Entropy Coding Based Image Compression using Set Partitioning in Hierarchical Trees (SPIHT)

Abstract

Images have become a vital part of our daily life. With the increase of images in use the storage and transmission has come into consideration. In the era where image compression was in consideration, emergence of Wavelet Transform (WT) has made it a lot easier as it represents a signal in terms of functions localized in both frequency and time domain, there are many wavelet transform based Image compression methods with progressive coding schemes. With the Embedded Zero Tree Wavelet (EZW) massive improvement was witnessed in image compression. EZW and SPIHT are used to attain better PSNR and compression ratios. EZW works on DWT to predict the absence of significant information by exploiting self-similarities across the scale. However, coefficient position is missing, didn’t cater for intra-band correlation and its performance with single embedded file was not much pronounced. And we had to share bit plane with it to the decoder. The improvements in EZW were brought in with the introduction of SPIHT, which is again a fully embedded codec algorithm. It uses principal of partial ordering by magnitude, set partitioning by importance of magnitude of the coefficients, self-similarity across the scale and ordered bit plan transmission. SPIHT encodes the transformed coefficients according to their significance. Statistical analysis has shown that the output bit-stream of SPIHT comprises of long series of zeroes which can be further compressed, therefore SPIHT cannot be used as sole mean of compression. To this end, additional compression is being done by making use of different kinds of entropy encoding schemes. One of the entropy encoding scheme which is concatenated with SPIHT for further compression is Huffman encoding. This research is motivated by the requirement of a viable solution for fast transmission and less storage space. This research concentrates on saving comparatively more number of bits without compromising the quality of the image by combining two encodings “Set Partitioning in Hierarchical Tree and Huffman coding. This is done by making deft use Huffman encoding where it yields the optimized results and saves more numbers of bits thereby reducing the storage space and increasing the transmission time.