A DISTRIBUTED PROBABILISTIC SYSTEM FOR ADAPTIVE REGULATION OF IMAGE-PROCESSING PARAMETERS

A distributed optimization framework and its application to the regula tion of the behavior of a network of interacting image processing algo rithms are presented. The algorithm parameters used to regulate inform ation extraction are explicitly represented as state variables associa ted with all network nodes. Nodes are also provided with message-passi ng procedures to represent dependences between parameter settings at a djacent levels. The regulation problem is defined as a joint-probabili ty maximization of a conditional probabilistic measure evaluated over the space of possible configurations of the whole set of state variabl es (i.e., parameters). The global optimization problem is partitioned and solved in a distributed way, by considering local probabilistic me asures for selecting and estimating the parameters related to specific algorithms used within the network. The problem representation allows a spatially varying tuning of parameters, depending on the different informative contents of the subareas of an image. An application of th e proposed approach to an image processing problem is described. The p rofessing chain chosen as an example consists of four modules. The fir st three algorithms correspond to network nodes. The topmost node is d evoted to integrating information derived from applying different para meter settings to the algorithms of the chain. The nodes associated wi th data-transformation processes to be regulated are represented by an optical sensor and two filtering units (for edge-preserving and edge- extracting filterings), and a straight-segment detection module is use d as an integration site. Each module is provided with knowledge conce rning the parameters to regulate the related processing phase and with specific criteria to estimate data quality. Messages can be bidirecti onally propagated among modules in order to search, in a distributed w ay, for the ''optimum'' set of parameters yielding the best solution. Experimental results obtained on indoor images are presented to show t he validity of the proposed approach.