Nonlinear pixel non-uniformity: emulation and correction

All infrared focal plane array (FPA) sensors suffer from spatial non-uniformity or fixed-pattern noise (FPN). The severity of the FPN depends on the underlying manufacturing materials, methods, and tolerances, and can greatly affect overall imager performance. A key part of sensor characterization is the ability to map a known input radiance to an observed output digital count value. The presence of FPN requires a per-pixel response to be measured and specified. With this forward model defined, the inverse can be used to correct the spatial variation and ensure FPN does not corrupt other measurement estimates. In general, both the forward and inverse models are nonlinear in nature and require special care to ensure correct implementation. In this correspondence we outline a least squares emulation and correction estimation method for linear and nonlinear correction terms. We discuss the tradeoffs between computational complexity for different non-linear functions and the potential gains in reduction of fixed pattern noise. The algorithms utilize centering and scaling to improve numerical stability and is computationally efficient. In support of the reproducible research effort, the Matlab functions associated with this work can be found on the Mathworks file exchange [1].