Bayesian inference for identifying solar active regions
M. Turmon, Saleem Mukhtar, Judit M. Pap · 1997
The solar chromosphere consists of three classes plagc, network, background- which couttibute ctiffcrcutly to ultraviolet ractiatiou reacl]iug the earth. Solar physicists are illtermted iu relatiug plage area aud iuteusity to UV irradiauce, as well as uuctcrstamtiug the spatial aud tcm~)oral evolutiotl of plage sha~)cs. WC clescribc a clata set of solar images, meaus of segumatiug the images iuto corlstitucllt classes, a[ld a novel higll-level re~)rcscutatiorl for compact ot)jccts based ou a spatial ‘mcmhcrship fullctio~l’ ctefiuecf via a triaugulatcxt planar graph. Segulclltatious are fouud usiug a discrete hiarkov rauclom fkkl setup, auct the high-level rqJrescutatious are lcarucct by a Markov cllaiu hIolite Carlo birtll/death process on the triaugulatious,