The Research Of Image Face Extraction Algorithm Based On Constraint Rule

With the continuous development of 3d reconstruction based on the geometrical information of the image,extracting the geometrical information of image is an important step in the realization of 3d reconstruction.It is important to study to obtain the geometric information from images.Extracting the points,lines and faces in the geometry information,especially extracting the face information is the important research direction.The method of extracting the points and lines is obtained by using the SUSAN cornor extraction algorithm and the Hough transformation.After obtaining the points and linear data of the image,extracting faces of image.In order to get the face of image,this paper established the basic theorem and assumption of the face extraction.Under the premise,put forward the strategy for extracting the faces of simple self-intersecting images and ordinary objects images.For simple self-intersecting images,algorithm will change the planar graph to the not self-intersecting images,and then search faces by depth first search.For the face extraction of ordinary objects,calculate the upper limit of the face by using the maximum rank equation to iterate,use the face adjacent theory and a series of face extraction constraint rules to obtain some face sets.And then algorithm will make the Heuristic search algorithm to get some face sets that the value of the standard is minimum,the face sets may be more than one group,in order to get the most reasonable face set,according to the human eye stereo vision features,get reasonable face set that is more likely to be true.Using different complexity images to test the algorithm,the experimental results show that the algorithm can effectively reduce the error occurring in the process of the extraction of point and line,and can be more accurate access to reasonable face of the polyhedron,but building's face extraction accuracy still needs to improve.