An effective technique for three dimensional reconstruction of craniofacial structure from the contours of CT slices is presented. Preprocessing is done to suppress the noise and distortion introduced during image acquisition. Discrete Curvelet transform is applied to extract the contours of individual slices. The edge points are stacked to obtain the 3D point cloud and it is Delaunay triangulated for reconstruction. The edge detection performance is compared with Laplacian, Sobel and Canny edge detectors. The final reconstructed 3D structure is compared with those edge-detector-based reconstruction (with the same set of 2D slices) on the basis of number of triangular indices and time complexity. Copyright © 2012 Inderscience Enterprises Ltd.