Fault detection using seismic attributes and visual saliency

In order to detect accurately faults in seismic inline sections, we propose a new bottom-up saliency based approach using different seismic attributes such as coherence, curvature, dip, and gradient in parallel. Each attribute is calculated independently from the original seismic section. The saliency maps of aforementioned attributes are computed using covariance matrix, which are later combined to form a consolidated saliency map that highlights the seismic fault regions. The covariance matrix is used to characterize the seismic patches and captures local structures. By thresholding the variance maps and optimizing the binary points for curve fitting, the proposed workflow yields good results for faults labeling.