Absil, Pierre-AntoineJacques, LaurentMil-Homens Cavaco, NicolasNicolasMil-Homens Cavaco2025-05-142025-05-142025-05-142024https://hdl.handle.net/2078.2/37783Differential Imaging is currently a widespread technique to post-process images of a remote planetary system, captured by ground-based telescopes during an observation campaign, so that exoplanets around the host star are directly visible after post-processing. To extract crucial information about the spectrum of the exoplanets, the light is further decomposed into several spectral channels during the data acquisition process, yielding a 4-D data cube of spectral frames and temporal frames. By combining those images cleverly, Angular and Spectral Differential Imaging (ASDI) allows to remove the so-called quasi-static speckles that dramatically affect our detection capabilities, since they have the same shape as an exoplanet. This Master’s thesis has multiple objectives. Firstly, a great importance is attached to the modeling of the 4-D data cubes used in ASDI. Then, we design new ASDI post-processing algorithms based on low- rank models, which take two forms : inverse problem-based approaches and greedy algorithms. Both are finally compared with an ASDI version of the classical PCA algorithm by massively injecting fake companions in a data cube without exoplanets from the VLT. It turns out that these new algorithms correctly detect the injected fake companions as well as PCA and even outperform the latter at small angular separation.Low-rank modelsDirect imagingExoplanet detectionImage processingtext::thesis::master thesisthesis:46105