INTERPOLATION AND WAVELET COMPRESSION
Richard M. Dansereau, K Kinsner · 1997
This paper presents a perceptual image representation technique based on fractal surface interpolation (FSI). This technique is motivated from the observation that images taken from the real world con- tain many textures that are self-similar, or fractal, in nature. The fractal surface inter- polation representation is then compressed using a zero-tree wavelet compression sub- system with lossless entropy encoding. The fractal surface interpolation technique described relies on the extraction and recon- struction of self-aflne fractal surfaces with measured Hurst exponents H*. This gives statistically self-similar fractal surfaces used to represent textures in a real world image. Fractional Brownian motion (fBm) through a modified midpoint displacement (MPD) algo- rithm provides the basis for generating these self-affine fractal surfaces between interpo- lation points.