Comparison of orthogonal basis functions for image compression.

This paper introduces a comparison of the orthogonal decomposition of two-dimensional signals (images) using a spectral transform in the different functions for compression and texture representation. The following basis functions DCT, Fourier, Haar, Hartley, and Walsh transform, are compared using system criteria analysis. This criterion includes error signal representation and computational cost based on the Bellman-Zade concept and fuzzy sets. It is shown that optimal values of the number of terms of the series were very close for the bases of trigonometric and Haar functions. The number of calculations in the Haar basis is one to two orders of magnitude less, which determines its predominant use in signal detection and recognition problems. But in tasks related to signal denoising, it is preferable to use the basis of trigonometric functions. [ABSTRACT FROM AUTHOR]