CS 536 Computer Graphics NURBS Drawing Week 4, Lecture 6

Transcription

1 CS 536 Computer Graphics NURBS Drawing Week 4, Lecture 6 David Breen, William Regli and Maxim Peysakhov Department of Computer Science Drexel University 1

2 Outline Conic Sections via NURBS Knot insertion algorithm The de Boor s algorithm for B-Splines for NURBS Oslo Algorithm Barycentric Coordinates Discussion of homework #3 2

3 Conic Sections via NURBS Obtained via projection of the 3D parabola onto a plane Note: 3D Case: rational curve is a 4D object 2D Case: rational curve is a 3D object assign w to each control point From 3

4 Conic Sections via NURBS Define the curve with three control points Weights of first/last control points are 1 For center control point w<1 gives an ellipse w>1 gives a hyperbola w=1 gives a parabola Knot vector is {0.0, 0.0, 0.0, 1.0, 1.0, 1.0} 4 From

5 Conic Sections via NURBS: A Circular Arc The two sides of the control polygon are of equal length The chord connecting the first and last control points meets each leg at an angle θ equal to half the angular extent of the desired arc (for instance, 30 for a 60 arc) The weight of the inner control point is equal to the cosine of θ Knot vector is {0.0, 0.0, 0.0, 1.0, 1.0, 1.0} 5 From

6 Conic Sections via NURBS: A Circle What if we need an arc of >180? Idea: Use multiple 90 or 120 arcs B stitch them together with knots 0 = B6 = { 0.866,0.5,1} B 5 = { 0,2,0.5} B 4 = { 0.866,0.5,1} Example: 3 arcs of 120 B { 1.732, 1,0.5} 1 = B { 0, 1,1 } 2 = B 0,0,0,1,1,2,2,3,3,3 6 { 1.732, 1,0.5} 3 = From

7 Conic Sections via NURBS Example: 4 arcs of { 2 } B 1 = { 1,1, 2 } B 3 = 1,1, 2 { 2 } B 5 = 1, 1, 2 { 2 } B 7 = 1, 1, From 7

8 Knot Insertion Issue: More control points mean more control How do we add more points and keep same curve? From 8

9 Knot Insertion Basic Approach Decide where we d like to tweak the curve Add a new knot Find affected d-1 control points Replace it with d new control points Example: New knot at u=2.6 9 From

10 Knot Insertion Given: n+1 control points (P 0, P 1,..., P n ), a knot vector of m+1 knots U = { u 0, u1,..., u m } and a degree d B- spline curve C(u). Insert a new knot t into the knot vector without changing the shape of the curve. If t lies in knot span [u k, u k+1 ), only the basis functions for (P k, P k-d ) are non-zero. Find d new control points Q k on edge P k-1 P k, Q k-1 on edge P k-2 P k-1,..., and Q k-d+1 on edge P k-d P k-d+1 All other control points are unchanged. Note that d-1 control points of the original control polyline are removed and replaced with d new control points. 10 See

11 Knot Insertion Algorithm Create new control point Q j = (1 α j )P j-1 + α j P j Where α is defined as α j = t u j u j +d u j See 11

12 Properties of Knot Insertion Increasing the multiplicity of a knot decreases the number of non-zero basis functions at this knot At a knot of multiplicity d, there will be only one non-zero basis function Corresponding point on the curve p(u) is affected by exactly one control point p i In fact p(u) is p i! 12 Compiled from Lecture notes of Dr. Ching-Kuang Michigan Technological University

13 The de Boor Algorithm Generalization of de Casteljau's algorithm It provides a fast and numerically stable way for finding a point on a B-spline curve Observation: if a knot u is inserted d times to a B-spline, then p(u) is the point on the curve. Idea: We just simply insert u d times and the last point is p(u)! See 13 Compiled from Lecture notes of Dr. Ching-Kuang Michigan Technological University

14 The de Boor Algorithm 14

15 The de Boor Algorithm 15

16 The de Boor Algorithm 16

17 The de Boor Algorithm 17

18 The de Boor Algorithm 18

19 De Boor s Algorithm If u lies in [u k,u k+1 ) and u!= u k, let h = d If u = u k and u k is a knot of multiplicity s, let h = d s Copy the affected control points p k-s, p k-s-1,..., p k-d+1, p k-d to a new array and rename them as p k-s,0, p k-s-1,0,..., p k-d+1,0 for r := 1 to h do for i := k-d+r to k-s do { Let a i,r = (u - u i ) / ( u i+d-r+1 - u i ) Let p i,r = (1 - a i,r ) p i-1,r-1 + a i,r p i,r-1 } p k-s,d-s is the point p(u). 19 Compiled from Lecture notes of Dr. Ching-Kuang Michigan Technological University

