The cutting detection device of the dynamic picture image data and the characteristic quantity selection means which select

摘要

PROBLEM TO BE SOLVED: To design an optimum cut detecting method by calculating a likelihood ratio of whether or not there is a cut from the appearance frequency of a feature quantity obtained from dynamic picture data, and performing likelihood ratio inspection, based on the likelihood ratio and detecting a cut from a dynamic image. SOLUTION: Data of a dynamic image are inputted from a dynamic image data input part 101. A feature quantity extraction part 102 extracts or calculates a specific feature quantity from the dynamic image. An appearance frequency counter 103 counts how many times a specific value of the feature quantity appears. A likelihood ratio calculation part 104 calculates a likelihood ratio of whether or not there is a cut from the appearance frequency data on the feature quantity counted by the appearance frequency counter 103. Then the likelihood inspection is performed by using the likelihood ratio thus obtained to detect the cut from the dynamic picture. Consequently, the optimum cut detecting method can be designed in a sense that, when one of the nondetection rate and excessive detection rate of the cut is fixed, the other can be minimized.