Gaussian¶
A Gaussian-smoothed copy of the image at a chosen scale (\(\sigma\)). As a standalone feature, it captures the image’s large-scale intensity trend with fine detail and noise averaged out; it’s also the building block several other filters on this page are defined in terms of (e.g. Laplacian-of-Gaussian). Computing it at several values of \(\sigma\) gives JuSPICE’s feature stack a genuinely multi-scale view of the image, from fine texture to coarse structure.
See also
Sobel / Scharr — gradient filters usually applied after Gaussian smoothing to reduce noise sensitivity
Laplace — the second-derivative filter this one underlies (Laplacian of Gaussian)
Feature Extraction — every other feature