Segmentation Algorithm Applied to Seismic Images for Obtaining Potentially Hidden Geobodies

Abstract

Image segmentation is the process of extracting information based on similarity criteria. In this paper we propose a segmentation algorithm applied in color images with seismic information in the CIELAB space. This algorithm, the IMP-2DMA, takes a set of initial values provided by the user, which are part of patterns in the image with certain characteristics. The association of pixels based on vertical expansion control variables and direction guides is performed. With the selected pixels, a set of 2D binary masks will be formed that will be part of a volume. By means of a three-dimensional projection, the resulting masks are visualized with effects of light and shadow, allowing to appreciate complex geobodies not visible at first sight. The results show that with the IMP-2DMA it is possible to extract different patterns in a similar way to those obtained manually and more accurately than with other segmentation algorithms. The Wilcoxon rank sum test was used to evaluate the performance of the IMP-2DMA. The 2D masks were compared against the ideal solution and the segmentation obtained by a threshold-based segmentation algorithm.