Channel Edge Detection Based On 2D Compactly-Supported Shearlets, An Application to a Channelized Seismic Data in South Caspian Sea

Shearlet transform, as a multi-scale and multi-directional transformation is highly capable of detecting features with different dips and has found several applications in image processing tasks. The anisotropy property of shearlet transform can be employed so as to detect edges where channels may occur in the seismic data. In this study, cone-adapted compactly-supported 2D shearlet transform was applied to both synthetic and real data from South Caspian Sea containing channels to decompose the data into different scales and directions. In order to detect the edges, the local maxima of shearlet coefficients in both horizontal and vertical cones were calculated and summed at the finest scale so as to obtain the final image of the local maxima. Global thresholding based on the histogram of the local maxima followed by thinning were applied to delineate the edge pixels. The results of the proposed algorithm were compared qualitatively and quantitatively with such classic image processing edge detectors as Sobel and Canny. In both synthetic and real data examples, our algorithm outperformed both Sobel and Canny operators.