Local Entropy¶
Local entropy computes the Shannon entropy of the intensity distribution within a small window around each pixel — high where the neighborhood’s intensities are spread out and unpredictable (rough, noisy, or highly textured regions), low where they’re concentrated on one or a few values (flat, uniform regions). It’s a fast, single-number way to flag “busy” vs. “quiet” areas of an image, complementing the more structured texture descriptors like Gabor and LBP.
See also
Local Binary Pattern (LBP) — a richer, pattern-based texture descriptor
Histograms — the intensity distribution entropy is computed from
Feature Extraction — every other feature