Lamb wave tomography offers a new dimension to the challenging field of in situ health monitoring of structures. The possibility of constructing tomograms from a network of sensors generating and sensing Lamb waves in thin, multi-layered, anisotropic composite plates is explored in the present study. It is shown that improved tomograms result when the anisotropic and attenuative characteristics of composite plates are accounted for by (a) using the newly identified energy of the earliest Lamb wave signals as the reconstruction parameter, (b) modifying the sensor configuration from conventional geometries (and thereby also optimizing the number of sensors), and (c) normalizing the Lamb wave energy data of the defective sample with respect to that of the defect-free sample.