The interpretation of borehole images begins with the detection and classification of features—a time-consuming manual process subject to variations between interpreters. In seeking to automate the detection part for the most frequently picked features (which in circumferential images from clastic rock environments are sinusoids corresponding to planar or subplanar bedding surfaces and fractures), it is not necessary to pick all instances, but it is necessary to pick sufficient representative instances to satisfy the interpretation objective, accounting for a broad range of apparent dips, and allowing for the likelihood of fractures crossing bedding surfaces. A key challenge in this context is the minimization of false picks, as manual corrections would potentially negate the principal benefit of automation. A fast nonsubjective method is described for the detection of prominent discontinuities and the calculation of associated dip angles. It combines a gradient based approach for edge detection with a phase congruency method for validation, followed by a robust sinusoid detection technique. It has been evaluated on microresistivity images from wireline and logging-while-drilling tools, these images having a wide range of features with varying degrees of geologic complexity; the proportion of false positives in the case of noisy data is less than 5%, improving to better than 2% in the case of good-quality data. In contrast to manual picking, the method is fast and gives reproducible results. With potentially thousands of sinusoids in a single image, the method dramatically improves efficiency.