Abstract: In order to solve the problems of fuzzy boundary information and noise influence in identifying special geological bodies such as seismic faults and gas chimney in seismic images, the LBP / ENT edge detection algorithm is proposed, which is an improved combining LBP( Local Binary Patterns) algorithm and local ENT (Entropy). Rotation invariant unified local binary pattern is used to construct the traditional LBP. The transverse discontinuity of abnormal geological bodies is adapted in seismic images, local entropy is used to describe the local discrete characteristics of seismic images to improve the robustness of seismic noise. Compared with Canny operator, LBP Operator, local VAR(Variance) and LBP / VAR operator, the proposed LBP / ENT method is used to study the complex Marmousi theoretical model and the gas chimney on the actual seismic image. The results show that LBP / ENT can more clearly describe the edge information of seismic image and have better robustness to noise. It is concluded that the proposed LBP / ENT algorithm provides a feasible method and technology for detecting the edge information of special geological bodies on seismic images. 

Key words: local binary patterns, local entropy, edge detection, seismic fault, gas chimney