Construction of an optimal signal-to-noise ratio quadtree using state space search techniques

This paper reports the results of a set of experiments that apply state-space search methods to explore efficient ways for online construction of an optimal quadtree. An optimal quadtree is defined as a quadtree that yields a minimum reconstruction error for a given compression rate. Although some researchers have proposed procedures for image compression with quadtrees most of these methods either refer to binary images or use an off-line method to fix the variance-threshold. The results of our experiments show that a very low distortion (measured in SNR) can be achieved in a reasonable computation effort. Comparison with the literature proves that in some cases, for a given compression, our quadtree construction algorithm yields less distortion than well known image compression methods such as DCT-based vector quantization, and wavelet-based vector quantization.