PROBLEM TO BE SOLVED: To provide a technique for robustly measuring variation level in facial expression from taken images regardless of changes in photographing environment.SOLUTION: A facial expression variation measurement device 1 comprises: image feature quantity extracting means 5 that extracts SURF (Speeded Up Robust Feature) in a taken image of an object as a characteristic point from a face region detected by face region detection means 4 clustering means 8 that classifies SURFs extracted from a reference face of the object into clusters, and that classifies SURFs extracted from input faces of the object into a most closest cluster with respect to the feature quantity in clusters classified SURFs of the reference face histogram creation means 9 that creates a histogram using the number of SURFs in a cluster as a frequency for each cluster with respect to each cluster classified on the reference face and each cluster classified on the input face respectively and difference calculation means 11 that calculates the difference between a reference face histogram 13 and an input face histogram 14 as facial expression variation level of the object.
