method of adaptive object extraction by feature element learning object dependent

摘要

The present invention relates to an object-dependent features cattle, more particularly, to extract the object to extract a more accurate object by learning an object-dependent characteristic small in the video image for search video method according to the adaptive object extraction method according to the study .; If the object extraction method using a conventional general features small it is difficult to accurately extract the object time also takes long, using object-dependent characteristic cattle has the characteristic minimum value different from itself according to the object, and therefore also different according to the environment Preview feature it is difficult to define the range of minimum value.; If the video or the still image frame to be extracted, type in the present invention, at first, extracts the object by using the general features small then some extent when the correct object is determined that the extraction since the learning-dependent additional feature of small of the extracted object, the object by extracting the object, as dependent more predetermined characteristics, the objects in the video image to more accurately extract the desired object-dependent characteristic to provide an adaptive object extraction method according to predetermined learning.