The USC Andrew and Erna Viterbi School of Engineering USC Signal and Image Processing Institute USC Ming Hsieh Department of Electrical Engineering University of Southern California

Technical Report USC-SIPI-226

“Image Compression with Full Wavelet Transform (FWT)”

by Kwo-Jyr Wong and C.-C. Jay Kuo

December 1992

Image compression based on a multiresolution approach has been intensively studied over the last ten years, including the Laplacian pyramid method by Burt and Adelson and the pyramidal wavelet transform (PWT) method by Mallat. In this research, we propose a modified wavelet transform known as the full wavelet transform (FWT) for image compression. By the FWT, we apply recursively the two-scale wavelet decomposition to all subimages so that an image is decomposed into blocks of the same size. It is shown experimentally that energy compaction is achieved in both the spatial and frequency domains via FWT, and can be effectively utilized to achieve high image compression ratio while preserving good image quality. Moreover, entropy coding is used for improve the overall performance. Numerical experiments show that our algorithm has a comparable performance with several existing methods. The relationship between the proposed method and other popular image compression methods such as DCT, PWT and SBC (subband coding) is also discussed.

To download the report in PDF format click here: USC-SIPI-226.pdf (1.2Mb)