In order to solve the noise pollution problems of the complex-background image in the transmission process, a new threshold noise reduction algorithm is put forward. Several methods of image de-noising had been analyzed based on the wavelet transform; wavelet threshold function prototype had been studied; a new improved wavelet threshold de-noising method was presented based on the traditional soft-threshold and hard-threshold de-noising method. This method can adjust the threshold function flexibly according to the actual environment and ameliorate the "over-kill" or "over-protection" phenomenon obviously while the images are reconstructed. In order to explain the noise reduction effect of the new method accurately, this paper put different images as the original images, put the salt & pepper and Gaussian noise as the noise source, and then the results of noise reduction with the new threshold function de-noising method and the others de-noising methods are analyzed, the results show that the new threshold method is superior to the other methods for noise reduction based on both the human eye visual standard and PSNR evaluation standard, verify the validity of the de-noising method.