Flattenable Mesh Surface Fitting on Boundary Curves

Illustration for vertex flattenability—the inner angles before and after flattening the triangles around a vertex are identical

Method overview: (a) the given boundary curve (blue knots are endpoints of line segments) and the normal vectors (red arrows) defined on segments, (b) the computed optimal planar boundary and tessellated planar mesh patch, (c) the warped mesh surface from the planar mesh as the initial fitting patch (where the yellow arrows represent the normal vectors on the boundary of the warped patch), (d) the color map of flattenablity at vertices of the initial fitting patch, (e) the final flattenable mesh computed from (c), and (f) the corresponding flattenability map of (e)

Angles on a boundary vertex vb: (a) the value of α(vb) can be predicted by the projected angle on the plane defined by vb and the normal vector 12(ne++ne−), and (b) the value of an optimal θ(vb)∊R2 has a lower bound as been addressed in Lemma 1

The closed-path constraint and the position coincident constraint on the planar boundary

The steps for computing an initial patch fitting to the given boundary: (a) the given boundary and normal vectors for cross-tangent planes, (b) the computed optimal planar boundary, (c) the inner region is filled with regular triangles, (d) the boundary region is tessellated by the CDT (29) and the inner vertices are repositioned through an area-based smoothing, (e) the planar mesh is warped to interpolate the given boundary and cross-tangent planes, and (f) the color map of flattenability at vertices of the warped patch

Different constraints on the boundary yield different surfaces by the RBF-based warping: (a) only the positions and the normal vectors are constrained, and (b) all the positions, the normal vectors, and the cross-tangents are constrained

The comparison of (a) direct Newton’s routine versus (b) our multiple-loop algorithm for computing a FL mesh surface from an input of Fig. 5, where the results after 200, 400, and 800 iterations of direct Newton’s routine are shown in (a) and the results after 5, 10, and 20 outer loop iterations (i.e., accumulated 25, 50, and 100 steps) are shown in (b)

The wrinkle patch example: (a) the given boundary curve and the cross-tangent normals defined on it, (b) the initial warped patch M0 and its color map for flattenability, and (c) the result FL mesh surface patch its the color map

The hole filling example: (a) the given moai model with a large hole at the back of the model, (b) the 2D mesh patch computed from the hole boundary, (c) the result FL mesh surface patch for filling, and (d) the color map for flattenability on the result FL mesh

An example for the flattenable surface modeling in apparel industry: (a) the curve network for fitting flattenable mesh surfaces—the normals of cross-tangent surfaces are defined on the curves (see the yellow arrows), (b) the mesh surface after initial fitting and its flattenability map, and (c) the final FL mesh surface patches fitting the given curve network

An example for modeling a shoe: (a) the input curve network with normal vectors, (b) the surface after initial fitting and the color map for illustrating vertex flattenability, and (c) the final FL mesh surface patches (unexpected wrinkles are generated at the region near arch)

A refined design for the shoe cover: (a) the input curve networks with normal vectors, (b) the surface after RBF warping (i.e., the initial fitting result), and (c) the final FL mesh patches that are automatically constructed by our approach

The 2D layout of leather patterns for making the shoe shown in the left, where the patterns are generated by Ref. 16 from the patches shown in Fig. 1

Discussion: (a) the computed optimal planar boundary may globally self-intersected (top) although the local self-intersection (bottom) has been prevented and (b) tangential continuity cannot always be preserved on the resultant patches (comparing normal vectors on the boundary curves at heel and other regions)