Applications of high dimensional model representations to computer vision
Emre Demiralp · WSEAS Transactions on Mathematics archive · 2009
A new and powerful method for matrix decomposition is developed in this work. It is similar to singular value decomposition and the main idea comes from the univariate approximation of a function, given on a planar grid's nodes, by two variable high dimensional model representation. The proposed method is less iteration dependent than the singular value decomposition and the components are determined via straightforward steps containing recursions. It seems to have more capabilities than the singular value decomposition as an alternative method. An illustrative application is also given.