Shapes and diffeomorphisms (Q5916078)

This is the second edition of a book previously published in 2010, and the author first presents the differences between the two editions chapter by chapter. From the previous edition, a chapter has disappeared, its content being partly added to another one, and a chapter has appeared. This second edition is composed of 13 chapters completed with 5 appendices and with a list of 316 references. \par Chapter 1 starts with generalities on parametrized plane curves, Jordan curves, regular curves and closed curves. Then the author defines the change of parameter, he proves the equivalence property of the reparametrization and he defines the notion of regular change of parametrization. He defines the unit tangent and normal vectors to a plane curve. Integration along a plane curve is introduced together with the analysis of arclength, and their properties are presented. Then the author presents the notion of curvature and its properties. He computes the above-defined quantities in Cartesian coordinates and in polar coordinates. He introduces the Frénet frame. He defines the enclosed area, he presents Green's formula and he derives the isoperimetric inequality. He defines the rotation index and winding number of a curve. He introduces an approximate value of the curvature based on finite differences and obtained through a least-squares method. He defines the distance map to a subset and especially to a curve and its properties. He defines the zero-level set of a function and the link with the preceding notion through the implicit function theorem. A long section of the first chapter is devoted to the impact of affine and projective transformations on the previously defined notions. The Chapter ends with the presentation of nonlocal representations and that of semi-local invariants. \par In the short Chapter 2, the author introduces the notion of medial axis or skeleton of a shape as the set of centers of discs of maximal radii inscribed in the shape. He presents examples and Voronoï diagrams as an efficient tool for the computation of the skeleton, and Chapter 2 ends with recovery tools. \par Chapter 3 moves to the 3D case. The author defines regular curves and their arclength and torsion. Then he defines $C^{p}$ parametrized regular surfaces and he presents classical examples. He introduces the tangent plane and differentials through local variations, the orientation and normals, the integration over orientable surfaces and regular surfaces with boundary. He then introduces the first fundamental form from which he deduces the notion of geodesics and the divergence theorem on surfaces and the Laplace-Beltrami operator. Through the second fundamental form, the author deduces the notion of curvature whose expression is computed in local coordinates. Chapter 3 ends with implicit surfaces and the Gauss-Bonnet theorem. \par Chapter 4 presents triangulated surfaces and the associated computations. The author computes an approximate curvature. He applies the Gauss-Bonnet theorem in this case and the variations of the area. He defines curvature measures first in the smooth case then in the present case and he introduces a discrete Laplace-Beltrami operator. Chapter 4 ends with the introduction of isocontours and isosurfaces for which the author presents different illustrations. \par In Chapter 5, the author analyzes the evolution of curves and surfaces with respect to a parameter which may represent the time. He starts introducing a general equation which describes the evolution of curves. He presents the grassfire equation as a simplified case and he derives the curvature equation for which he proves an existence result and he presents properties of this solution. He finally derives an implicit representation of the curvature motion. In the case of surfaces, he also introduces a general equation which describes this evolution, the general gradient flows and Chapter 5 ends with the analysis of the evolution of active contours and the associated numerical simulations. \par The short Chapter 6 is devoted to the analysis of deformable templates which are shapes obtained through transformations of a given prototype and the author here only considers the case of curves. \par In Chapter 7, the author defines the homeomorphisms and diffeomorphisms on an open subset $\Omega $ of $\mathbb{R}^{d}$ and the flows and groups of diffeomorphisms associated to ordinary differential equations which lead to existence of variations and of admissible Banach spaces. \par In Chapter 8, the author shows how to build admissible spaces, mainly Hilbert spaces, which induce groups of diffeomorphisms. He first defines the notion of reproducing kernel Hilbert spaces in the scalar case, for which he uses the scalar spline interpolation. In the vector case, he considers the vector spline interpolation and he computes derivatives of the kernel. He shows how to define a Hilbert space of smooth functions on vector fields using inner products associated to operators and the Fourier transform and he proves invariance properties of the inner products considering the operator or the kernel sides. Chapter 8 goes on with Mercer's theorem in the case of a positive definite kernel, the presentation of a thin-plate interpolation and it ends with a discussion on asymptotically affine kernels. \par In Chapter 9, the author investigates the relations between deformations and the objects they affect. He starts defining deformable objects and matching functionals associated to the discrepancy between two objects (the target one and the deformed one). He introduces the definition of a $(p,\infty )$-compliant function which allows to define the differentiation with respect to diffeomorphisms. He presents the relations with matching functionals and the chapter ends with different discussions: labeled point, image, measure, densities, curves and surfaces, vector fields, fields of frames and finally tensor matchings. \par Chapter 10 is devoted to diffeomorphic matching. The author starts adding with a penalization term to the matching functional (associated to discrepancy) which ensures the existence of a smooth solution to a matching problem. He briefly analyzes the Monge-Kantorovich problem. He then comes back to the links between diffeomorphisms and flows of ordinary differential equations for which he presents a penalty term which ensures the existence of a minimizer of the penalized functional. He derives the associated Euler-Lagrange equations that he links to a momentum conservation equation. He analyzes the properties of this equation. Moving to a more general expression of the functional to be minimized, he introduces optimization strategies for flow-based matching and Chapter 10 ends with the numerical treatment of such problems. \par In Chapter 11, the author introduces distances between deformable objects and their relations with group actions. He defines the notion of distance equivariant under the left or right action of a group and he analyzes the properties of such distances. Then he defines the invariant distance between point sets, of parametrized invariant distance between plane curves and of invariant metrics on diffeomorphisms, which leads to the geodesic equation. Chapter 11 ends with the presentation of Riemannian submersions and the relation with group action. \par Chapter 12 deals with special metrics called metamorphoses associated to the action of diffeomorphisms on objects. The deformations are supposed to belong to a Lie group with Lie algebra acting on a Riemannian manifold. The author defines the notion of metamorphosis. From the definition of infinitesimal metamorphosis, he derives a new Riemannian metric on the manifold which involves a minimizing procedure and he derives the associated Euler-Lagrange equations. Chapter 12 ends with applications to labelled point sets, images, densities and curves. \par In the final Chapter 13, the author presents shape analysis methods, considering datasets in which each object is a shape. He presents a principal component analysis and he finally considers the case where the dataset is a time series. \par The first appendix gathers elements from functional analysis. The second appendix gathers elements from differential geometry. The third appendix gathers elements concerning ordinary differential equations. The fourth appendix gathers elements about optimization and optimal control theory. The fifth appendix gathers elements dealing with principal component analysis. \par The book presents and discusses tools which are useful for researchers working in different areas but which all have to deal with curve or surface treatment, for example image treatment. The author who is a well-known specialist of this field presents many applications which illustrate the different notions which are introduced.