Detection And Removal Of Linear Interference In The Radial Trace Of Pre-stack Seismic Data With Ridgelet Transform

Denoising of seismic data, especially prestack seismic data, is a critical part of seismic data interpolation, for it is a single-shot record which will easily suffer regular and irregular interference. There are many mature ways to deal with irregular interference; however, it is relatively difficult to remove regular interference. This thesis discusses the removal of linear noise (regular interference) with ridgelet in the radial trace domain. The traditional way of removing linear noise may cause loss of data because of the interpolation in the radial trace, but the radial trace prediction can not accurately determine the direction of the radial trace. Therefore, with the thought to accurately detect linear noise with ridgelet, this thesis studies the detection of linear noise in seismic data and the removal of it with radial trace prediction technology.The second chapter of this thesis gives a brief introduction to the development and theory of ridgelet and the current research status of ridgelet.Aiming at the detection of seismic image which contains linear noise with ridgelet, chapter three introduces the problems encountered in the detecting process of ordinary linear and line segment with ridgelet. The ridgelet can only determine the direction and position of the line segment, while it is hard to ascertain the length of it; therefore, with the method of image block and then ridgelet transformation, this thesis accurately detects the length of line segment. The "linear" noise in the practical detection of linear interference of seismic data does not mean "straight line" in the tradition sense. It is composed by several "straight lines" of the similar rate of slope. The thought of image block can not only determine the length of the linear interference, but also can turn the thought of processing linear interference into processing the linear interference domain. Based on the results obtained in chapter three, chapter four makes a radial trace transformation and prediction to remove the linear noise in this field.Chapter four mainly introduces the theoretical algorithm and the present developing direction of radial trace transformation. Based on the results gained in chapter three, interpolation techniques will be adopted because non-integer sampling will be encountered during the radial trace transformation process. However, owing to the loss of seismic data resulted from interpolation techniques, which will lead to the unsatisfactory treatment effect, this thesis employs radial trace prediction instead of radial trace transformation. Within the local domain, and along the direction of coherent interference, the energy of coherent interference between traces is nearly the same. And differences caused by signals will emerge due to the superimposition of the coherent noise and the factual signals in the direction of coherent interference. If the coherent interference in this direction is deemed as random signals, the equalizing value of the signals transmitted will be zero. Thus the equalizing value of the linear interference in a certain direction can be worked out. And then the effective signals may be calculated by subtracting the equalizing mean from the primary record. This thesis gets a satisfactory denoising effective through this method and will be further proved in practical seismic data.