Words possibly included in a scene image shot by a mobile camera can be efficiently extracted using a word dictionary or a map database. Positional information acquiring means 101 measures a current position of the device to acquire positional information. Directional information acquiring means 102 detects a direction of the device to acquire directional information. Character recognizing means 104 determines a range of shooting of a scene image based on the current positional information and the directional information. The character recognizing means 104 extracts from a map database 103 information such as store names, building names, and place names associated with positions in the shooting range. Then the character recognizing means 104 conducts character recognition using word knowledge such as the extracted store names, building names, and place names.