Reducing the complexity in a fractal-based image coder

Abstract

Presents a method for reducing the complexity of encoding still images with fractal based coders. The high computational effort of the exhaustive search of the transform parameters can be substantially reduced by taking advantage of the statistical properties of the discrete cosine transform. The selection of the important transform coefficients, from a visual point of view, provides a reduction of the dimensionality of the transform space that can be employed as a feature selection to design a vector quantizer. The encoder proposed provides a substantial reduction of the execution time of the algorithm, without a noticeable loss of quality.<>