In this thesis we present an effective algorithm，which can extract holes automatically from IC images，and implement the practical tasks satisfactorily． First，holes are coarsely located by using wavelet decomposition to eliminate some non-hole regions．Then further elimination is done by using SOBEL operator， which is sensitive to horizontal and vertical edges．Using moment invariants and K-L transform，we can find the holes'essential features which is invariant to holes'size， shape and orientation and use these features to finally recognize holes．Meantime，we use histogram equalization to make up the slightly difference of gray level between images．
修改评论