Bayesian Inference for Identifying Solar Active Regions

The solar chromosphere consists of three classes plage, network, background which contribute differently to ultraviolet radiation reachina thn am-th ~.nl~r nhv&-iota cama intmwatd in --a “Y” U-V.&. Y”*yYJ”.“‘uY” .“A.8 .Y”YIVY”VU IY relating plage area and intensity to UV irradiante, as well as understanding the spatial and temporal evolution of plage shapes. We describe a data set of solar images, means of segmenting the images into constituent classes, and a novel high-level representation for compact objects based on a spatial ‘membership function’ defmed via a triangulated planar graph. Segmentations axe found using a discrete Markov random field setup, and the high-level representations are learned by a Markov chain Monte Carlo birth/death process on the triangulations.