부산대학교 도서관

Gaussian function has been frequently used in image and signal processing, especially on image denoising and edge detection. In this paper, we present an approximation model of the Gaussian function using the powers of a half-period of a sine wave. In the previous research, there is only exponential formulation of the Gaussian kernel with two parameters mean µ and standard deviation σ. The proposed model is based on harmonic decomposition of powers of sine with a preliminary accuracy better than 10 −7 with two parameters n and f 0 . Moreover, the complexity of implementation of the proposed Gaussian model is reduced with more efficiency. On the other hand, we introduce a novel formulation of the 2D Gaussian kernel based only on sine wave for image processing applications. The performance and robustness of the suggested Euler Harmonic Gaussian model are illustrated on image filtering and edge detection applications compared to the existing algorithms.