Geometric measures on arbitrary dimensional digital surfaces - PowerPoint PPT Presentation

Geometric measures on arbitrary dimensional digital surfaces Jacques-Olivier Lachaud, Anne Vialard LaBRI (Laboratoire Bordelais de Recherche en Informatique) - France Geometric measures on arbitrary dimensional digital surfaces – p.1/22

Outline 1- 3D example 2- nD digital surface: Definition Tracking Contours 3- Discrete tangent line: Definition Recognition 4- Geometric estimators: Normal vector Tangent planes Area Geometric measures on arbitrary dimensional digital surfaces – p.2/22

3D example Problem: how to compute the normal vector to each element of a digital surface ? [Lenoir96] , [Tellier / Debled-Renesson99] , [Coeurjolly02] 3D digital surface: boundary of a voxel set Surfel Geometric measures on arbitrary dimensional digital surfaces – p.3/22

3D example Problem: how to compute the normal vector to each element of a digital surface ? [Lenoir96] , [Tellier / Debled-Renesson99] , [Coeurjolly02] 3D digital surface: boundary of a voxel set (1 ) Exactly two 4-connected contours cross at a given surfel Geometric measures on arbitrary dimensional digital surfaces – p.3/22

3D example Problem: how to compute the normal vector to each element of a digital surface ? [Lenoir96] , [Tellier / Debled-Renesson99] , [Coeurjolly02] 3D digital surface: boundary of a voxel set (1 ) Exactly two 4-connected contours cross at a given surfel Geometric measures on arbitrary dimensional digital surfaces – p.3/22

3D example Problem: how to compute the normal vector to each element of a digital surface ? [Lenoir96] , [Tellier / Debled-Renesson99] , [Coeurjolly02] 3D digital surface: boundary of a voxel set (1 ) Exactly two 4-connected contours cross at a given surfel Geometric measures on arbitrary dimensional digital surfaces – p.3/22

3D example (2) Tangent computation on each 2D contour: discrete line segment recognition [Debled95] , [Vialard96] Geometric measures on arbitrary dimensional digital surfaces – p.4/22

3D example (2) Tangent computation on each 2D contour: discrete line segment recognition [Debled95] , [Vialard96] Geometric measures on arbitrary dimensional digital surfaces – p.4/22

3D example (2) Tangent computation on each 2D contour: discrete line segment recognition [Debled95] , [Vialard96] Geometric measures on arbitrary dimensional digital surfaces – p.4/22

3D example (2) Tangent computation on each 2D contour: discrete line segment recognition [Debled95] , [Vialard96] Geometric measures on arbitrary dimensional digital surfaces – p.4/22

3D example (2) Tangent computation on each 2D contour: discrete line segment recognition [Debled95] , [Vialard96] Normal vector: cross product of the two tangent vectors Geometric measures on arbitrary dimensional digital surfaces – p.4/22

3D example (3) Outer tangent plane: orthogonal to the normal vector and containing P P is the projection of the surfel centroid on the upper tangent lines (highest point). Upper tangent lines Geometric measures on arbitrary dimensional digital surfaces – p.5/22

Outline 1- 3D example 2- Multidimensional digital surface: Definition Tracking Contours 3- Discrete tangent line: Definition Recognition 4- Geometric estimators: Normal vector Tangent planes Area Geometric measures on arbitrary dimensional digital surfaces – p.6/22

� Multidimensional digital surface Digital space: cellular decomposition of R n into a regular grid [Khalimsky90] , [Kovalevsky89] , [Herman92] , [Udupa94] 4 ( x K , y K ) topo y x 3 11 01 01 (3,3) 2 2 00 01 10 (4,2) 10 00 01 (2,1) 1 ( y K ) 0 ( x K ) 0 1 2 3 4 Spel: n -cell (pixel in 2D, voxel in 3D) Surfel: n − 1 -cell Digital surface: set of oriented surfels Geometric measures on arbitrary dimensional digital surfaces – p.7/22

Object boundary Object O : set of spels Boundary of O : surfels separating spels of O from the background with an orientation Boundary / Coboundary operators: − q c ∈ ∆+ p ∇ c c c y + p x ∆ c = ∪ ∆ y c ∆ x c ∂O = � ∆+ p, p ∈ O Geometric measures on arbitrary dimensional digital surfaces – p.8/22

Boundary tracking Given a bel of O , find ∂O by following the surface of the object ⇒ adjacency between bels Direct followers of a bel along coordinate j : j (3) (3) c d + l ∈ ∆ c d c (2) Link l : (2) − l ∈ ∆ d (1) (1) Interior direct adjacent bel: the first direct follower that is a bel of ∂O c Geometric measures on arbitrary dimensional digital surfaces – p.9/22

