Taming Generative Diffusion Model for Task-Oriented Infrared Imaging

Figure 1. Overview of the proposed framework and comparative results. The sub-figure (a) presents the main components of the proposed framework and highlights the key design features that support high-quality visual enhancement and improved semantic performance. The sub-figure (b) presents a comprehensive evaluation from three perspectives: visual fidelity, semantic metrics, and computational efficiency. The sub-figure (c) further presents examples of visualization results for infrared reconstruction, detection, and segmentation tasks.