1: \begin{abstract}

2: Globular clusters (GCs) are considered strong candidates for hosting rogue (free-floating) planets. Since they are not bound to a star, they are undetectable by any traditional detection methods: transit, radial velocity, or direct imaging. Gravitational microlensing (ML), which causes transient brightening of background stars by passing foreground masses, is, on the other hand, an established method of detecting planets and proves promising for application in GCs. By employing the image subtraction technique, differential photometry on the time-series images of GCs could extract variability events, build light curves and inspect them for the presence of microlensing. However, instrumental anomalies and varying observing conditions over a long observational campaign period result in the distortion of stellar Point Spread Function (PSF), which affects the subtraction quality and leads to false-positive transient detection and large-scale noise structure in the subtracted images. We propose an iterative image reconstruction method as a modification to the Scaled Gradient Projection (SGP) algorithm, called the Flux-Conserving Scaled Gradient Projection (FC-SGP), to restore the shapes of stars while preserving their flux well within the photometrically accepted tolerance. We perform an extensive empirical comparative study of FC-SGP with different image restoration algorithms like the Richardson-Lucy (RL) and the original SGP algorithms, using several physically motivated metrics and experimental convergence analysis. We find that FC-SGP could be a promising approach for astronomical image restoration. In the future, we aim to extend its application to different image formats while maintaining the performance of the proposed algorithm.

3: 

4: \end{abstract}

5: