Locally - adaptive object recognition for digital image material

摘要

A method for the locally - adaptive object recognition for digital image material,a digital image by means of a plurality of different binarizing sch waves is binarized, so that are produced in accordance with a plurality of objective,in these individual binary and wherein a correlation analysis is carried out in binary form as a coherent regions in order to be segmented object hypotheses,characterized by the,that to all of the segmented object hypotheses features in the form of grey value vectors are temporarily stored as well as the corresponding rois,that for the elimination of more compartment hypotheses a clustering process is carried out of the rois needs,that within each cluster is a selection of a single of interest roi shall,and, on the associated object of interest roi hypothesis represents the object to be identified.