We proposed a novel framework on how to improve the compression performance through the idea related to the decreased spatial resolution down-sampled by the LP (Laplacian Pyramid) transform, and only code the low-pass band. Before being compressed, our method intentionally down-sampled the image as part of the pre-processing step, and then applies the directionlet interpolation as post-processing step of the compressed image. At low bit-rates an appropriately down-sampled image compressed using multi-codec and later interpolated via directionlet, can be visually and objectively better than the image compressed directly with the codec scheme at the same number of bits. Down-sampling and up-sampling play a key role in our proposed novel code means which support spatial scalability. This scheme can be utilized to efficiently optimize the overall quality of the reconstructed image, especially the rich textures or edges preservation. We compared with the conventional JPEG2000 coding method and showed our framework can get comparable good results at low bit rate and is independent with the coding method.

Date Published: 30 October 2009
PDF: 8 pages
Proc. SPIE 7498, MIPPR 2009: Remote Sensing and GIS Data Processing and Other Applications, 749851 (30 October 2009); doi: 10.1117/12.833587

Published in SPIE Proceedings Vol. 7498:
MIPPR 2009: Remote Sensing and GIS Data Processing and Other Applications
Faxiong Zhang; Faxiong Zhang, Editor(s)