A velocity spectrum or semblance spectrum is often used to estimate normal-moveout (NMO) velocity. The concept of velocity spectrum or semblance was initially introduced as a measure of the correlation of multichannel seismic data (Neidell and Taner, 1971; Taner and Koehler, 1969). Nowadays, semblance-based NMO velocity analysis is still popular and is the most practical and trustworthy method for velocity model building from seismic data. However, geophysicists are not satisfied with the performance of the conventional semblance because of the long-standing limitations of the semblance calculations; for example, the low resolution that causes erroneous picking, expensive computation cost for the semblance in anisotropic media, and poor adaptability to the seismic data with AVO. As a result, many improved methods of semblance have been proposed to address these problems (Chen et al., 2015; Fomel, 2009; Hu et al., 2015; Luo and Hale, 2012; Sarkar et al., 2001,2002).
Sarkar et al. (2001,2002) introduced a new semblance calculation that took into account AVO information. Fomel (2009) reformulated this AVO semblance as a squared correlation coefficient between the data and a trend function, and called it AB semblance. Luo and Hale (2012) proposed a resolution-improved semblance by weighting the data to honor large-offset data and dampen small-offset data. Hu et al. (2015) used a fast butterfly algorithm that handles the large computational power required for 3D anisotropic velocity analysis. Chen et al. (2015) utilized a different weighting strategy for improving the resolution of the semblance by weighting the data according to the local similarity of each trace with a reference trace.
Weighted stacking of seismic AVO data
using hybrid AB semblance and local similarity