Change-point Detection and Image Segmentation for Time Series of Astrophysical Images

Many astrophysical phenomena are time-varying, in the sense that their intensity, energy spectrum, and/or spatial distribution of emission suddenly change. This paper develops a method for modeling time series images. Under assumption arrival times photons follow Poisson process, data binned into 4D grids voxels (time, band, and x-y coordinates), viewed as non-homogeneous The assumes at each point, corresponding multiband image stack is an unknown 3D piecewise constant function including noise. It also all stacks between any two adjacent change points (in domain) share same function. proposed designed to estimate number locations domain), well functions pairs points. applies minimum description length principle perform this task. A practical algorithm developed solve complicated optimization problem. Simulation experiments applications real sets show enjoys very promising empirical properties. Applications sets, XMM observation flaring star emerging solar coronal loop, illustrate usage scientific insight gained from it.