When the transcendental function is used in the calculation of the piece-wise linear approximation algorithm, the accuracy can not be determined in advance and part of interval resources are wasted. In order to solve the problems, an improved piece-wise linear approximation algorithm of transcendental function was proposed. The proposed algorithm calculated a linear approximation function with predefined interval endpoints and adjusted the function by the concave of the function to be approximated and then calculated the maximum error between the function adjusted and the function to be approximated. The approximation intervals would be adjusted automatically based on the precision requirement and a fewer number of segments would be obtained by such iterative process. The proposed algorithm was simulated in Matlab. The simulation results show that the number of segments of the proposed algorithm is reduced by 60% compared with the equal division method. Thus, the proposed algorithm can reduce the resource consumption of the Look Up Table (LUT) greatly on the premise of guaranteeing accuracy.