Image feature extraction method, object search method, image feature extraction apparatus, object retrieval device

摘要

PPROBLEM TO BE SOLVED: To achieve a method of extracting a feature amount used for more accurately specifying a moving state of an object in video, and provide a method of predicting overlap, collision and disappearance of objects by using the feature amount, as well as an efficient video encoding method. PSOLUTION: An object extraction part 101 extracts position data 11a and low level metadata 11b of an object from a video signal 10; and a velocity calculation part 103 calculates velocity data 13 and an acceleration calculation part 104 calculates acceleration data 14, by using object feature amount data 15 last extracted for the object and already stored in an object feature amount buffer 106. A feature amount generation part 105 stores these data as object feature amount data 15 in the object feature amount buffer 106. A video state determination device determines a moving state of the object by using the object feature amount data 15. PCOPYRIGHT: (C)2011,JPO&INPIT