IMAGE RECOGNITION APPARATUS FOR OBJECTS IN GENERAL AND METHOD THEREFOR, USING EXCLUSIVE CLASSIFIER

摘要

Provided is an image recognition apparatus, wherein capability and reliability in image recognition of objects in general can be improved. In the image recognition apparatus (1), an image classification information accumulating means (20) has stored therein feature information that indicates image features of each of a plurality of types of objects obtained through a learning process, and when an input image is to be classified, an image feature-amount calculating means (18) extracts a descriptor indicating a feature-amount from the input image and votes for an image vocabulary that corresponds to the descriptor, and a classifying means (19) calculates an existence probability of one or more objects on the basis of the voted result, and specifies the types of objects that exist on the basis of the existence probability. When existence probabilities of each of the objects are to be calculated, an exclusive classifier is made to adjust the existence probabilities, using exclusive relationship information that indicates combinations of a plurality of different-type objects (object labels) that were assumed as not coexisting within the same image.