Perceptual image compression through fractal surface interpolation
Richard M. Dansereau, Witold Kinsner · 2002
This paper presents a perceptual image compression technique using fractal surface interpolation (FSI). This technique relies on the extraction and reconstruction of self-affine fractal surfaces with a specified Hurst exponent H*. Fractional Brownian motion (fBm) through midpoint displacement (MPD) provides the basis for generating these fractal surfaces between interpolation points. The MPD algorithm is designed to eliminate creasing often associated with MPD in two dimensions. Experimental results show that when applied to 512/spl times/512 8-bit gray-level images, perceptually good results are achieved with a bit rate of 0.35 bpp.