Coupled strata and granulometry detection on indurated cores by gray-level image analysis

Abstract

In this paper, we propose an approach combining both strata detection and grain analysis on images of indurated cores. We use high-resolution images (47 µm/pixel) for granulometry and low-resolution images (270 µm/pixel) for strata detection, stored in indexed database. The quality of these images is sufficient for an interpretation identical and even higher than the geologist’s eye. The methods we have implemented and adapted for strata detection are based (1) on time-frequency transformations (Fourier), and (2) on scale-frequency transformations (wavelet). A comparison between these two methods is provided for representative image examples—we demonstrate that the wavelet transform method is more efficient because it allows to better identify strata and to distinguish these from fractures which are due to drilling operations or to geological processes (tectonic movements, recrystallization). For the detection and analysis of grain size, by contrast, we decided to use the most classical mathematical morphological operator in imagery—the regional maxima, widely used in granulometry, and well adapted to the detection of small grains. The sequential description (stratification) on low-resolution images, and the granulometry assessment on high-resolution images constitute two distinct tasks, performed in parallel. For each layer detected, we can perform a specific granulometric analysis. The efficiency, reliability and repeatability of our methods have been tested without any subjectivity. The samples used for this purpose were taken from a borehole in the Cretaceous to Upper Eocene volcano-sedimentary series of the New Caledonian west coast.