Creases Transformation (CT)

Updated 3 July 2026

Creases Transformation (CT) is a unified framework combining geometric, mechanical, and computational approaches to model localized high-curvature features.
Analytical models and numerical simulations in CT reveal critical bifurcation parameters and scaling laws governing the transition from wrinkles to creases.
CT principles enable practical applications in computer vision and biometrics, such as precise shape matching, synthetic crease generation, and efficient polyhedral flattening.

Creases Transformation (CT) refers to a broad class of geometric, mechanical, and computational frameworks unifying the modeling, formation, synthesis, and exploitation of creases—localized high-curvature features—in a range of physical, digital, and algorithmic contexts. Across materials science, applied mathematics, computer vision, and digital biometrics, CT encapsulates analytic, variational, and data-driven formalisms for understanding creases as singularities, instabilities, geometric generators, and functional features.

1. Analytic and Mechanical Models of Crease Transformation

Creases in continuum mechanics are treated as singular transformations of deformed solids, distinct from wrinkling and linear buckling. In incompressible hyperelastic materials, crease nucleation is governed by bifurcation from the flat state triggered by a singular, three-fold braced-incremental-deformation (“bid”) eigenmode under lateral compression. The bid-field’s stability is quantified by a shape factor $S(\lambda)$ constructed from two path-independent conservation integrals. For neo-Hookean solids, the configurational stability criterion $S<1$ is maintained up to a critical compressive strain $\varepsilon_c\approx0.356$ ; at $S=1$ , a higher-order bifurcation initiates self-similar crease growth. The wrinkle instability, by contrast, corresponds to a regular (first-order) bifurcation at higher strain ( $\varepsilon_w\approx0.456$ ). Strain-stiffening models, such as the Gent law, elevate crease resistance by altering the $S(\lambda)$ relation and increasing $\varepsilon_c$ , eventually inhibiting crease formation entirely for strong enough stiffening (Song et al., 2023).

In swelling gels, a coupled series–asymptotic expansion approach reduces the nonlinear eigenproblem of crease formation to explicit ODEs. Analytical solutions yield three possible crease-formation pathways—setback (same-mode), mode-jump, and direct crease—depending on layer thickness. The critical crease strain $\epsilon^{\rm crease}\approx0.43$ and the shape-to-wavelength ratio at onset are independent of thickness. The bifurcation is subcritical, with a post-bifurcation snap-through transition and universal scaling laws for the evolution from wrinkles to creases (Chen et al., 2014).

The competition between wrinkles (smooth, scale-invariant, antisymmetric modes) and creases (symmetric, subcritical, localized cusps) is fundamental in soft solids. For instance, in the buckling of a triangular prism under axial compression, the transition from sinusoidal to creased modes is controlled by a critical ridge angle $\phi^*\approx90^\circ$ at which the wrinkle and crease bifurcation strains coincide ( $\epsilon_c(\phi^*)\approx0.44$ ). Scale-invariance alone does not guarantee localization; nonlinear energy focusing is necessary for the abrupt formation of a crease (Lestringant et al., 2017, Sigaeva et al., 2018).

2. Mathematical and Energetic Theories of Creased Surfaces

In differential geometry and the mechanics of thin sheets, curved-crease transformation (CT) links local curvature, folding angle, and developability. In isometric sheets, the geometry of a crease is parameterized by a space curve $S<1$ 0 with curvature $S<1$ 1, torsion $S<1$ 2, and a local fold angle $S<1$ 3. Isometry constraints relate the geodesic curvature $S<1$ 4 of the crease to its 3D configuration and opening angle. Explicit algebraic relations ensure the compatibility and developability of each sheet adjoining the crease. The total bending energy of the creased domain involves logarithmic contributions in the boundary cutoff, with stress focusing at the crease ends but finite overall energy (Mowitz, 2020).

The geometric mechanics of curved creases governs multistability and snapping transitions in shells. On a curved shell, a crease with nonzero normal curvature $S<1$ 5 generates a local bistable energy landscape, forbidding continuous folding between unfolded and folded states and enforcing fast snap-through transitions. Continuous folding is possible only along lines of curvature ( $S<1$ 6). Explicit design rules for multistable and monostable morphologies depend on the imposed crease geometry and the shell's material parameters (Bende et al., 2014).

3. Instability Theory: Wrinkles-to-Creases and Critical Transitions

The transformation from smooth wrinkling to sharp creasing is a hallmark of the nonlinear instability regime in soft solids. Linear analysis predicts a wrinkle instability (e.g., Biot's critical strain $S<1$ 7), but experiments and nonlinear computations reveal creasing nucleates at lower, subcritical strains ( $S<1$ 8 for neo-Hookean solids). In sectors under bending or eversion, finite element simulations show creases appear several percent before wrinkles, with amplitudes saturating rapidly due to self-contact. The critical location and spacing of creases typically correspond to the most dangerous mode predicted by linear bifurcation, although actual nonlinear evolution involves abrupt localization, period-doubling, and topological changes in the pattern (Sigaeva et al., 2018, Lestringant et al., 2017).

