Abstarct: Seismic imaging is highly dependent on the quality of seismic data. The presence of noise in seismology is inevitable that reduces continuity of the reflection events and effectiveness of the various stages of data processing such as deconvolution, velocity analysis and migration has negative effects. There are many noises in seismic data that are divided into two general categories of coherent and incoherent. Major categories of noise in seismic data, are coherent noises that are coherent from trace to trace and always decrease signal to noise ratio together with the main events and many problems in the processing and interpretation of seismic data will be created. The main objective of this thesis undermine this category is the noise of seismic data. Many different methods have been introduced to suppression of coherent noises. Most methods have been introduced, undermining the process in a domain other than the domain of the time-offset. Including converting that to change the domain of seismic data used to this purpose, is two-dimensional discrete wavelet transform and radial traces transform that individually are a number of disadvantages. In this thesis, has been used a new method baxsed on joint application of the two-dimensional discrete wavelet and radial trace transforms in order to eliminate the coherent noise in seismic data. Process is that first seismic data are taken from time-offset domain into the radial trace domain using the radial trace transform, ground roll noises are stretched in this domain and the apparent frequency of these events in the new domain because the strain is reduced. However, the frequency content of other events such as reflections in the new domain remain largely untapped. This feature radial trace transform as a good tool, improved performance of two-dimensional discrete wavelet transform and limitations the wavelet transform greatly reduced. With two-dimensional discrete wavelet transform is applied in radial trace domain, the seismic signal is decomposed into four components and ground roll noise mostly for almost vertical seismic placed on record, The vertical component of the discrete wavelet transform appears that removing this component and return to the domain of time- offset, ground roll noise is attenuate from the seismic record. Another new method in this thesis to undermine coherent noise is used, is the use curvelet transform in radial trace domain. One of the weaknesses of the wavelet transform is the inability to rebuild the edges. Curvelet transform fix these weaknesses, and in other words, improves of the edge quality after the reconstruction. Wavelets are not able to describe the anisotropic components and therefore, will not be able to describe the edges, because this tool has not been directional elements. While curvelets tools multi-scale and are able to choose the direction. Curvelet transform is divided the frequency domain into scale, directions and different positions. Thus, according to the characteristics of the ground roll waves of low frequency, low speed and high amplitude compared to be reflective of relatively good separation between these waves can be produced in the curvelet domain. So, remove the curvelet coefficients related to the scales and directions that ground roll waves are present and back in time, can be found the better sections for the interpretation. The method algorithm is such as two-dimensional discrete wavelet transform method with the difference that converts two-dimensional discrete wavelet transform is replaced curvelet transform. These methods were applied on real seismic data, and results were examined and in order to more determine, the results were compared with f-k speed filter. The results indicate that the method baxsed on joint application of the two-dimensional discrete wavelet and radial trace transform compared of the f-k filtering has better performance in improve the quality of seismic data. On the other hand, from between the algorithms used in radial trace domain, the curvelet transform method compared of two-dimensional wavelet transform in attenuation ground roll coherent noise been more successful.