode the encoded text, simply pass the compressed text and the input alphabet to the decoder.
./LZW/LZW_decoder LZW_encoded.txt LZW_alphabet.txt
The decoded output is stored in LZW_decoded.txt, which should correspond to the original input provided, in this case, samples/mapreduce.pdf.
A more mature, improved implementation of the above "slow" LZW. Debug data is kept to a minimum. The compressed output LZW_encoded.txt contains all 256 ASCII symbols (not just '0' and '1' as above).
The code can also be found in the LZW directory. Usage is even easier, since the alphabet is included in the compressed output. However, unlike the above LZW, this version uses a header, where the input alphabet, its length and the length of the compressed output is stored.
Encoding is done with the following line and produces the compressed file LZW_encoded.fastlzw.
./LZW/LZW_fast_encoder samples/mapreduce.pdf
To decompress, one now only needs to provide the compressed output, since the alphabet already included in there. The deccompressed file is stored in LZW_decoded.txt.
./LZW/LZW_fast_decoder LZW_encoded.fastlzw
The header of LZW_encoded.fastlzw stores the length of the alphabet, the alphabet (similar to the above LZW_alphabet.txt), the length of the compressed output
The compressed output follows right after the header.
Compresses sequences in which the same character occurs multiple times consecutively.
Currently only supports text files containing '0' and '1'.
Running the following line produces RLE_encoded.txt, the Run-Length encoded output.
./RLE/RLE_encoder samples/zero_one_sample.txt
To decode the encoded text, simply pass the compressed text to the decoder.
./RLE/RLE_decoder RLE_encoded.txt
The decoded output is stored in RLE_decoded.txt, which should correspond to the original input.
An entropy coder.
Supports any file format, .txt, .jpg, .pdf, .gif, .png, .avi, .zip, etc.
Let's compress a picture of a tiger wearing sunglasses.
./HUFFMAN/HUFFMAN_encoder samples/deal_with_it.jpg
This produces HUFFMAN_encoded.txt, the Huffman encoded output and HUFFMAN_code_words_mapping.txt, the mapping of symbols to codewords. (Some other files are also produced, you may ignore those.)
Note, that similarly to LZW, the Huffman encoding HUFFMAN_encoded.txt is stored as a text file using only the characters '0' and '1'. In practice, the output would be stored using all 256 symbols, so the effective size of the compressed output is actually one eight of HUFFMAN_encoded.txt.
To decode the encoded text, simply pass the compressed text and the codeword mapping to the decoder.
./HUFFMAN/HUFFMAN_decoder HUFFMAN_encoded.txt HUFFMAN_code_words_mapping.txt
The decoded output is stored in HUFFMAN_decoded.txt. As you can see by running
diff HUFFMAN_decoded.txt samples/deal_with_it.jpg
the decoded output corresponds to our original picture.