Haralick

Haralick is a surname. In technical literature, it is most commonly associated with Haralick texture features, a set of statistical descriptors used to quantify texture in grayscale images. The features were introduced in a 1973 paper by Robert M. Haralick, Kenneth Shanmugam, and Ida Dinstein, based on the concept of a gray-level co-occurrence matrix (GLCM).

To compute them, the image is quantized to a finite number of gray levels. For a given offset and direction, a GLCM is built by counting how often pairs of pixel values occur at the specified spatial relationship. From the matrix, multiple features can be derived, including angular second moment (energy), contrast, entropy, homogeneity, and correlation; additional features such as maximum correlation coefficient and various variances and entropies have been proposed. In practice, features are often computed for several directions (e.g., horizontal, vertical, and diagonals) and for several gray-level quantizations.

Applications include texture classification, segmentation, and material inspection; they are used in medical imaging and remote sensing. Limitations include sensitivity to image noise, rotation and scale changes, and the need for gray-level quantization. They form a foundational approach in texture analysis and have inspired many subsequent feature sets.