Academic Journals Database
Disseminating quality controlled scientific knowledge

A High Performance Modified SPIHT for Scalable Image Compression

Author(s): Bibhuprasad Mohanty, Abhishek Singh, Sudipta Mahapatra

Journal: International Journal of Image Processing
ISSN 1985-2304

Volume: 5;
Issue: 4;
Start page: 390;
Date: 2011;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: Zero Shifting | 2D SPIHT | Lifting Wavelet Transform | Context Modeling | Resolution Scalability | SSIM

In this paper, we present a novel extension technique to the Set Partitioning in Hierarchical Trees(SPIHT) based image compression with spatial scalability. The present modification and thepreprocessing techniques provide significantly better quality (both subjectively and objectively)reconstruction at the decoder with little additional computational complexity. There are twoproposals for this paper. Firstly, we propose a pre-processing scheme, called Zero-Shifting, thatbrings the spatial values in signed integer range without changing the dynamic ranges, so that thetransformed coefficient calculation becomes more consistent. For that reason, we have to modifythe initialization step of the SPIHT algorithms. The experiments demonstrate a significantimprovement in visual quality and faster encoding and decoding than the original one. Secondly,we incorporate the idea to facilitate resolution scalable decoding (not incorporated in originalSPIHT) by rearranging the order of the encoded output bit stream. During the sorting pass of theSPIHT algorithm, we model the transformed coefficient based on the probability of significance, ata fixed threshold of the offspring. Calling it a fixed context model and generating a Huffman codefor each context, we achieve comparable compression efficiency to that of arithmetic coder, butwith much less computational complexity and processing time. As far as objective qualityassessment of the reconstructed image is concerned, we have compared our results with popularPeak Signal to Noise Ratio (PSNR) and with Structural Similarity Index (SSIM). Both thesemetrics show that our proposed work is an improvement over the original one.

Tango Rapperswil
Tango Rapperswil

     Affiliate Program