Image Sequence Segmentation Based on a Similarity Metric

1 Research supported by NIH/NINDS Grant HL65425 Abstract--In this paper we present a new approach for clustering of time-sequence imaging data. The clustering metric used is the normalized cross-correlation, also known as similarity. The main advantage of this metric over the more-traditional Euclidean distance, is that it depends on the signal’s shape rather than its amplitude. Under an assumption of an exponential probability model that has several desirable properties, the expectation-maximization (EM) framework is used to derive two iterative clustering algorithms. In numerical experiments based on a simulated dynamic PET brain study, the proposed method achieved better performance than several existing clustering methods.