Reliable reconstruction strategy with higher grid resolution for limited data tomography

• Sensitivity analysis of four discretization grids and best combination of grid and solution technique.
• Generalized spatial filtering technique for iterative reconstruction method.
• Improved reliability and accuracy in reconstruction with high grid resolution.
• Reconstruction images show smoother boundaries thus better depiction of inner profiles.

A computerized tomography (CT) procedure is presented for reliable image recovery from limited data with the aim to improve depiction of the inner profile. Proposed strategy focuses on attaining (a) accuracy and reliability of the reconstructed properties like the shapes, sizes and densities, and (b) stability in presence of noise by optimally choosing various reconstruction parameters pertaining to smearing and grid size. Sensitivity analysis is performed with different grid types and their spatial resolutions. Generalized spatial smearing is proposed to alleviate artifacts due to the presence of non-participating pixels. Best results are obtained when an optimal smearing window size (w.r.t. the l2l2 norm) is employed. The proposed procedure is tested successfully for limited view data from both, the cyber domain and real world. It is observed that (i) explicit smearing is essential in limited data cases and (ii) optimal smearing parameters exist along with an interval of confidence within which reconstruction results can deemed to be reliable. We recommend using explicit optimal smearing for improved and reliable recovery from limited data. The contribution herein can be employed in the development of miniature CT setups for application where collection of full data is not possible.

Figure optionsDownload as PowerPoint slide