Fig. 5.

Typical image processing operations with scikit-image. Data are synthetic, unless stated otherwise. a Filtering−Top non-local means denoising of an image with a fine-grained texture, acquired by in situ synchrotron microtomography during glass melting [21]. Bottom total-variation denoising of an image with two phases, corresponding to phase-separating silicate melts observed by in situ tomography [7]. b Feature extraction−Top Hubble deep field (NASA, public domain), blob detection using the Laplacian of Gaussian method. Bottom ridge detection using the leading eigenvalue of the Hessian matrix, neuron image from CREMI challenge (https://cremi.org/data/). c Segmentation—Top super-pixel segmentation of a CT slice of the human head [13], using Felzenszwalb’s algorithm [18]. Bottom random walker segmentation (right) of noisy image (top-left corner), using histogram-determined markers (bottom-left corner). d Measures—Top visualization of local diameter (color-coded on the skeleton curve) of an interconnected phase (represented in violet). Bottom particles color-coded according to their extent