490 likes | 672 Views
Splines III – Bézier Curves. based on: Michael Gleicher : Curves , chapter 15 in Fundamentals of Computer Graphics, 3 rd ed. (Shirley & Marschner ) Slides by Marc van Kreveld. Interpolation vs. approximation.
E N D
Splines III – Bézier Curves based on: Michael Gleicher: Curves, chapter 15 in Fundamentals of Computer Graphics, 3rd ed. (Shirley & Marschner) Slides by Marc van Kreveld
Interpolation vs. approximation • Interpolation means passing through given points, approximation means getting “close” to given points • Bézier curves and B-spline curves p2 p2 p1 p1 p0 p3 p0 p3 p1and p2are interpolated p1and p2are approximated
Bezier curves • Polynomial of any degree • A degree-d Bezier curve has d+1 control points • It passes through the first and last control point, and approximates the d – 1 other control points • Cubic (degree-3) Bezier curves are most common; several of these are connected into one curve
Bezier curves • Cubic Bezier curves are used for font definitions • They are also used in Adobe Illustrator and many other illustration/drawing programs
Bezier curves • Parameter u, first control point p0 at u=0 and last control point pd at u=1 • Derivative at p0 is the vector p0p1 , scaled by d • Derivative at pd is the vector pd-1pd , scaled byd • Second, third, …, derivatives at p0 depend on the first three, four, …, control points
Cubic Bezier curve example 3 p1 p2 p0 p2p3 p2p3 p0p1 p0p1 p3 3
Quintic Bezier curve example 5 p2 p3 p1 p4 p5 p4p5 p4p5 p0p1 p0p1 p0 5
Cubic Bezier curves • p0 = f(0) = a0 + 0 a1 + 02a2 + 03a3p3 = f(1) = a0 + 1 a1 + 12 a2 + 13a33(p1 – p0) = f’(0) = a1+ 20 a2 + 302a33(p3 – p2) = f’(1) = a1 + 21 a2 + 312a3 basis matrix
Cubic Bezier curves • f(u) = (1 – 3u + 3u2 – u3) p0 + ( 3u – 6u2 + 3u3) p1 + ( 3u2 – 3u3) p2 + ( u3) p3 • Bezier blending functionsb0,3= (1 – u)3 b1,3 = 3 u (1 – u)2 b2,3 = 3 u2 (1 – u) b3,3 = u3
Bezier curves • In general (degree d):bk,d(u) = C(d,k) uk(1 – u)d-kwhere , for 0 k d(binomial coefficients) • The bk,d(u) are called Bernstein basis polynomials
Bezier curves degrees 2 (left) up to 6 (right)
Properties of Bezier curves • The Bezier curve is bounded by the convex hull of the control points intersection tests with a Bezier curve can be avoided if there is no intersection with the convex hull of the control points
Properties of Bezier curves • Any line intersects the Bezier curve at most as often as that line intersects the polygonal lie through the control points (variation diminishing property)
Properties of Bezier curves • A Bezier curve is symmetric: reversing the control points yields the same curve, parameterized in reverse p2 p3 p1 p4 p3 p2 p5 p0 p4 p1 p0 p5
Properties of Bezier curves • A Bezier curve is affine invariant: the Bezier curve of the control points after an affine transformation is the same as the affine transformation applied to the Bezier curve itself (affine transformations: translation, rotation, scaling, skewing/shearing) p2 p2 p3 p3 p1 p1 p4 p4 p5 p5 p0 p0
Properties of Bezier curves • There are simple algorithms for Bezier curves • evaluating • subdividing a Bezier curve into two Bezier curves allows computing (approximating) intersections of Bezier curves the point at parameter value u on the Bezier curve
Bezier curves in PowerPoint • The curve you draw in PowerPoint is a Bezier curve; however you don’t specify the intermediate two control points explicitly • Select draw curve • Draw a line segment (p0 and p3) • Right-click; edit points • Click on first endpoint and move the appearing marker (p1) • Click on last endpoint and move the appearing marker (p2)
Splines from Bezier curves • To ensure continuity • C0 : last control point of first piece must be same as first control point of second piece • G1 : last two control points of first piece must align with the first two control points of the second piece • C1 : distances must be the same as well q2 p2 p1 p3 q3 q0 q1 p0
Intuition for Bezier curves • Keep on cutting corners to make a “smoother” curve • In the limit, the curve becomes smooth p1 p0 p2
Intuition for Bezier curves • Suppose we have three control points p0 ,p1, p2;a linear connection gives two edges • Take the middle p3 of p0p1, and the middle p4 of p1p2 and place p’1 in the middle of p3 and p4 • Recurse on p0, p3, p’1 and also on p’1 , p4, p2 p1 p3 p4 p’1 p’1 p0 p0 p0 p2 p2 p2 gives a quadratic Bezier curve
