In order to accurately infer the actions or attributes of a subject on the basis of a plurality of images that have spatial or temporal continuity, the present invention comprises an image acquisition unit that acquires a plurality of images that have spatial or temporal continuity, a channel allocation unit that uses prescribed rules to allocate different channels to at least a portion of color gradient information and/or brightness gradient information that can be acquired from each of the plurality of images, a composite image generation unit that extracts the gradient information to which the channels have been allocated from each of the plurality of images, synthesizes the gradient information, and thereby generates a composite image that makes it possible to identify at least a portion of the gradient information from each image on the basis of the channels, and an inference unit that analyzes the composite image and makes an inference about the plurality of images.
展开▼