A heuristic method for separating clusters from noisy background

Abstract When points of clusters are confronted with points having a uniform density (noisy background), the separation of the clusters from the background is computed by heuristic statistical method, where the distribution of each point in the clusters is assumed to be bivariate normal. We try to maximize the log likelihood function of the clustering configuration. A modified hill-climbing pass algorithm is studied and the simulation results indicate that the algorithm is reliable and efficient. Also real data from astronomical photographs are tested with good results.