A Survey of Mathematical Structures for Extending 2D Neurogeometry to 3D Image Processing

In the Big Data landscape, learning algorithms are often “black-boxes” and as such, hard to interpret. We need new constructive models, to eventually feed the Big Data framework. The emerging field of Neurogeometry provides inspiration for models in medical computer vision. Neurogeometry models the neuronal architecture of the visual cortex through Differential Geometry. First, Neurogeometry can explain visual phenomena like human perceptual completion. And second, it provides efficient algorithms for computer vision. Examples of applications are image completion (in-painting) and crossing-preserving smoothing. In medical computer vision, Neurogeometry is less known. One reason is that one often deals with 3D images, whereas Neurogeometry is essentially 2D (our retina is 2D). Moreover, the generalization to 3D is not mathematically straight-forward. This article presents the theoretical framework of a 3D-Neurogeometry inspired by the 2D case. The aim is to provide a “theoretical toolbox” and inspiration for new models in 3D medical computer vision.

[1]  R. Duits,et al.  Left-invariant parabolic evolutions on SE(2) and contour enhancement via invertible orientation scores. Part II: Non-linear left-invariant diffusions on invertible orientation scores , 2010 .

[2]  J. Koenderink,et al.  Representation of local geometry in the visual system , 1987, Biological Cybernetics.

[3]  Peter W. Michor,et al.  The Riemannian geometry of orbit spaces. The metric, geodesics, and integrable systems , 2001 .

[4]  N. Bourbaki,et al.  Lie Groups and Lie Algebras: Chapters 1-3 , 1989 .

[5]  Remco Duits,et al.  Left-Invariant Diffusions on the Space of Positions and Orientations and their Application to Crossing-Preserving Smoothing of HARDI images , 2011, International Journal of Computer Vision.

[6]  Giovanna Citti,et al.  A Cortical Based Model of Perceptual Completion in the Roto-Translation Space , 2006, Journal of Mathematical Imaging and Vision.

[7]  R. Duits,et al.  Left-invariant parabolic evolutions on SE(2) and contour enhancement via invertible orientation scores. Part I: Linear left-invariant diffusion equations on SE(2) , 2010 .

[8]  J. P. Jones,et al.  An evaluation of the two-dimensional Gabor filter model of simple receptive fields in cat striate cortex. , 1987, Journal of neurophysiology.

[9]  D. Hubel,et al.  Orientation columns in macaque monkey visual cortex demonstrated by the 2-deoxyglucose autoradiographic technique , 1977, Nature.

[10]  Giovanna Citti,et al.  Image Completion Using a Diffusion Driven Mean Curvature Flowin A Sub-Riemannian Space , 2008, VISAPP.

[11]  Jean-Paul Gauthier,et al.  Hypoelliptic Diffusion and Human Vision: A Semidiscrete New Twist , 2014, SIAM J. Imaging Sci..

[12]  Jean-Paul Gauthier,et al.  Anthropomorphic Image Reconstruction via Hypoelliptic Diffusion , 2010, SIAM J. Control. Optim..

[13]  Vijay Kumar,et al.  On the generation of smooth three-dimensional rigid body motions , 1998, IEEE Trans. Robotics Autom..

[14]  Jean Petitot,et al.  Neurogeometry of neural functional architectures , 2013 .

[15]  Jean-Jacques Risler,et al.  Sub-Riemannian Geometry , 2011 .

[16]  Francesco Rossi,et al.  Existence of planar curves minimizing length and curvature , 2009, 0906.5290.

[17]  Xavier Pennec,et al.  Computing Bi-Invariant Pseudo-Metrics on Lie Groups for Consistent Statistics , 2015, Entropy.

[18]  M. Postnikov,et al.  Geometry VI: Riemannian Geometry , 2001 .