Investigation and application of a Gaussian-based noise model in infrared computational imaging

Infrared imaging systems play an important role in many fields. With the increasing demand for system miniaturization, infrared computational imaging technology achieves aberration correction through image restoration algorithms, becoming an effective way to simplify system configuration. The accurate modeling of detector noise is a critical step in infrared computational imaging technology, and its precision directly impacts the efficacy of image restoration. This article elucidates the random noise model of pyroelectric detectors through experimental research and proposes a dynamic noise model to enhance the performance of infrared computational imaging systems. The study initially established a noise testing system for pyroelectric detectors, designed to operate under various temperature conditions in accordance with the specifications of infrared detector noise testing. Subsequently, it conducted a comprehensive analysis of the detectors' noise characteristics. The test results indicate that the random noise conforms to a Gaussian distribution without Poisson components. Furthermore, as integration time and detector temperature increase, the noise demonstrates linear and exponential growth trends respectively. Building upon these findings, dynamic noise models were integrated into imaging noise simulations. Results show that compared with traditional single random noise models, the newly proposed dynamic model improves peak signal-to-noise ratio by 9.17dB, demonstrating its effectiveness in enhancing performance within infrared computing imaging systems. Through extensive exploration into random noises associated with pyroelectric detectors and proposing dynamic models for such noises, this paper not only advances understanding regarding their characteristics but also presents novel strategies for optimizing infrared computational imaging systems—offering significant theoretical significance and practical value.