In this paper, some conventional methods used to reduce random noise are reviewed. These methods are as follows: Wavelet transform and curvelet transform which by applying these transforms on the signal, we can determine the various components of the frequency of the signal and the time of their occurrence with proper accuracy. Empirical mode decomposition which in this method, the noisy signal is decomposed into components called intrinsic mode functions (IMFs), and those IMFs that represent the random noise are eliminated and the signal is reconstructed using IMFs. Singular value decomposition which in this method, seismic data are considered as a matrix, and then this matrix is written as a sum of a number of matrices, which are called eigenimages. Each eigenimage represents part of the seismic data, we can process seismic data by removing eigenimages related to random noise gathering other eigenimages. FX deconvolution filter, the data will be denoised according to the signal as a predictable function of offset. The ability of different methods for seismic images quality improvement will investigate and compare.