Embedded applications use more and more sophisticated computatious. These computations can integrate composition of elementary functions and can easily be approximated by polynomials. Indeed, polynomial approximation methods allow to find a trade-off between accuracy and computation time. Software implementation of polynomial approximation in fixed-point processors is considered in this paper. To obtain a moderate approximation error, segmentation of the interval I on which the function is computed is necessary. This paper proposes a new method to compute the values of a function on I using non-uniform segmentation and polynomial approximation. Non-uniform segmentation allows one to minimize the number of segments created and is modeled by a tree-structure. The specifications of the segmentation set the balance between memory space requirement and computation time. Besides, compared to table-based methods or the CORDIC algorithm, our approach significantly reduces, the memory size and the function evaluation time respectively.