Impact of noise on reconstruction performance. (a) Mean absolute error (MAE) and cross-correlation coefficient (CC) of the reconstructions as a function of the applied noise level σ, where the noise amplitude is scaled by the mean brightness of each frame. The model remains robust up to noise levels of approximately 5% of the mean brightness (﹩\overline{{\rm{SNR}}}=20﹩), after which the performance gradually degrades. Values represent the mean over the full time series, with error bars indicating the standard deviation. (b) Example of a noisy input frame illustrating that noise becomes dominant in the faint outer regions of the corona, leading to the reduced reconstruction quality observed at higher noise levels.