An improved ROBust OpTimization-based (iROBOT) fusion model for reliable spatiotemporal seamless remote sensing data reconstruction

• iROBOT is proposed as an improved version of the ROBOT fusion model for spatiotemporal reconstruction. • It replaces fixed rectangular blocks with spectrally and spatially homogeneous segments to suppress block artifacts. • An adaptive gap-filling strategy, coupled with low-quality information detection, is introduced to improve robustness under cloud-affected conditions. • Experiments across two representative areas show consistent improvements over ROBOT, OL-HSTFM, and FSDAF. • It achieves higher accuracy while maintaining efficiency, supporting reliable seamless data cube reconstruction at global scales. Spatiotemporal fusion is recognized as a feasible approach for reconstructing high spatiotemporal resolution remote sensing imagery. The recently proposed ROBust OpTimization-based (ROBOT) fusion model has attracted attention for its efficiency, accuracy, and simple parameterization, and has been applied to seamless global data cube reconstruction. However, its sliding-window strategy often leads to block artifacts, while cloud contamination in auxiliary images further degrade the reconstruction quality and reliability. To address these limitations, this study proposes an improved version of ROBOT (iROBOT). By replacing fixed rectangular patches with spectrally and spatially homogeneous segments, iROBOT effectively suppresses block artifacts and better preserves structural details. In addition, an adaptive gap-filling strategy is applied to temporally nearest auxiliary images prior to fusion, and a low-quality information detection module is introduced to filter unreliable cloud-contaminated data from the regression and residual allocation steps, thereby enhancing reliability under cloud contamination. A total of 28 experiments conducted across two representative areas demonstrate that iROBOT consistently outperforms ROBOT, OL-HSTFM, and FSDAF under both cloud-free and cloud-contaminated scenarios. Moreover, compared with ROBOT, iROBOT exhibits superior capability in selecting reliable information from multi-temporal auxiliary images, with its accuracy advantage becoming more pronounced as the number of auxiliary inputs increases, and with notably enhanced robustness in cloud-prone environments. These findings highlight the potential of iROBOT for improving the accuracy of global-scale spatiotemporal image reconstruction. Code is found at https://github.com/Andy-cumt/iROBOT_spatiotemporal-fusion .