Segmentation of Images and Time Series Based on Spatial ARMA Processes

RONNY VALLEJOS; SILVIA OJEDA

Advances in Image Segmentation

Pei-Gee Ho

Año: 2012;

This chapter describes two problems, the first one is related to a certain algorithm to perform image segmentation. The second problem is about clustering of serially correlated sequences. In both cases we proposed new developments to approach these issues. The techniques we suggest in both problems rely on the estimation of spatial/temporal ARMA processes. Variants of the image segmentation algorithm rise from the way in which the prediction window is defined on the plane for a unilateral autoregressive process. On the other hand, a new distance measure is proposed to perform hierarchical clustering of time series. Numerical examples with real images and time series support our findings and illustrate how the new proposals work in practice.