Integrated method to image compression using the discrete wavelet transform

Yi Qiang Hu*, Hung Hseng Hsu, Bing-Fei Wu

*Corresponding author for this work

Research output: Contribution to journalConference articlepeer-review

1 Scopus citations


In this paper, we propose an integrated image compression method, which contains the discrete wavelet transform, scalar quantization and some lossless codings, to gain higher compression ratios while maintaining the image fidelity. The discrete wavelet transform, a multiresolution technique, has the properties of entropy reduction and energy concentration in high frequency subimages. Scalar quantizer is applied to the high frequency components since those histograms can be modelled to the generalized Gaussian distribution. These subimages with small relative energy can be dropped entirely to compensate the compression ratio if the optimal scalar quantization is adopted. An innovative approach, named as revised run-length coding, is proposed to improve the compression performance. The idea of this approach is to represent the appearance of symbols of run-length codes in exponential expression for saving the storage in bits. One coding method, differential pulse coded modulation, is introduced to reduce the entropy of the lowest frequency subimage performed after the discrete wavelet transform and to achieve the high compression effect losslessly. The experimental results are compared with some well-known methods, for example, JPEG, (entropy constrained) vector quantization, fractal-based compression method and wavelet with variable-length coding.

Original languageEnglish
Article number622093
Pages (from-to)1317-1320
Number of pages4
JournalProceedings - IEEE International Symposium on Circuits and Systems
StatePublished - 12 Jun 1997
EventProceedings of the 1997 IEEE International Symposium on Circuits and Systems, ISCAS'97. Part 4 (of 4) - Hong Kong, Hong Kong
Duration: 9 Jun 199712 Jun 1997


Dive into the research topics of 'Integrated method to image compression using the discrete wavelet transform'. Together they form a unique fingerprint.

Cite this