A new thresholding function which is used in image denoising is proposed in this paper to solve the problems existed in the traditional wavelet thresholding functions. Compared with the traditional hardthresholding function and softthresholding function, the new thresholding function overcomes the shortcoming of uncontinuity of hardthresholding function and remains the continuity of softthresholding function as well as decreases the invariable dispersion in the soft thresholding function. The new thresholding function is simple in expression and has a high order derivative, it is suitable to do some kinds of mathematical disposals. The simulation results for image show that this method has better denoising effect and higher practical value than the traditional hard and softthresholding methods and soft and hard thresholding compromised methods.