We present simulated Large Binocular Telescope (LBT) infrared narrow-band observations of a star-jet system, in conjunction with improved and optimized deconvolution and image reconstruction algorithms, considering two cases of interest: single-dish direct imaging with an AO-assisted camera and imaging through a Fizeau interferometer that combines the beams of the two mirrors of LBT. We aim at understanding what accuracy can be obtained with the use of present AO-assisted large telescopes (such as LBT) and what improvements an interferometric instrument (such as LINC-NIRVANA) will be able to provide. The proposed deconvolution method is based on the target decomposition as a sum of a point source (the star) and an extended source (the jet). By assuming Poisson noise we add to the negative logarithm of the likelihood a regularization term enforcing smoothness of the jet component. Finally, we use a Richardson-Lucy-like method for the minimization of this function. This approach is an improvement of a method proposed by Lucy in 1994 for accurate photometric restoration of HST images and called two channel photometric restoration. We denote the new method as the multi-component Richardson-Lucy (MC-RL) method. The analysis of the reconstructed objects shows that the MC-RL method applied to the interferometric observations allows us to evaluate the width and the spatial intensity profile of the jet down to 20 mas with an accuracy better than about 20% in the best case of a central star fainter than 10 mag. These limits allow us to obtain a very good reconstruction of the jet acceleration region very close to the exciting source, which would provide fundamental scientific information on the jet collimation degree and eventually on its launching mechanism. As concerns the proposed MC-RL method, it demonstrates a good performance in the reconstruction of images with a very high dynamic range. It can be improved in several directions, by increasing both its efficiency, thanks to recently proposed acceleration techniques, and its accuracy by means of more sophisticated regularization terms. We are also planning to apply the method to simulated observations of upcoming super giant earth-based telescopes.