Numerical Algorithms Прикладная математика
Implementation of divided differences to shape preservation
Аннотация
Abstract Divided difference methods are used in many engineering and scientific applications to solve numerical interpolation problems. Since shape-preserving approximation has many applications in computer-based geometric design, image processing, geodesy, chemistry, and robotics, we aim to implement divided difference methods to study the shape-preserving properties of a class of blending-type operators, which are constructed via certain shape parameters and represented in terms of divided differences. Considering this new representation, we establish the shape-preserving properties of operators, such as linearity, positivity, end-point interpolation and specifically monotonicity and convexity-preserving properties in connection with a function on the interval $$\varvec{\left[ 0,1\right] }$$ 0 , 1 . Our theoretical, computational, and numerical results reveal that while operators preserve monotonicity entirely on the interval $$\varvec{\left[ 0,1\right] }$$ 0 , 1 for the defined intervals of shape parameters, they are unable to preserve the convexity of functions for some values of a shape parameter if the function has a monotonic behavior. Therefore, we give a modified convexity preservation result for the operators by imposing auxiliary assumptions on a function on the interval $$\varvec{\left[ 0,1\right] }$$ 0 , 1 . We conclude our analysis by providing a comparison of the convexity preservation abilities of operators equipped with the shape parameter $$\varvec{\lambda }$$ λ .