Thumbnail-based image coding utilising the fractal transform

In thumbnail-based retrieval from image databases the display of any fully featured (FF-) image at a workstation is commonly preceded by its selection from a display of reduced-size, thumbnail image (T-image). The effective implementation of thumbnail-based retrieval requires that the data supplied for the T-image be efficiently utilised in the production of the FF-image. We describe two image codecs based on the fractal transform that are thumbnail-based; one, called thumb fractal is a variant of block-oriented fractal coding; the other, vector quantization via fractal transform, VQFT. For both codecs signed correction tiles are derived via the fractal transform. Both codecs have as an initial image the exploded thumbnail, to which VQFT applies correction tiles, while thumb fractal iteratively applies similar correction tiles to a rapid convergence. Experimental data is presented for fixed range block size of 4/spl times/4, but is indicative of results for other image blocking schemes, such as quadtree.