An accurate model for quadtrees representing noiseless images of spatial data

In this paper we propose and analyze a new meaningful branching sequence to generate random quadtrees representing binary images. In particular, we show that this sequence produces expected distributions of external and internal nodes much closer to real data than all previous proposed approaches in the literature to model both random binary images and quadtrees. This new model provides a good compromise in representing images belonging to various classes, more or less structured. The effectiveness of the new proposed model is shown through a comparison with respect to nodes distributions of representative real spatial data images. The introduction of this new realistic model can have a large impact on the analysis of expected performances of a large class of algorithms for spatial data processing. First experimental results show that this new model closely simulate real cases