Tellurium Sub-Oxides Infrared Phototransistors for Adaptive Super-Resolution Image Reconstruction

Infrared (IR) detection using crystalline silicon or III-V compounds is commonly utilized but often challenged by bulkiness and inefficiency. With the development of autonomous driving and machine vision, there is a growing need for IR technology to incorporate compact neural architectures. In this study, IR-sensitive p-type disordered tellurium sub-oxides (TeO_x) thin films are deposited via an inorganic blending strategy. By integrating a luminescent dielectric layer, synergistic charge transfer and photon-induced secondary excitation endow TeO_x-based IR-visible adaptive sensors (IVAS) with broadband detection and memory capabilities. The IR-driven modulation of IVAS convolutional weights enables super-resolution image reconstruction even under suboptimal conditions. This IVAS-based system achieves a peak signal-to-noise ratio of 27.55 dB (compared to 26.85 dB conventionally), a structural similarity index measure of 0.94 (compared to 0.88 conventionally), and a 13.8% reduction in mean absolute error. These findings highlight TeO_x-based IVAS as a robust and adaptive solution for IR machine vision systems.