METHOD FOR ESTIMATING SCENE FROM IMAGE

PROBLEM TO BE SOLVED: To obtain a method for scene estimation which can efficiently and accurately estimate characteristics of the scene that an image represents by repeatedly transmitting local probability information to an adjacent node of a network, reading probability density at each node of the network, and estimat ing the scene. SOLUTION: Patches and their relative probability density are organized into a Markov network representing the statistical relationship between the scene and image. Through the Markov network 200, respective scene nodes update their reliability on the basis of local evidences which are gathered from other nodes and accumulated. The Bayes-like reliability of each node is repeatedly sent to an adjacent node with a message 161. Then the best estimation 171 of each node is read out as to a scene which is a base and hidden. Consequently, it is known what scene value the best estimation as to a target scene right below the position is.