In a statistical-physics framework, surface creasing is mapped to a Kosterlitz-Thouless–type transition. Creases (cusped "sulci") are viewed as unbound quasi-particle excitations ("ghost fibers") whose emergence occurs above a critical strain, analogous to the unbinding of vortex–antivortex pairs. The transition is controlled by the balance of elastic self-energy and disorder-driven effective temperature, yielding universal critical strains closely matching experimental observations. The formation of finite-length creases ("ghost slabs") and two-dimensional crease arrays are described by electrostatic analogies and support pattern crystallography (Engstrom et al., 2017).

4. Algorithmic and Computational Creases Transformation

CT frameworks appear in computational geometry, shape processing, and computer vision. In polyhedral geometry, CT provides a rigorous method for planar flattening of arbitrary 3D manifolds. Allowing countably infinite, moving crease patterns enables the continuous, isometric flattening of any finite polyhedral surface—even nonconvex or non-orientable—without self-intersection and while preserving intrinsic distances. The construction involves recursive slicing into prismoidal slabs and the controlled placement and migration of crease lines, ensuring that the sum area supporting "moving creases" approaches zero as the subdivision refines (Abel et al., 2021).

In non-isometric shape matching, CT is realized via hybrid functional maps combining intrinsic (Laplace–Beltrami) and extrinsic (elastic thin-shell Hessian) spectral bases. This approach enables robust correspondence under deformations involving creation or sharpening of creases, beyond the reach of purely intrinsic or elastic descriptors. Empirically, such hybrid approaches significantly reduce mean geodesic errors, especially in datasets containing sharp features, such as finger creases or wrinkle lines, relative to baseline methods (Bastian et al., 2023).

5. Trait-Specific Image Synthesis and Biometric Applications

CT applies to visual biometrics, notably in trait-specific image synthesis for user verification. For example, synthetic generation of forehead creases exploits geometric modeling with piecewise-polynomial curves: principal (long) creases are represented by B-splines, and non-prominent (short) creases by quadratic Bézier curves. These are rendered as edge prompts on a 6×6 grid and input to a Brownian-Bridge diffusion model, conditioned to produce high-fidelity synthetic images. Intra-subject diversity is achieved via control point perturbation and targeted image-level augmentations (e.g., dropout, elastic deformations). Synthetic and real datasets are merged under a curriculum protocol to improve verification networks, demonstrating improved cross-database performance (Tandon et al., 23 Jan 2025).

6. Universalities, Scaling Laws, and Unifying Themes

Across systems, CT is linked to universal scaling laws, bifurcation types, and invariances. In swelling gels, the dimensionless product $S<1$ 9 (with $\varepsilon_c\approx0.356$ 0 the crease mode number) and critical strain are independent of thickness at onset. The scaling of profile shapes, wavelength, and amplitude are likewise thickness-invariant. In shells, the geometric design of crease curves controls snapping and multistability, with critical parameters (e.g., radius ratio $\varepsilon_c\approx0.356$ 1) set by material and geometric factors such as the Föppl–von Kármán number.

Algorithmically, CT constructions rely on isometry, compatibility, and developability constraints—whether in origami-inspired surface folding, creased shell actuation, or polyhedral flattening. Energetically, all frameworks trace transitions from delocalized to singular, high-curvature features to a fundamental balance between elastic energy, geometric localization, and topological constraints.

7. Open Problems and Computational Directions

CT remains an active area of research with diverse open problems. In mechanics, challenges include analytically characterizing the full nonlinear evolution from wrinkle to crease, especially in multisector or curved-geometry settings, and quantifying the energetic cost and scaling of crease arrays. In computational geometry, extension of CT-based flattening to smooth (non-polyhedral) manifolds and dynamic adaptation of crease networks in programmable materials present ongoing challenges. In computer vision, transferability of hybrid spectral CT frameworks to learning on non-manifold and topologically noisy datasets is an emerging direction. Biometrically, harnessing CT for synthesis and recognition in unstructured, heterogeneous populations continues to drive methodological and algorithmic innovation.

Table: Summary of Creases Transformation Formulations

Field	Core CT Model	Key Mathematical Objects / Mechanisms
Continuum Mechanics	Bid-field eigenmode, shape factor $\varepsilon_c\approx0.356$ 2	Singular perturbations, conservation integrals (Song et al., 2023)
Gel Physics	Coupled ODE reduction, subcritical bifurcation	Scaling laws, invariance at crease onset (Chen et al., 2014)
Shell Geometry	Curved-crease compatibility, snapping rules	Fold angle, normal curvature, multistability (Bende et al., 2014, Mowitz, 2020)
Soft Solids	Kosterlitz-Thouless (ghost fibers)	Quasi-particle unbinding transition (Engstrom et al., 2017)
Polyhedral Computation	Infinite moving crease construction	Piecewise isometry, local gadgets (Abel et al., 2021)
Computer Vision	Hybrid functional maps	Intrinsic/extrinsic spectral basis (Bastian et al., 2023)
Biometrics	Piecewise-polynomial visual synthesis	B-splines/Bézier, diffusion modeling (Tandon et al., 23 Jan 2025)

Each row corresponds to a distinct mathematical or algorithmic realization of CT, yet is united by the focus on localization, sharp features, and the management or exploitation of geometric singularities.