De Casteljau algorithm • Generalization of the subdivision scheme just presented; it works for any degree • Given points p0, p1, …, pd • Choose the value of u where you want to evaluate • Determine the u-interpolation for p0p1, for p1p2, … , and for pd-1pd , giving d – 1 points • If one point remains, we found f(u), otherwise repeat the previous step with these d – 1 points
De Casteljau algorithm u = 1/3 p2 p1 p3 p0
De Casteljau algorithm u = 1/3 p2 p1 p3 p0
De Casteljau algorithm u = 1/3 p2 p1 p3 p0
De Casteljau algorithm u = 1/3 p2 p1 p3 p0 one point remains, the point on the curve at u = 1/3
Splitting a Bezier curve • The De Casteljau algorithm can be used to split a Bezier curve into two Bezier curves that together are the original Bezier curve p2 p1 r1 r2 r0 q3 q2 p3 q1 r3 p0 q0
Splitting a Bezier curve p2 p1 p3 r1 p0 r2 r0 q3 q2 Question: Recalling that Bezier splines are C1 only if (in this case) the vector q2q3 is the same as r0r1 , does this mean that the spline is no longer C1 after splitting?!? q1 r3 q0
Splitting a Bezier curve p2 p1 p3 r1 p0 r2 r0 q3 q2 Answer: q0q1q2q3 parameterizes the part u [0, 1/3] and r0r1r2r3 parameterizes the part u [1/3, 1] The condition for C1 continuity, q2q3 = r0r1 , applies only for equal parameter-length parameterizations q1 r3 q0
Splitting a Bezier curve • Splitting a Bezier curve is useful to find line-Bezier or Bezier-Bezier intersections p2 p1 u = ½ p3 p0
Intersecting a Bezier curve • To test if some line L intersects a Bezier curve with control points p0, p1, …, pd , test whether L intersects the poly-line p0, p1, …, pd • If not, L does not intersect the Bezier curve either • Otherwise, split the Bezier curve (with u = ½ ) andrepeat on the two pieces p2 p1 p3 p0
Intersecting a Bezier curve • If the line L separates the two endpoints of a Bezier curve, then they intersect • Repeating the split happens often only if the line L is nearly tangent to the Bezier curve p2 p1 p3 p0
Intersecting a Bezier curve • If the line L separates the two endpoints of a Bezier curve, then they intersect • Repeating the split happens often only if the line L is nearly tangent to the Bezier curve p2 p1 p3 p0
Intersecting a Bezier curve • When determining intersection of a line segment and a Bezier curve we must make some small changes p2 p1 p3 p0
Splitting a Bezier curve for rendering • Splitting a Bezier curve several times makes the new Bezier curve pieces be closer and closer to their control polygons • At some moment we can draw the sequence of control polygons of the pieces and these will approximate the Bezier curve well (technically this approximation is only C0)
Splitting a Bezier curve for rendering p2 p1 u = ½ p3 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 p3 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 p3 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 p3 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 p3 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 p3 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 u = 1/2 p3 u = 3/4 u = 1/4 p0
Splitting a Bezier curve for rendering p0 p3 p1 p2 p2 p1 u = 1/2 p3 u = 3/4 u = 1/4 p0
Splitting a Bezier curve for rendering p0 p3 u = 1/2 p3 u = 3/4 u = 1/4 p0
Splitting a Bezier curve for rendering p3 p0 p3 p0 p0 p3 p3 p0
3D Bezier surfaces The 16 blending functions for cubic Bezier surfaces
Summary • Bezier curves are elegant curves that pass through the start and end points and approximate the points in between • They exist of any order (degree) but cubic is most common and useful • Continuity between consecutive curves can be ensured • The De Casteljau algorithm is a simple way to evaluate or split a Bezier curve
Questions • Consider figure 15.11, bottom left. It looks like a circular arc, but it is not. Determine whether the quadratic Bezier curve shown here goes around the midpoint of a circular arc or not • Can we ensure higher degrees of continuity than C1 with cubic Bezier splines? Discuss your answer • Suppose we want to define a closed Bezier curve of degree d. What properties must the control points have to make a C1 continuous curve? What is the minimum degree of the Bezier curve that is needed for this? What if we want a closed Bezier curve with an inflection point (boundary of a non-convex region)?
Questions • Apply the De Casteljau algorithm on the points (0,0), (4,0), (6,2), and (4,6) with u = ½ by drawing the construction (note that this is a cubic Bezier curve) • Apply the De Casteljau algorithm on the points (0,0), (4,0), (6,2), (6,8), and (10,4) with u = ½ by drawing the construction (note that this is a quartic Bezier curve)