Digital images can be used in many aspects of life. The acces and transfer process from one media digital image to another media is often perfomed. The processes, however, often experience distortions which cause noises in the digital images, so that the quality of images received by someone will be less or not as good as the original. Maximum Aposteriori-Gaussian Scale Mixtures is a technique for filtering digital images in order to subtract or decrease the noise level, so mthat the quality can be improved. Noise used in this research is the additive Gaussian noise, impulsive noise and Laplacian noise with a dixed probability, generated by a noise generator. The digital image will be decomposed into 4 subband (LL, LH, HL, HH) first an the subband that will be prosessed are LH, HL, and HH. The process is conducted using GSM method with maximum aposteriori for estimating the multiplier and local wiener for estimating the central coefficient. Performance parameter tested is PSNR (Peak Signal-to-Noise Ratio). The analysis result asserts that Maximum Aposteriori-Gaussian Scale Mixture method is a proper method for decreasing additive gaussian noise as well as Laplacian, impulsive and multiplicative Gaussian noise. To get better PSNR, a bigger size of MAP size can be used. Keywords: Gaussian Scale Mixture (GSM), Maximum Aposteriori (MAP), Noise, Discrete Wavelet Transform (DWT), PSNR