Application Of Anisotropic Diffusion In Edge-Preserving Smoothing For Seismic Image

The quality of seismic data is the basis of structural interpretation and lithologic interpretation. Although the conventional means of improving the quality of seismic data can improve signal-to-noise ratio, they will reduce the horizontal resolution of the seismic data, for example, they smear faults and stratigraphic edge. However, all these information above is very important for the structural interpretation, stratigraphy and sedimentary interpretation, and they are the basis of reservoir discovery and description. Therefore, we need a class of filters that improves the quality of seismic data, as well as effectively preserving the faults and the stratigraphic edge.These filters therefore require three ingredients: (1) orientation analysis, (2) edge detection, (3) edge-preserving oriented smoothing. The gist of these filters is a smoothing operation parallel to the seismic re?ections that does not operate beyond re?ection terminations (faults). In other word, once the dip/azimuth has been estimated, a filter can enhance signal along the reflector. Most important, no smoothing takes place if a discontinuity is detected, thereby preserving the appearance of major faults and stratigraphic edges.Anisotropic diffusion filtering is just to meet the requirements, because it combines the smoothing and edge detection. In an anisotropic diffusion process, the input (seismic) image serves as the initial condition, and the processed image is given by solving a partial differential equation through time. In the diffusion equation, "structure tensor" is introduced to extract local structure information. Based on the local structural information, "diffusion tensor" is designed. And diffusion tensor is used to control the diffusion process, making the smoothing take place parallel to the reflector, but no smoothing take place perpendicular to the reflector. As a result, the anisotropic diffusion filtering improves signal-to-noise ratio of seismic data as well as enhance the continuity of reflection events and preserve the important geological structure.In this paper, we present anisotropic diffusion pioneered by Weickert and use this edge-preserving filtering for the seismic image to improve the quality of seismic data. Based on the basic principles, the numerical discretization, parameter adjustment and iteration stopping criteria are discussed. By selecting the appropriate method and adjusting every parameter, we try to achieve optimal filtering.In interpretation software GeoScope, we develop a tool according the algorithm of anisotropic diffusion using the C++ programming. We use this tool to process seismic data in some projects. The results show the remarkable effects of this filter, and it is very helpful for fault interpretation and sedimentary interpretation.