The sequential study type non stationary image detection device, the program null which implements sequential study type non stationary image method of detection and its manner

摘要

PPROBLEM TO BE SOLVED: To provide the detection of a nonstationary video image in real time while performing sequential learning of video images input consecutively. PSOLUTION: A nonstationary cut is detected as a result of a determination using a part 101 for inputting a video image comprising consecutive time-series frames, a part 102 for obtaining a cut from the input video images and then extracting a feature amount from the frames to generate feature vectors based on the feature amount, a part 103 for selecting feature vectors corresponding to a specific time range from among the generated feature vectors to calculate a conversion parameter for mapping in a most appropriate identification space for identification based on a statistical distribution of the input feature vectors, and a part 104 for mapping the generated feature vectors in the identification space based on the conversion parameter to determine nonstationariness. PCOPYRIGHT: (C)2008,JPO&INPIT