An unsupervised strategy for the global tagging of individual diffractions

As a by-product to a zero-offset section with a significantly improved signal-to-noise ratio, multi-parameter stacking schemes like the common-reflection-surface (CRS) stack also extract a set of physically meaningful wavefront attributes from the data. Although these attributes vary laterally along the events, an analysis of their local similarity allows the global identification of measurements, which stem from the same diffractor or passive source, i.e., from the same subsurface region. We present a fully automatic scheme to globally identify and tag diffractions in simple and complex data by evaluating local attribute similarity. Since wave propagation is a smooth process and due to the assumption of local attribute similarity, this approach is not restricted to settings with moderate subsurface heterogeneity. Results for simple synthetic 2D and 3D diffraction data and an application to complex marine field data confirm the potential of the proposed method.