Ultrasound image segmentation using local statistics with an adaptive scale selection

We propose a region-based segmentation method based on local statistics. The adaptive spatial locality is defined using the Intersection of Confidence Intervals (ICI) approach. This pixel dependent local scale is estimated, conditionally on the current segmentation, in the sense of minimizing the mean-square error of a Local Polynomials Approximation (LPA). In other words, the scale is `optimal´ since it gives the best trade-off between the bias and the variance of the estimates. We provide a comparison with the single scale local region-based model. Results on simulated and real ultrasound images show that the proposed adaptive scale selection gives a robust solution to the attenuation problem.