Image Decomposition by Data Dependent Systems

This paper presents a new digital image processing approach called image decomposition. This approach is based on a recently developed system modeling, prediction, and analysis methodology called Data Dependent Systems (DDS). DDS is an innovative approach to the application and interpretation of the well known stochastic Autoregressive Moving Average (ARMA) models. The Green’s function form of these models provides a modal decomposition of the image data with boundary features captured in the model residuals and regional feature dynamics captured by the components of the Green’s function. This approach is unique in that it provides a method for image representation and scene identification which is not dependent on geometric descriptions.