Abstarct: Attenuation of noise from seismic data is one of the important steps in processing, which is effective in better interpretation and accuracy of the obtained results. It causes access to the results with higher and more accurate confidence. In land and sea seismic data, there are different types of noises. One of them in the marine seismic data is the Swell noise, which has a high amplitude and low frequency and is observed in the form of vertical bands in the marine seismic data. It seems necessary to develop methods that can optimally attenuate the Swell noise along with the least amount of damage to the signal in seismic data. Various methods have been used to attenuate Swell noise, such as FX filter, median filter, time-frequency filter, Wiener predictive filter and bandpass filter, thresholding, and an approach equipped with a deep learning model. In this thesis, the method of mode decomposition is used to filter the noise. Mode decomposition methods include empirical mode decomposition and variational mode decomposition. Due to the fact that these methods are able to decompose the signal into its constituent components, it enables any filtering process to be performed on the constituent modes. Due to the nature of Swell noise, which includes a specific frequency band, this noise will be visible on one or more modes and according to this issue, it can be filtered using mode decomposition methods. In this thesis, the performance of empirical and variational mode decomposition methods in filtering Swell noise on synthetic and real data has been investigated, and acceptable results have been obtained.