ON THE ONE METHOD OF A THIRD-DEGREE BEZIER TYPE SPLINE CURVE CONSTRUCTION Stelia O., Potapenko L., Sirenko I. Faculty of Computer Sciences and Cybernetics, Taras Shevchenko National University of Kyiv, Kyiv, Ukraine [email protected], [email protected], [email protected] Abstract. A method is proposed for constructing a spline curve of the Bezier type, which is continuous along with its first derivative by a piecewise polyno- mial function. Conditions for its existence and uniqueness are given. The con- structed curve lies inside the convex hull of the control points, and the segments of the broken line connecting the control points are tangent to the curve. To construct the curve, we use the approach proposed earlier for constructing a parabolic spline. The idea is to use additional points with unknown values of some function. Additional points are used as spline nodes, and the function val- ues are determined from the condition of the first derivative continuity of a piecewise polynomial curve. In multiple interpolation nodes, the function takes the given values and the values of the first derivative, which are determined by the control points. Examples of constructing a spline curve are given. Keywords: THIRD-DEGREE SPLINE, BEZIER CURVE. 1 Introduction The history of constructing and using smooth curves is quite old. The algorithms for constructing piecewise polynomial smooth curves of low degrees received rapid de- velopment with the advent and development of computer technology. On the one hand, computers can significantly accelerate the construction of such curves for vari- ous technical applications, and on the other hand the computers themselves are "con- sumers" of various algorithms in computer graphics systems [1, 2]. The most fre- quently Bezier curves, spline curves and B-splines are used for these purposes. Curves subsequently called Bezier curves [3, 4, 5, 6] were developed in the 1960s independently by Pierre Bezier of the Renault car company and Paul de Castillo of Citroën, where they used for the design of car bodies. Bezier curves are used in com- puter graphics to draw smooth bends, animations and other applications. The wide application of Bezier curves for approximation problems is related with their convenience as an analytical description and a visual geometric construction. With respect to the computer graphics, this means that the user can set the shape of the curve interactively, i. e. moving the curve-defining points on the screen. Also Bezier curves can be used in cartography when it is required to build a route on the map with given initial and end points and which passes near some points 2 In work [7], an algorithm is given for calculating the coefficients of a parametric spline, which approximates the sequence of experimental data. The Bezier curves were used as the approximating spline. B-splines [4, 5, 8] is a generalization of Bezier curves. Modeling on the basis of inhomogeneous rational B-splines has the advantage over other methods. With the help of such splines it is easier to simulate the surfaces of natural objects, as well as surfaces that have complexly curved profiles. The models of such splines provide the best quality of the rounded edges of objects visualization [8]. In the book [9] an algorithmic processing of Bézier curves and spline curves is considered. In this publication, we propose and justify an algorithm for constructing a curve that has a number of properties of both Bezier curves and spline curves. In connection with this, the resulting curve can be called a third-order spline curve of Bezier type. 2 A constructive method for proving the existence and uniqueness of a spline curve We consider a certain interval [a,b] , on which we define the grid :a 1 2 ... N b . At the nodes i we set the values Fi . To construct the curve, we use the approach proposed in [10] for parabolic splines. Along with the grid , we introduce a grid x :1 x1 x2 ... xN1 N , where i1 xi i . The values of a certain function at the points xi will be denoted by fi . We introduce the notation hi i i1 , i i xi Then xi i1 hi i . Taking these notations into account, we get: fi Fi1(hi i ) Fi i / hi , fi (Fi Fi1) / hi . We will construct a spline curve of the third order, for which the points i will be nodes of the spline, and the points xi will be multiples of the interpolation nodes. To construct the curve, we construct a cubic spline of defect 2 on the interval [ a,b ], satisfying the following conditions: 3 S (xi ) fi , (1) 3 (S )(xi ) fi , i 2, N . (2) We denote , i 1, N the unknown values of the function S 3 (x) at the nodes i i of the spline, 3 1 2 i Theorem. Under the conditions i , wherei , the interpolation cubic 3 3 hi 3 spline of defect 2 S (x) for grids x , on the interval [ a,b ], satisfying conditions (1) - (2), there exists and unique one. Proof. To construct a spline, we write the third-order interpolation Hermite poly- nomial [11] on each of the intervals [ i1, i ], i 2, N 1. 3 (x xi )(x i ) (x xi )(x i1) S (x) i1 i ( i1 xi )( i1 i ) ( i xi )( i i1) (3) (x i )(x i1) fi Q1(x i )(x i1)(x xi ) (xi i )(xi i1) for x [ i1, i ] , 3 (x xi1)(x i1) (x xi1)(x i ) S (x) i i1 ( i xi1)( i i1) ( i1 xi1)( i1 i ) (4) (x i )(x i1) fi1 Q2 (x i )(x i1)(x xi1) (xi1 i )(xi1 i1) for x [ i , i1] . It is obvious from the relations (3) - (4) that the condition (1) is satisfied. To determine the quantities Q1 and Q2 we use condition (2), i.e. the following rela- tions must be satisfied: 3 3 (S )(xi ) fi , (S )(xi1 ) fi1 . 3 We consider (S )(x) for x [ i1, i ] 3 (x xi ) (x i ) (x xi ) (x i1) (S )(x) i1 i ( i1 xi )( i1 i ) ( i xi )( i i1) (5) (x i ) (x i1) fi Q1(x xi )(x i ) (x i1) (x i )(x i1) . (xi i )(xi i1) 3 Similarly we consider (S )(x) for x [ i , i1] (x x ) (x ) (x x ) (x ) (S 3 )(x) i1 i1 i1 i i ( x )( ) i1 ( x )( ) i i1 i i1 i1 i1 i1 i (6) (x i ) (x i1) fi1 Q2 (x xi1 )(x i ) (x i1 ) (x i )(x i1 ) . (xi1 i )(xi1 i1 ) 4 3 xi i xi i1 (S )(xi ) i1 i ( i1 xi )( i1 i ) ( i xi )( i i1) (xi i ) (xi i1) Fi Fi1 fi Q1(x i )(x i1) , (xi i )(xi i1) hi x x (S 3 )(x ) i1 i1 i1 i i1 i ( x )( ) i1 ( x )( ) i i1 i i1 i1 i1 i1 i (7) (xi1 i ) (xi1 i1) Fi1 Fi fi1 Q2 (xi1 i )(xi1 i1) . (xi1 i )(xi1 i1) hi1 Using the above notation for the steps hi and i , we obtain: i hi i hi 2i Fi Fi1 i1 i fi Q1(hi i )i , (8) (hi i )hi hi i (hi i )i hi wherefrom 1 1 hi 2i Fi Fi1 Q1 i1 2 i 2 fi 2 2 . (9) (hi i ) hi hi i (hi i ) i hi i (hi i ) Similarly i1 hi1 i1 hi1 2i1 i I 1 fi1 (hi1 i1)hi1 hi1i1 (hi1 i1)i1 (10) Fi1 Fi Q2 (hi1 i1)i1 , hi1 wherefrom we get 1 1 Q2 i 2 i1 2 (hi1 i1) hi1 hi1i1 (11) hi1 2i1 Fi1 Fi fi1 2 2 . (hi1 i1) i1 hi1i1(hi1 i1) To ensure the smoothness of the resulting curve, that is, the continuity of the first 3 3 derivative, we require the fulfillment of the relation (S )( i 0) (S )( i 0) , where 3 i hi i hi (S )( i 0) i1 i fi Q1i hi , (12) (hi i )hi hi i (hi i )i 5 2h h (S 3 )( 0) i1 i1 i1 i1 i i (h )h I 1 h i1 i1 i1 i1 i1 (13) hi1 f1I Q2 (hi1 i1)hi1 . (hi1 i1)i1 Equating expressions (12) and (13) and substituting the values Q1 and Q2 from (9) and (11), we obtain a system of linear algebraic equations with a tridiagonal matrix for determiningi : i hi i hi i i i1 i fi i1 2 (hi i )hi hi i (hi i )i (hi i ) i (hi 2i )hi Fi Fi1 2hi1 i1 hi1 i1 fi 2 i i1 (hi i ) i (hi i ) (hi1 i1)hi1 hi1i1 (14) hi1 i (hi1 i1) fi1 i1 2 (hi1 i1)i1 (hi1 i1) i1 (hi1 2i1)hi1 Fi1 Fi fi1 2 . (hi1 i1)i1 i1 We write expression (14) in the following form: A (B B ) C , (15) i 1i 1 2 ii i1i i where 2 2 i 2hi i 3hi1 i1 (hi1 i1) A 2 , B1 , B2 , C 2 , (16) (hi i ) hi ihi (hi1 i1 )hi1 hi1i1 h (h 2 )h F F h f i f i i i i i1 f i1 i i (h ) i (h )2 (h ) i1 (h ) i i i i i i i i i1 i1 i1 (17) (hi1 2i1)hi1 Fi1 Fi fi1 2 . (hi1 i1)i1 i1 To close up the system of equations, we add the conditions: 1 f1 , N f N . (18) Obviously, the values A, B1, B2 ,C are positive and 0 i 1 . If B1 A and B2 C , then system (15) will have a diagonal predominance.