In this paper，a fast algorithm for the fractal signal wavelet decomposition and reconstruction is put foreward. In accordance with the selfsimilar and long-term related characteristics of the fractal signals，and by means of discrete wavelet transformation (DWT)，multi-scale decomposition is carried out so as to make them become similar stationary signals and estimate them with the usual wiener filtering and Kalman filtering method，Then multi-scale reconstruction is carried out with DWT in order to estimate the primary signals polluted by noises. This paper stresses the algorithm design of the DWT filtering process，and the computing complexity is also considered.