Data compression with Huffman coding; an efficient dynamic implementation using file partitioning

The authors present a further improvement to the Huffman method which is based on the dynamically changing frequencies of characters within a document. This approach divides the document into a number of partitions and reads one partition at a time into a buffer. Then a frequency table is prepared for the partition, and the table is stored along with the compressed data. This approach has shown great improvement, both in terms of compression time and storage, over the two earlier implementations despite the fact that it consumes a little extra space by storing the frequency table a number of times (albeit with differential contents). This approach is fast, since the document is read only once. The new method also provides the ability to access any partition randomly and decompress it, thus avoiding decompression of an entire (lengthy) document