The 3D Hough transform in seismic data

Abstract

The Hough transform has been widely used in feature detection and recognition in image processing. It has several attractive properties, including robustness to noisy and incomplete data and the potential to detect features which are hard to discover using other techniques (including manual analysis of the data). However, the primary focus in work on the Hough transform has been on linear and curved features in 2D images. It is possible to extend the Hough transform to three dimensional datasets, enabling the detection of planar features. The properties mentioned above makes the transform well suited for use on seismic data, which is notoriously noisy. This thesis focuses on exploring the possiblity of using the 3D Hough transform to detect planar features in seismic data. To accomplish this it was necessary to do a comprehensive literature study on the Hough transform and its applications, advantages and drawbacks. A prototype application simply named "3D Hough" was also developed and tested on real world data from the Gullfaks oil field. This prototype application can successfully detect planar structures in seismic data, as well as visualising the data along with the detected planes. However, there are several challenges involved in successfully using this technique. These include, among others, finding and setting useful parameters, pre-processing the seismic dataset to yield desirable results within the user's current region of interest, and the high running time of the algorithm itself. In spite of these drawbacks, with further research the 3D Hough transform could potentially prove useful in the analysis of seismic data.