Application of multiresolution spatial filters to long-axis tracking

A method was developed for automatically tracking the long axis of thin objects which have nonuniform width and arbitrary orientation in a two-dimensional image space. This method is used to determine the length of isolated contractile smooth muscle cells, but has applications in other medical areas such as angiographic imaging. Pattern recognition techniques that determine object size and orientation are used to identify the long axis of an imaged object from its responses to difference of Gaussian and orientation filters. This method needs no a priori knowledge of object location, adapts to varying image magnification, requires little human interaction, and yields reproducible results.<
>