Data compression techniques for Bangla text

Abstract

In recent years BangIa has been being used in computers. For efficient use of this language in computers it is very important to be able to store texts economically so that in terms of both storage requirement and transmission cost it is ! competitive. In this study efforts have been made to obtain economical coding of BangIa texts using static and dynamic Huffman codes, arithmetic codes and other important coding techniques. Performances of various coding techniques in coding BangIa texts of different types and formats have been considered in. terms of compression efficiency, coding and decoding times. Our result shows that arithmetic coding with scaling symbol counts has outperformed all the remaining coding techniques for off-line coding on general texts in BSCII format in term~ of coding efficiency. Compression efficiency for this algorithm varies between 24.80% - for lkb file and 34.92% _ for 200kb file. Although Vitter algorithm is the slowest in terms of coding and decoding times, it has been found best i~ terms of coding efficiency among all on-line coding algorithms having efficiency 28.40% for lkb file and 34.84% for 200kb file of general BSCII format texts. (i ) 1,\ There is a significant variation of efficiency and coding ~nd decoding times with respect to text formats. Non documellt BSCII format texts have been found to be. the most efficielt II and fastest whereas document BNA format texts are the slowest and most inefficient. Static Huffman coding techniques have been terms of coding and decoding times requiring "!I I found faster I" n roughly 16% tile, less than the arithmetic coding. Among the dynamic coding algorithms Vitter algorithm, being the slowest, takes rouihly 28% time more than the fastest static Huffman algorithm.