Modeling Parameter Space Behavior of Vision Systems Using Bayesian Networks

摘要

The performance of most vision systems (or subsystems) is significantly dependent on the choice of its various parameters or thresholds. The associated parameter search space is extremely large and nonsmooth; moreover, the optimal choices of the parameters are usually mutually dependent on each other. In this paper we offer a Bayesian network-based probabilistic formalism, which we call the parameter dependence networks (PDNs), to model, abstract, and analyze the parameter space behavior of vision systems. The various algorithm parameters are the nodes of the PDN and are associated with probabilistic beliefs about the optimality of their respective values. The links between the nodes capture the direct dependencies between them and are quantified by conditional belief functions. The PDN structure captures the interdependence among the parameters in a concise and explicit manner. We define information theoretic measures, based on these PDNs, to quantify the global parameter sensitivity and the strength of the interdependence of the parameters. These measures predict the general ease of parameter tuning and performance stability of the system. The PDNs can also be used to stochastically sample the parameter space, to select optimal parameter sets (e.g., in performance evaluation studies), and to choose parameters, given constraints on the choice of some parameters. We also offer a strategy based on stochastic learning automata to generate training data to automatically build these PDNs. The team of learning automata stochastically samples the parameter space in a nonuniform manner with more samples near optimal values. These nonuniform samples are used to infer both the dependency structure and the conditional probabilities of the PDN. We demonstrate the process of construction of the PDN for an isolated vision module with 4 parameters (an edge detector), a coupling of two vision modules with a total of 7 parameters (a small edge grouping module), and a combination of three vision modules with 21 parameters (a complex perceptual organization module).