Outgoing Longwave Radiation (OLR) is an important component of the Earth's radiation budget and a key parameter for coupled models of the atmosphere, ocean, land, and other systems. It is of significant importance in studies related to Earth sciences such as weather forecasting, climate research, and disaster monitoring. Since narrowband sensors are more widely available and have higher spatial resolutions than broadband sensors, high-resolution OLR data are currently frequently estimated using narrowband sensors. This study proposes a novel physical method, namely the Multi-Dimensional matrix MAPping algorithm (MDMAP) framework, inspired by the scene classification ideas of Cloud and Earth's Radiant Energy System (CERES) and the differential absorption theory. The new framework aims to accurately retrieve OLR from the multi-channel infrared sensor, such as Moderate Resolution Imaging Spectroradiometer (MODIS). Corresponding to traditional algorithms, such as the polynomial regression algorithm (POLY) and lookup table algorithm (LUT), the new framework provides two distinct implementations of the MDMAP algorithm framework (MDMAP:POLY and MDMAP:LUT). The performances of both the traditional and the newly proposed algorithms are evaluated based on the radiative transfer simulation dataset and CERES SSF OLR products. The results show that the MDMAP algorithms behave more accurately than the traditional ones under most conditions, especially under clear-sky conditions. Specifically, a comprehensive analysis indicates that the new algorithms demonstrate smaller RMSEs than the traditional ones under various conditions, particularly in desert regions with the RMSE reduction exceeding 3 W/m2 (>30%). Moreover, the two new algorithms reveal enhanced robustness to noise uncertainties, and demonstrate remarkable generality and computational efficiency, implying their potential and better applicability in deriving believable OLR from most infrared sensors.