Conventional algorithms suffer the problem of ill-conditioning against noisy and sparse data for limited data computerized tomography. A simple iterative solution technique may converge towards a wrong solution if it fails to cope against any such effect. An adaptive algorithm may improve this situation by adjusting the number and location of the nodes during the discretization step. A simple adaptive scheme, in general, with simple solution technique works well with sufficient data problems. We report that it is not the same with limited data problems. Synthetic and real world data for limited view and detector tomography (LVDT) is presented in the analysis. Two different approaches of spatial filtering schemes are imbedded first time with adaptive process: (a) optimal smearing and (b) adaptive optimal smearing. A sensitivity analysis is also performed between uniform and non-uniform grids after incorporation of these schemes. Discretization frameworks are based on finite element methods (FEM). Entropy maximization is used due to its robust support towards any changes in the grid. Better results (than conventional adaptive schemes) are obtained with these modifications. The level of error can be improved further if discretization scheme avoids generation of inactive pixels.
展开▼
