Total Variation regularized reconstruction for enhancing the quality of few-view industrial computed tomography applied to image analysis and metrology

Abstract

The few-view image reconstruction problem is one of the challenging research areas in industrial Computed Tomography (CT). On the one hand, acquiring enough data for reconstruction leads to a longer scanning time, which may not be applicable in industrial CT. On the other hand, reconstructing under-sampled data leads to artifacts and inaccurate image analysis. To get a usable reconstruction for image analysis, from the insufficient data caused by quick cone beam CT scans, we use a Total Variation (TV) regularized optimization problem. A split Bregman implementation is used to solve the TV regularized CT reconstruction problem. To evaluate the quality of the few-view reconstruction produced by the split Bregman, we perform gray value analysis and porosity analysis. The comparison is done against the established reconstruction algorithms OSSART and FDK approximate formula, using real and realistically simulated CT data in cone beam geometry.