Comparative Study of Edge Detection Algorithms

TLDR: In this paper, the authors evaluate the sensitivity of classic derivative, Canny and Torreao and Amaral edge detection algorithms used in image processing in their ability to extract the full spectrum of geological features in the presence of random noise and acquisition footprint.

Abstract: Summary Important geological features like faults and channels often manifest themselves as edges in seismic data. Several classes of edge detection algorithms have been developed to extract the edges of seismic data. Two of the most basic and straightforward classes are that of the gradient edge detectors and Gaussian edge detectors. In this paper we evaluate the sensitivity of classic derivative, Canny and Torreao and Amaral edge detection algorithms used in image processing in their ability to extract the full spectrum of geological features in the presence of random noise and acquisition footprint. Geological features of interest may include linear but arbitrary oriented faults and fractures, sinuous channels, and highly irregular but organized karasting and dewatering features. Through application to real data we conclude that Torreao and Amaral, Canny, and derivative algorithms exhibit better performance respectively. We show that these algorithms perform better in detecting faults and channels when noise is suppressed.