A comparison of 2-d moment-based description techniques

摘要: Moment invariants are properties of connected regions in binary images that invariant to translation, rotation and scale. They useful because they define a simply calculated set region can be used for shape classification part recognition. Orthogonal moment allow accurate reconstruction the described shape. Generic Fourier Descriptors yield spectral features have better retrieval performance due multi-resolution analysis both radial circular directions In this paper we first compare various moment-based description techniques then propose method that, after previous image partition into classes by morphological features, associates appropriate technique with each class, i.e. recognizes class. The results clearly demonstrate effectiveness new regard techniques.