Its multiplicative nature complicates speckle noise reduction in images because of the effort required for separation of noisy pixels from other pixels. In this study, a novel adaptive anisotropic diffusion filter algorithm based on Haar wavelet transform has been proposed. Initially, Haar transform of image to be filtered was taken and then median absolute deviation of wavelet coefficients was used to tune the conductance parameter, K of diffusion filter with different diffusion functions. The suggested strategy has been tested with different images and different noise variances. Moreover, experimental results have been compared with conventional diffusion filters, and also Lee filter and Wiener filter which are frequently used for despeckling.