Adaptive Triangular Mesh Image Representation for X-ray Tomographic Reconstruction
Résumé
In this work is presented a new approach to Xray tomographic reconstruction method based on an adaptive triangular mesh (ATM). Unlike, what is done commonly in tomography, where the reconstructed volume is discretezed using a square regular grid. In the proposed method the volume is discretezed by a set of non-overlapping triangles in 2D, whose is associated a gray value. The reconstruction is performed by alternatively two steps: firstly, estimating the unknowns parameters using iterative method, and secondly updating the mesh in conformity with the estimate boundaries of the object, which are evaluated using a variational method. The latter is implemented using a single two-phase level set formulation. As result, the reconstructed image contains both the density map and the segmented regions of the object. The mesh image significantly reduces the number of unknowns parameters to be estimated, resulting in reduced computation time.