Image compression using an adjustable basis function

  • Rune Dalmo
  • Jostein Bratlie
  • Peter Zanaty

Abstract

We investigate the performance of image compression using a custom transform, related to the discrete cosine transform, where the shape of the waveform basis function can be adjusted via setting a shape parameter. A strategy for generating quantization tables for various shapes of the basis function, including the cosine function, is proposed. Two signal fidelity measures, peak signal-to- noise ratio and mean structural similarity index, respectively, are computed for a few selected photos to benchmark the results.

Published
2015-02-18
Section
Articles