New Set of Non-separable Orthogonal Invariant Moments for Image Recognition

摘要: It is known that the rotation, scaling and translation invariant property of image moments has a high significance in recognition. For this reason, seven presented by Hu are widely used field analysis. These finite order; therefore, they do not comprise complete set descriptors. we introduce paper another series infinite order, which based on normalized central moments. The non-orthogonal these causes redundancy information. To overcome problem, propose new construction technique non-separable orthogonal polynomials two variables recurrence formula present moments, to translation, rotation. approaches tested several well-known computer vision datasets including moment’s invariability, retrieval classification objects, latter fuzzy K-means clustering algorithm. performance for compared with some recent such as invariants multi-channel radial-substituted Chebyshev quaternion rotational Radon space Legendre–Fourier space. experimental results made using four databases images, namely Columbia Object Image Library (COIL-20) database, MPEG7-CE shape COIL-100 database ORL show our have done better than other descriptors tested.