Dynamic Attention Map by Ising Model for Human Face Detection Masaru Tanaka.pdfVIP

Dynamic Attention Map by Ising Mo del for Human Face Detection Masaru Tanakay Kazuhiro Hottaz Takio Kuritay mtanakaetlgojp hottameicssaitamauacjp kuritaetlgojp z Taketoshi Mishima mishimameicssaitamauacjp y Electrotechnical Lab oratory z Saitama University Umezono Tsukuba ShimoOkub o Urawa Ibaraki JAPAN Saitama JAPAN Abstract implement the Dynamic Attention Map we make use of the Ising mo del which is the simplest mo del We pr esent a method to narrow down the search of magnetization The Ising mo del consists of spins space for scaleinvariant human fac e detection which interacting each other in some manner each of which uses Dynamic Attention Map implemented by Ising dy can take only one state of up or down at a time namics Combining the pr oposed method and the scale The spins will b e ipp ed through a certain dynamics invariant fac e detection method which is based on both dep ending on b oth the states of their neighb or spins HigherOr der Local Autocorrelation HLAC fe atures and an external magnetic eld It is natural to lo ok of LogPolar image and Linear Discriminant Analysis up on two states as faceand not face in face detec for face and notfac e classication it is shown that tion In our face detection metho d the distance in the face region in the imag