Seismic image segmentation and its application to geologic interpretation

Image segmentation is a standard low-level task in computer vision. We demonstrate how the segmentation of two-dimensional (2-D) seismic images can assist in geologic interpretation. An image segmentation technique is applied to 2-D images of the acoustic impedance. A salient feature of the resulting segments is that the hydrocarbon bearing regions are characterized by a finer segmentation than the other regions. A modified apparent polarity attribute is introduced and images of this attribute are segmented. The boundaries of the resulting segments pass through the minima and the maxima of the acoustic impedance in a manner which allows us to segment the acoustic impedance image into layers of high and low acoustic impedance. Our methods can assist in segmentation of acoustic impedance images according to lithology or geologic facies.