ABSTRACT: Digital Color Image Analysis of Core
Rocco Difoggio, Kathryn Burleigh
Geologists often identify sands, shales, or UV-fluorescent zones by their color in photos of slabbed core or sidewalls. Similarly, they observe porosity as blue-dyed epoxy in thin sections.
Of course, it is difficult to accurately quantify the amount of sand, shale, fluorescence, or porosity by eye. With digital images, a computer can quantify the area of an image that is close in shade to a selected color, which is particularly useful for determining net sand or net fluorescence in thinly laminated zones.
Digital color photography stores a video image as a large array of numbers (512 × 400 × 3 colors) in a computer file. With 32 intensity levels each for red, green, and blue, one can distinguish 32,768 different colors.
A fluorescent streak or a shale has some natural variation in color that corresponds to hundreds of very similar shades. Thus, to process a digital image, one picks representative shades of some selected feature (e.g., fluorescence).
The computer then calculates the eigen values and eigen vectors of the mean-centered covariance matrix of these representative colors. Based on these calculations, it determines which parts of the image have colors similar enough to the representative colors to be considered part of the selected feature.
Our results show good agreement with independently measured thin-section porosity and with specially prepared images having known amounts of a given color.
AAPG Search and Discovery Article #91003©1990 AAPG Annual Convention, San Francisco, California, June 3-6, 1990