20 De Boor s Algorithm (cont) for u := 0 to u max do { for r := 1 to h do for i := k-p+r to k-s do { Let a i,r = (u - u i ) / ( u i+p-r+1 - u i ) Let p i,r = (1 - a i,r ) p i-1,r-1 + a i,r p i,r-1 } p k-s,p-s is the point p(u). } 21 Compiled from Lecture notes of Dr. Ching-Kuang Michigan Technological University

21 Example of de Boor s Algorithm Degree 3 B-spline curve (i.e., d = 3) Defined by seven control points p 0,..., p 6 And knot vector: u = 0.4 u 0 = u 1 = u 2 = u 3 u 4 u 5 u 6 u 7 = u 8 = u 9 = u a 4,1 = (u - u 4 ) / ( u u 4 ) = 0.2 a 3,1 = (u - u 3 ) / ( u u 3 ) = 8/15 = 0.53 a 2,1 = (u - u 2 ) / ( u u 2 ) = 0.8 p 4,1 = (1 - a 4,1 )p 3,0 + a 4,1 p 4,0 = 0.8p 3, p 4,0 p 3,1 = (1 - a 3,1 )p 2,0 + a 3,1 p 3,0 = 0.47p 2, p 3,0 p 2,1 = (1 - a 2,1 )p 1,0 + a 2,1 p 2,0 = 0.2p 1, p 2,0 a 4,2 = (u - u 4 ) / (u u 4 ) = 0.3 a 3,2 = (u - u 3 ) / (u u 3 ) = 0.8 p 4,2 = (1 - a 4,2 ) p 3,1 + a 4,2 p 4,1 = 0.7p 3, p 4,1 p 3,2 = (1 - a 3,2 ) p 2,1 + a 3,2 p 3,1 = 0.2p 2, p 3,1 a 4,3 = (u - u 4 )/ (u u 4 ) = 0.6 p 4,3 = (1 - a 4,3 )p 3,2 + a 4,3 p 4,2 = 0.4p 3, p 4,2 22 Compiled from Lecture notes of Dr. Ching-Kuang Michigan Technological University

22 Similar but Different De Casteljau's: Dividing points are computed with a pair of numbers (1 - u) and u that never change Can be used for curve subdivision Uses all control points De Boor's These pairs of numbers are different and depend on the column number and control point number Intermediate control points are not sufficient d-1 affected control points are involved in the computation 23 Compiled from Lecture notes of Dr. Ching-Kuang Michigan Technological University

23 De Boor s: Curves 25 Animated by Max Drexel University

24 Oslo Algorithm A subdivision algorithm for B-splines, the basic idea: Take the curve with m+1 control points P 0 to P m Insert a knot in any point (0.5 maybe?) As a result you will have 2 new points P k and P k Take curves with m+1 control points P 0 P k, P k P m-1 and P 1 P k, P k P m Apply procedure recursively on each part 26

25 Oslo Algorithm 27 Recorded from:

26 Barycentric Coordinates By Ceva's Theorem: For any point K inside the triangle ABC Consider the existence of masses w A, w B, and w C, placed at the vertices of the triangle Their center of gravity (barycenter) will coincide with the point K. August Ferdinand Moebius ( ) defined (1827) w A, w B, and w C as the barycentric coordinates of K K = w A A + w B B + w C C B, w B A, w A K C, w C 28

27 Properties of Barycentric Coordinates Not unique Can be generalized to negative masses Can be made unique by setting w A + w B + w C = 1 w A = 0 for points on BC w B = 0 for points on AC w C = 0 on AB B, w B A, w A K C, w C 29

28 Area Ta: (b-p) x (c-p) /2 30

29 Calculating the Weights Given vertices A, B, C and Centroid K What are the weights, w A, w B, w C? x K = w A x A + w B x B + w C x C y K = w A y A + w B y B + w C y C Substitute w C = 1 - w A - w B x K = w A x A + w B x B + (1 w A w B )x C y K = w A y A + w B y B + (1 w A w B )y C 33

30 Calculating Weights (cont.) Solve for w A and w B w A = (x B x C )(y C y K ) (x C x K )(y B y C ) (x A x C )(y B y C ) (x B x C )(y A y C ) w B = (x A x C )(y C y K ) (x C x K )(y A y C ) (x B x C )(y A y C ) (x A x C )(y B y C ) w C = 1 - w A - w B 34

31 Programming Assignment 2 Process command-line arguments Read in 3D curve points and tangents Compute Bezier control points for curves defined by each two curve points Use HW1 code to compute points on each Bezier curve Each Bezier curve should be a polyline Output points by printing them to the console as an IndexedLineSet with multiple polylines, and control points as spheres in Open Inventor format 35