Recovering saturated images for high dynamic kernel-phase analysis

Kernel-phase observables extracted from mid- to high-Strehl images are proving to be a powerful tool to probe within a few angular resolution elements of point sources. The attainable contrast is limited, however, by the dynamic range of the imaging sensors. The Fourier interpretation of images with pixels exposed beyond the saturation has so far been avoided. In cases where the image is dominated by the light of a point source, we show that we can use an interpolation to reconstruct the otherwise lost pixels with an accuracy sufficient to enable the extraction of kernel-phases from the patched image. We demonstrate the usability of our method by applying it to archive images of the Gl 494AB system taken with the Hubble Space Telescope in 1997. Using this new data point along with other resolved observations and radial velocity measurements, we produce improved constraints on the orbital parameters of the system, and consequently the masses of its components.