Abstarct: The presence of noise in the reflected seismic data causes data degradation and as a result causes inaccuracy in subsurface modeling. An important group of unwanted and destructive factors in seismic data are random noises. Random noises appear at any time and frequency and have no order. Due to the destructive role of random noises, their destructive effects were cleared from the data with the help of adaptive-noise multi-scale diffusion filter. Anisotropic filters using diffusion tensor can be smoothed to maintain structures. An anisotropic adaptive-noise diffusion filter uses a Gaussian filter in its algorithm to create windows with specific dimensions, examines the data quality and the amount of noise impregnated with it, and according to the data conditions, determines the best diffusion function to perform defogging. Diffusion and noise cancellation are done with the help of this function. By estimating the diffusion function, other parameters are automatically selected by the filter. In this dissertation, in order to better understand and the effect of each parameter on the results, a different range of numbers for these parameters were tested and evaluated and the amount of noise reduction with different values of these parameters were compared with each other. In order to achieve the desired results, by applying the wavelet transform algorithm, the mentioned anisotropic diffusion filter can be multidimensional. The adaptive-noise multi-scale diffusion filter first breaks down the data into different components at different scales by means of a wavelet transform, and then attenuates random noises by applying a adaptive-noise diffusion filter to each of these scales. In this dissertation, at first, the adaptive-noise anisotropic emission filter was applied to artificial and real data, and contrary to predictions, the desired results were not recorded, and this filter failed to improve the results compared to other emission filters. Then, in order to achieve better results, a filter with a multi-scale approach was applied to artificial and natural data. The use of this filter, in comparison with other methods of attenuation of noise with the help of diffusion filter, caused a significant increase in the signal-to-noise ratio and preserved the edges and borders of structures.