A least area approach of Hough Transform for pattern recognition in seismic reflection
The intensive computational complexity, storage requirement, recognition of envelopes with parametric equations have rendered the standard Hough Transform interior to the intellectual mechanism of human's visual recognition. In this paper, a Modified Generalized Hough Transform is proposed which implements parametric equation of a straight line to establish relationship among various samples lying on a seismic reflector. It performs thresholding demarcation of interfaces in each window, establishment of continuity of interfaces from one window to another and mid-point smoothing. Based on its application on a zero-offset synthetic seismogram and a migrated CDP section presented here, the algorithm is found to be sensitive to the choice of window length. A time window of six samples or less with an equal degree of scrolling seems to the best suited for pattern recognition using this technique. However, a trade-off is necessary between the window length and the amount of overlap to obtain better reflection continuity.
Bibliogr. 6 poz.