Local likelihood regression of acoustic logging data with adaptive selection of multiple bandwidth

摘要

This paper deals with a regression model in a two-dimensional (2-d) domain and a local likelihood approach for estimating parameters in the model. The localization is realized by introducing a kernel function. Cross validation (CV) estimates give an insight into obtaining optimal bandwidth for the localization. We apply the method to acoustic well-logging image data and demonstrate noise reduction with the optimal bandwidth selection on a 2-d (space, time) domain. The method is generic enough to apply to various other types of data without restriction to geophysical data.