Caculation Of Local Slop On Seismic Data Via Multi-scale Principal Component Analysis

Seismic exploration is one of important methods of geophysical research. The traditional analysis of seismic data is to construct models based on accurate speed information, acquiring which, however, costs too much time. Meanwhile accurate speed information involves several unstable parameters leading to divergence of results. The local slope contains almost the whole information of the data and is easier to obtain. Thus the method based on the local slope has played a more and more important role in the seismic exploration.This paper proposed a novel method of the calculation of local slope on seismic data based on the principal component analysis(PCA) an d the multi-scale pyramid decomposition. The image edges and image gradients are acquired in different scales and the image gradient will be broken down into non-overlapping blocks in each scale in the same way. We will get the local slope of each sub-block by singular value decomposition(SVD) as the result that the PCA is equivalent with SVD. And then an iteration algorithm is applied to transfer slope information from low resolution images to high resolution images different scales which ensures the method insensitive to noise. In this paper, the noise immunity will be validated via the processing of real seismic data and synthetic data with noise or not.