Automating the interpretation of post-migrated seismic data would considerably reduce the time and cost of the interpretation process. In this paper, we study the role of texture perception in automating the delineation of an important sub-surface structure: salt dome. First, we propose a new seismic attribute, the gradient of texture, based on which we develop a framework for detecting and delineating salt domes. Since the new attribute can be defined based on a variety of dissimilarity measures, we define a perceptual dissimilarity measure and compare its performance against several other perceptual and non-perceptual texture dissimilarity measures in delineating salt domes in a real seismic data set. Our experimental evaluation reveals that results based perceptual measures are more consistent with human-interpreted results. Therefore, an important contribution of this paper is confirming that the human interpretation process does not only rely on geological and geophysical knowledge, but it also relies on the visual perception of 2D seismic data. Further, the proposed perceptual dissimilarity measure yields the best result and computational efficiency among all studied measures.