Fig. 5.

Three-dimensional (3D) recontraction of leaf morphology based on explicit assumptions. A. Point cloud data of the leaf surface (left) and results of surface reconstruction using Poisson reconstruction (middle) and B-spline surface fitting (left). In this case, the duplicated reconstruction was recognized as the result of Poisson reconstruction, which is often used for this purpose, and resulted in over-estimation of the leaf area. The B-spline surface fitting avoided the problem by assuming that a leaf is a single two-dimensional surface in a 3D space. B. Simulation data of a leaf (left) and a leaf reconstructed using curve features (right). The contours and reconstructions of the simulation data are almost identical. There is a lot of noise at both ends of the slope. The contour is represented as a set of curve fragments.