Yakui Dong

The InGaAs short-wave infrared (SWIR) cameras are now widely used, but there is poor horizontal comparability between cameras. The current test and evaluation of cameras is mainly based on photon transfer theory. The specific test methods are the European Machine Vision Association (EMVA) image sensor and camera characterization test standards and the camera test protocol of teledyne photometrics (TPs). Among them, the conversion gain $K$ [e ${}^{-}$ /digital number (DN)] is a key parameter for uniformly quantifying the performance of SWIR cameras. In practical applications, a single test method is prone to deviations and fails to provide a comprehensive analysis of the error sources. In this article, we compare the theoretical foundations and experimental results of the EMVA and TP testing standards. By plotting the complete noise photon transfer curve (PTC), the mean–variance PTC, and the conversion gain curve, we conduct a comprehensive analysis of the noise and gain characteristics of the SWIR camera. Furthermore, we find that under nonideal conditions and high read noise, the results obtained from the two testing methods can validate each other, thereby determining the true conversion gain of the camera. This work contributes significantly to the in-depth understanding of camera testing standards and provides valuable insights for evaluating the performance of SWIR cameras.

Blind pixels are an inevitable problem in short-wave infrared (SWIR) InGaAs focal plane arrays (FPAs). An experiment using a laser beam-induced current method was set up to characterize the blind pixels in InGaAs FPAs. According to the result, the blind pixels are classified into four types: single overhot pixels, single dead pixels, crossing dead pixels, and paired blind pixels. Additionally, the causes of the different types of blind pixels are analyzed. It is indicated that the blind pixels are caused by defects and impurities, disconnection of the readout circuit, unformed PN junctions, and a wrong connection between the PN junction and readout integrated circuit, respectively. Based on the characterization, a blind pixel calibration method that greatly improved the quality of images taken by InGaAs FPAs is proposed.