Low-rank plus sparse decomposition for exoplanet detection in direct-imaging ADI sequences Article Swipe
YOU?
·
· 2016
· Open Access
·
· DOI: https://doi.org/10.1051/0004-6361/201527387
· OA: W2281424698
Data processing constitutes a critical component of high-contrast exoplanet\nimaging. Its role is almost as important as the choice of a coronagraph or a\nwavefront control system, and it is intertwined with the chosen observing\nstrategy. Among the data processing techniques for angular differential imaging\n(ADI), the most recent is the family of principal component analysis (PCA)\nbased algorithms. PCA serves, in this case, as a subspace projection technique\nfor constructing a reference point spread function (PSF) that can be subtracted\nfrom the science data for boosting the detectability of potential companions\npresent in the data. Unfortunately, when building this reference PSF from the\nscience data itself, PCA comes with certain limitations such as the sensitivity\nof the lower dimensional orthogonal subspace to non-Gaussian noise. Inspired by\nrecent advances in machine learning algorithms such as robust PCA, we aim to\npropose a localized subspace projection technique that surpasses current\nPCA-based post-processing algorithms in terms of the detectability of\ncompanions at near real-time speed, a quality that will be useful for future\ndirect imaging surveys. We used randomized low-rank approximation methods\nrecently proposed in the machine learning literature, coupled with entry-wise\nthresholding to decompose an ADI image sequence locally into low-rank, sparse,\nand Gaussian noise components (LLSG). This local three-term decomposition\nseparates the starlight and the associated speckle noise from the planetary\nsignal, which mostly remains in the sparse term. We tested the performance of\nour new algorithm on a long ADI sequence obtained on beta Pictoris with\nVLT/NACO. Compared to a standard PCA approach, LLSG decomposition reaches a\nhigher signal-to-noise ratio and has an overall better performance in the\nreceiver operating characteristic space. (abridged).\n
