Segmenting Chromospheric Images with Markov Random Fields
The solar chromosphere roughly consists of three types of region: plage, network, and background. Thresholding individual pixel intensities is typically used to identify these regions in solar images. We have incorporated spatial information by using a Bayesian setup with an image prior that prefers spatially coherent labelings; resulting segmentations are more physically reasonable. These priors are a first step in developing an appropriate model for chromospheric images.
Unable to display preview. Download preview PDF.