SUBJECT IDENTIFYING METHOD, SUBJECT IDENTIFYING PROGRAM, AND SUBJECT IDENTIFYING DEVICE

摘要

An AdaBoost processing section selects the best discriminator exhibiting the lowest error rate from among binarization discriminators corresponding to respective predetermined feature values used for separating subject image samples and non-subject image samples, determines a weighting factor corresponding to the selected best discriminator, and repeats update of the sample weight based on the determined weighting factor so that the error rate of the best discriminator is 0.5 at the next learning. An integrated discriminator deriving section makes linear discrimination analysis using unbinarized data held in each of the selected best discriminators forming a discriminator group and thereby derives an integrated discriminator corresponding to the discriminator group. An integrated weighting factor determining section determines an integrated weighting factor corresponding to the integrated discriminator so that the error rate of the derived integrated discriminator is 0.5 at the next learning. A sample weight updating section updates the sample weight used by the AdaBoost processing section according to the determined integrated weighting factor. A face image identifying device is constructed to realize the above constitution.