Direct parametric object detection in tomographic images

摘要

This paper describes a new method of parametric segmentation of tomographic data without actual image reconstruction. It is possible to detect lines, circles, ellipses and other geometric figures from the raw tomographic data itself. If, for instance in non-destructive testing, we are interested in determining faults such as cracks, it is possible to do so without reconstruction and with lesser computation. Experimental results are provided for both simulation and real data. This technique is very useful in non-destructive testing for finding faults in metallic parts.