Accurate and robust Q estimation based on low-rank constraint of the attenuation function and DBSCAN clustering

Abstract Seismic attenuation decreases the resolution of the observed seismic data, which can affect the performance of subsequent seismic processing and interpretation. The attenuation can be assessed by the quality factor Q, which it is a prerequisite of high-resolution seismic attenuation compensation. However, the precision and stability of Q estimation are often influenced by various factors, including noise interference and super-parameter selection. Thus, we propose a novel Q estimation algorithm that uses the priors of the attenuation function and implementation-related Q estimations. The attenuation between the amplitude spectra of the observed and reference wavelets is of low-rank property, based on which we propose a low-rank constrained inversion algorithm to estimate the attenuation function. After that, a linear fitting method is used to derive equivalent Q. Besides, a DBSCAN clustering method is introduced to cluster implementation-related Q estimations. The clustering center is set as the final equivalent Q, where the correct Q is clustered and the incorrect one is divergent. The layered Q is achieved via a constrained transformation. Synthetic layered model and simulated VSP data quantitatively demonstrate the Q estimation validity of the proposed method. Field VSP data processing further qualitatively proves its effectiveness for reliable Q estimation.