摘要 |
A single image is obtained from among a plurality of temporal series images. At least one of a plurality of types of classifiers, each type of classifier judging each of a plurality of predetermined states of a predetermined subject, is employed to discriminate the state of the subject within the obtained image. At least one state is predicted for the subject within the obtained image, based on stepwise changes of the state of the subject obtained by previously discriminated states within temporal series images preceding the obtained image. The classifier corresponding to the predicted state is prioritized or weighted, when applying the classifiers to perform discrimination. |