Sub-Pixel Object Rendering Algorithm for Wide FOV Infrared Scene Generations

Infrared scenes are simulated by modeling 3D objects, such as air platforms and missiles, materials, background, atmosphere, and sensors with respect to their radiometric characteristics in different fields of view. When a scene is generated, the distinction of objects of interest versus other background sources, the geometric resolution, and the pixel-level radiation accuracy of these objects are of great importance for the reliability and usability of the simulation results. Traditional rasterization algorithms are optimized to ensure high visual fidelity without considering significant features related to distant and infinitesimal geometry. Even though this limitation is usually insignificant in perceived visual quality, it presents a challenge for accurate infrared modeling of distant objects, especially for wide-angle scenes. Due to the decline in spatial resolution of objects at long distances, these objects constitute sub-pixels while containing critical and significant radiance information, which is represented in the final image in an unstable and inaccurate fashion. This adversely affects the detection ranges of IR objects when such simulated IR scenes are used in algorithm development. In this paper, we propose a sub-pixel rendering approach utilizing hierarchical textures and radiometry concepts, which can be easily integrated into existing rendering pipelines, to accurately reflect the radiance contribution of such barely visible objects. Our algorithm, while being time efficient and able to support multiple geometries, ensures radiance and intensity consistency at varying distances from the viewer with temporal consistency. Infrared scenes with different objects against arbitrary backgrounds and environments have been simulated using uniform and complex geometries. We demonstrate its efficiency across various scenarios, including multiple distances and aspect angles.