Abstarct: Enhancement reflective signals by attention unwanted energy which called noise is the one of main goals in reflective seismic data processing. Generally, noise in seismic data can be divided into two categories of coherent and incoherent or random noise. Incoherent noise, known as random noise, is one of the most important categories of noise which is not correlated from trace to trace and is unlike the coherent type of noise and characterized by random oscillation in seismic data over the time, So that suppression of this wide-band noise is one of the challenging issues in seismic data processing. Often random noise has white frequency spectrum and includes all of frequency. However, signal and noise have the overlap in frequency domain. Inherently, earth is non-elastic and acts as a low pass filter and changes the frequency content of seismic signals with time. So that seismic signals are non-stationary and for attention random noise in seismic data it is better to using the Time-Frequency transforms and methods baxsed on them. In this thesis, at first, Synchrosqueezing wavelet transform introduce as a new version of the wavelet transform which incorporated features of empirical mode decomposition and frequency reassignment methods. It provides a high-resolution time-frequency representation allowing the identification of instantaneous frequencies in seismic signals to highlight individual components. The seismic data are transformed into sparse subspace using the synchrosqueezed wavelet transform, then the obtained sparse time-frequency representation is decomposed into semi-low-rank and sparse components using the Optshrink algorithm because sparse transform baxsed denoising assumes that the seismic data are sparse in a certain transformed domain and the performance of proposed method compared with the Semi-Soft GoDec algorithm, classical f − x SSA, and prediction Wiener filter. In the following of thesis, baxsed on decomposition time-frequency representation a new noise suppression algorithm (ISLR) will propose for seismic data demising and a comparison is performed between the proposed method and the WSST-GoDec algorithm and classical f-x SSA. The results in both method visually and quantitatively confirmed the superiority of the proposed methods in contrast to the other well-established noise reduction methods.