Where does thinning collide?

In image processing, thinning is a morphological operation that repeatedly strips boundary pixels from a binary shape until only a connected, one-pixel-wide skeleton remains, so that the object's essential topology is kept while its bulk is discarded. Zhang and Suen's 1984 A fast parallel algorithm for thinning digital patterns gave a widely used parallel rule set that deletes contour pixels without breaking the pattern apart. The referent is an algorithm that reduces a digital pattern to a medial skeleton, and it has nothing in common with cutting trees from a forest or discarding samples from a Markov chain beyond the word's meaning of making something thinner.[1]

Unit

The resulting one-pixel-wide skeleton of a binary pattern

Recovery means

Not applicable

Measured by

Judged by preservation of connectivity and topology and by skeleton width and length after iteration

In Markov chain Monte Carlo, thinning means retaining only every k-th draw from a correlated sampler and discarding the rest, originally to cut the autocorrelation and storage of a long chain. Geyer's 1992 Practical Markov chain Monte Carlo both describes the practice and cautions that thinning usually throws away usable information rather than improving estimates. The referent is a subsampling rule applied to a dependent sequence of draws, a formal-statistics construct unrelated to removing stems from a stand or pixels from an image apart from the shared idea of thinning out.[2]

Unit

The thinning interval k and the count of retained draws

Recovery means

Not applicable

Measured by

Evaluated by the autocorrelation of retained draws and the effective sample size relative to the full chain

In silviculture, thinning is the deliberate removal of part of a forest stand before final harvest so that the retained trees have more light, water, and growing space and put on diameter faster. Smith, Larson, Kelty, and Ashton's The practice of silviculture sets out the thinning methods (low, crown, selection, and mechanical) by which a manager redistributes growth onto fewer, better stems. The referent is a stand-tending treatment measured in stems and basal area removed, the same word that in statistics means discarding draws and in image processing means eroding a shape, but a wholly different thing.[3]

Unit

The proportion of stems or basal area removed and the residual stand density

Recovery means

Not applicable

Measured by

Tracked by pre- and post-treatment stand density, residual basal area, and spacing between retained trees