Spatial and frequency domain metrics for assessing fixed-pattern noise in infrared images

In this paper, two complementary metrics for assessing the spatial structure and the amount of additive Non-Uniformity (NU) estimated from raw Infrared (IR) images have been proposed and analyzed. The first metric, which has been defined in the spatial domain, quantifies the spatial similarity between an estimate and a sample of the NU noise using the cosine distance. The second metric characterizes the spectrum of the NU noise, and uses this characterization to quantify the fraction of NU noise that has been properly estimated. To test the effectiveness and consistency of the proposed metrics in assessing the NU noise, five previously reported NU correction algorithms have been ranked. To evaluate metrics´s simplicity, the most computing intensive and resource demanding metric has been implemented on an inexpensive FPGA with low resource utilization.