Channel boundary detection using partial area effect with sub-pixel resolution

Abstract

Channel facies, of the most prevalent stratigraphic features, are essential from the perspective of hydrocarbon exploration. However, measuring edge features using the gradient vector's calculation in an individual pixel mainly produces edge borders with pixel resolution even in images with low noise levels. This paper introduces a novel seismic attribute based on the partial area effect algorithm assuming a particular discontinuity in the edge location to extract the buried channel boundaries. First, the Sobel-based edge detector extracts the pixels' partial derivatives as probable channel boundaries. Then, the direction and the dimension of the postulated partial area effect mask in the local maxima points are deliberated to elicit the edge curve coefficients as the perfect edge positions. The proposed algorithm was implemented on a synthetic time slice comprising two-channel events with varying thicknesses and sinuosity. The procedure was also investigated on two high-cut frequency filtered datasets from the Penobscot prospect in the Nova Scotia Basin accommodating canyon-channel events. The proposed algorithm was then compared to the prominent gradient-based edge detectors (Canny, Prewitt, and Sobel) and some conventional seismic attributes (apparent dip, most negative curvature, and similarity). The results showed that applying the nominated algorithm could provide an improved map of buried channels less affected by the coherent noise and acquisition footprints. The partial area effect algorithm also successfully localized more true-positive edge points and fewer false-positive ones in the synthetic and field seismic data examples.