Contour over a surface Given a bel c of ∂O and j � = ⊥ ( c ) , a contour over the boundary is the sequence of direct interior adjacent bels starting from c and going along directions ⊥ ( c ) or j . Such a contour is a 2D 4-connected discrete path (bels → edges, links → points). Geometric measures on arbitrary dimensional digital surfaces – p.10/22

Outline 1- 3D example 2- nD digital surface: Definition Tracking Contours 3- Discrete tangent line: Definition Recognition 4- Geometric estimators: Normal vector Tangent planes Area Geometric measures on arbitrary dimensional digital surfaces – p.11/22

Discrete 2D tangent line 4-connected discrete line of characteristics ( a, b, µ ) ∈ Z 3 : { ( x, y ) ∈ Z 2 , µ ≤ ax − by < µ + | a | + | b |} y U’ U ( a, b, µ ) = (1 , 2 , − 1) L ’ x L Leaning lines: ax − by = µ , ax − by = µ + | a | + | b | − 1 Geometric measures on arbitrary dimensional digital surfaces – p.12/22

Discrete 2D tangent line Update of a segment line when adding a point M : (1) M is in between the leaning lines: OK (2) ax M − by M = µ − 1 : M is just over the upper leaning line M U’ y y (1 , − 2 , − 1) (2 , 3 , − 1) U’ U U L ’ L ’ L x x L (3) M is just under the lower leaning line: similar Geometric measures on arbitrary dimensional digital surfaces – p.13/22

Discrete 2D tangent line Symmetric tangent centered on an edge ( a, b, µ ) = (0 , 1 , 0) Geometric measures on arbitrary dimensional digital surfaces – p.14/22

Discrete 2D tangent line Symmetric tangent centered on an edge ( a, b, µ ) = ( − 1 , 2 , − 2) Geometric measures on arbitrary dimensional digital surfaces – p.14/22

Discrete 2D tangent line Symmetric tangent centered on an edge ( a, b, µ ) = ( − 1 , 2 , − 2) Geometric measures on arbitrary dimensional digital surfaces – p.14/22

Discrete 2D tangent line Symmetric tangent centered on an edge ( a, b, µ ) = ( − 1 , 2 , − 2) Geometric measures on arbitrary dimensional digital surfaces – p.14/22

Discrete 2D tangent line Symmetric tangent centered on an edge ( a, b, µ ) = ( − 2 , 5 , − 5) Geometric measures on arbitrary dimensional digital surfaces – p.14/22

Outline 1- 3D example 2- nD digital surface: Definition Tracking Contours 3- Discrete tangent line: Definition Recognition 4- Geometric estimators: Normal vector Tangent planes Area Geometric measures on arbitrary dimensional digital surfaces – p.15/22

Normal vector at a bel Normal vector at bel c : unit vector orthogonal to the n − 1 tangent vectors at c . i = ⊥ ( c ) τ i : orientation of the cell of ∇ i c with greatest x i τ j : orientation of the cell of ∆ j c with greatest x j u ( c ) � � n ( c ) = � � u ( c ) � α j ( c ) ∀ j � = i, � u ( c ) · � e j = τ j β j ( c ) � u ( c ) · � e i = τ i Geometric measures on arbitrary dimensional digital surfaces – p.16/22

Visualization of 3D discrete objects Geometric measures on arbitrary dimensional digital surfaces – p.17/22

Tangent planes to an nD surface Centroid of c : � x c 0 . 5 α j − µ j | α j |− 1 On each contour: z + , z − j = z + j = j − 1 − β j β j The inner tangent plane passes through e i max j � = i z + � x c + τ i � j Geometric measures on arbitrary dimensional digital surfaces – p.18/22

Smoothing of 3D digital surfaces Geometric measures on arbitrary dimensional digital surfaces – p.19/22

Area of an nD digital surface Corrected area: � dσ ( c ) = � n ( c ) · � e i Averaged area: dσ ( c ) = 1 / ( � n − 1 k =0 | � n ( c ) · � e k | ) Geometric measures on arbitrary dimensional digital surfaces – p.20/22

Area of a 3D sphere 1.05 1.035 corrected area corrected area averaged area averaged area 1.03 1.04 1.025 1.03 1.02 1.015 1.02 ratio ratio 1.01 1.01 1.005 1 1 0.995 0.99 0.99 0.98 0.985 10 15 20 25 30 35 40 45 50 55 60 10 15 20 25 30 35 40 45 50 55 60 radius radius Geometric measures on arbitrary dimensional digital surfaces – p.21/22

Conclusion Results: Definition of a set of geometric estimators for multidimensional surfaces Efficient and generic implementation [Lachaud03] Convergence of the estimators to the continuous values Further works: Can we compute the normal vector field in a time linear with the number of surfels ? Curvature definition and computation Geometric measures on arbitrary dimensional digital surfaces – p.22/22