Image segmentation as the search for the best description in terms of primitives

A new paradigm is presented for the segmentation of images into piecewise continuous patches. Data aggregation is performed via model recovery in terms of variable-order bi-variate polynomials using iterative regression. All the recovered models are potential candidates for the final description of the data. Selection of the models is achieved through maximization of a quadratic Boolean problem. The major novelty of the approach is in combining model extraction and model selection.