基于在均匀介质中的波射线传播理论，应用直射线追踪方法进行了射线走时正演计算，并选用了SIRT算法根据不同的点源和不同的检波点的走时观测值进行快速反演成像，并对不同的反演结果进行比较，获得了在不同走时观测值所反演得到速度异常区域的位置。

介绍了基于神经网络的多属性分析在地震图像的共同区域划分中的应用。首先优选并计算了能刻画地震图像复杂度的4种属性:纹理能量、纹理对比度、纹理随机性和纹理分形维属性,然后采用神经网络对具有统计学特征的纹理属性进行聚类分析,从而得到地震图像共同区域划分图。

Noise muffling of seismic records is a significant step in seismic data processing and it is of great importance in processing and interpreting seismic data subsequently.Based on the analysis of principles of a traditional total variation image denoising model,an improved total variation approach to image denoising with diffusion tensor appended was proposed.Through enhancing the anisotropic filtering of diffusion terms along the orientation of stratigraphic textures in the total variation model,this approach could consistently strengthen the formation continuity along the orientation and control the diffusion state perpendicular to stratigraphic bedding so as to preserve the edge structure information.The actual result of this approach in application to some seismic data from western Sichuan showed that it could effectively remove the noise in seismic images,enhance the continuity of seismic events and increase the signal to noise ratio of seismic data.