Automatic Object Detection on Tomographic Projections

In this study we present an approach used to truncate the empty space outside ROI on a stack of 2D tomographic projections by the means of locating the specific region of interest (ROI). In this work we shall consider the following definition of the region of interest-it is the area of the detector which only receives rays incident to the object during each measurement. Our goal is defined as-using a stack of 2D projection to find ROI and use it to truncate the amount of data processed by reconstruction block. We evaluate our approach on a phantom data to get the idea of how the amount of truncated pixels influence reconstruction time and signal-to-noise ratio. We use SART method as our primary tool of reconstruction. The approach is also applied to the real dataset. Significant improvement of the reconstruction time is achieved in both cases.