Images recorded with the Tunable Magnetograph (TuMag) instrument on board the Sunrise III balloon observatory are affected by temporal contrast fluctuations. These variations are related to high-frequency vibrations (jitter) of the platform that cannot be entirely compensated by the image stabilization unit of the telescope. Jitter smooths out mid- to high-frequency details and degrades the contrast of images that are integrated over hundreds of milliseconds to achieve the signal-to-noise levels required by the instrument. The temporal variations of jitter hamper the reconstruction of the data and limit the accuracy of the instrument. We present a technique that allows us to infer and compensate temporal changes of the amplitude and direction of jitter from two images of the same scene, provided they are recorded close enough in time to prevent the scene from evolving significantly. We assess the performance of our method through a dynamical magnetohydrodynamical simulation of the solar scene that is degraded with diffraction effects, noise, and jitter. We evaluate the performance of the algorithm as a function of the jitter amplitude and as a function of the time gap between the two images. We finally apply it to a time series recorded by TuMag to sense and restore the temporal evolution of jitter on the images. Our numerical experiments show that the algorithm is able to sense jitter with an accuracy better than 0 (łesssim 0.1 px) for images recorded less than 10 s apart and assuming a worst-case scenario with a low signal-to-noise ratio of 100 and induced rms jitter amplitudes as large as 0 in the X and Y directions. When increasing the time interval to 30 s, the errors found are still below 0 For the chosen TuMag series we find that the jitter rms amplitudes vary strongly within the range 0 (sim1-4 px) and are anticorrelated to the contrast of the recorded images. Correction of jitter improves the contrast of wavefront-reconstructed images by 2.5,% in some cases and reduces the rms variation of contrast along the series by half.
Bailén et al. (Wed,) studied this question.