The noise power spectrum (NPS) of an image sensor provides the spectral noise properties needed to evaluate sensor performance. Hence, measuring an accurate NPS is important. However, the fixed pattern noise from the sensor's nonuniform gain inflates the NPS, which is measured from images acquired by the sensor. Detrending the low-frequency fixed pattern is traditionally used to accurately measure NPS. However, detrending methods cannot remove high-frequency fixed patterns. In order to efficiently correct the fixed pattern noise, a gain-correction technique based on the gain map can be used. The gain map is generated using the average of uniformly illuminated images without any objects. Increasing the number of images n for averaging can reduce the remaining photon noise in the gain map and yield accurate NPS values. However, for practical finite n, the photon noise also significantly inflates NPS. In this paper, a nonuniform-gain image formation model is proposed and the performance of the gain correction is theoretically analyzed in terms of the signal-to-noise ratio (SNR). It is shown that the SNR is O (√n). An NPS measurement algorithm based on the gain map is then proposed for any given n. Under a weak nonuniform gain assumption, another measurement algorithm based on the image difference is also proposed. For real radiography image detectors, the proposed algorithms are compared with traditional detrending and subtraction methods, and it is shown that as few as two images (n = 1) can provide an accurate NPS because of the compensation constant (1 + 1/n).

In this paper, we propose the block-coordinate Gauss-Newton/regression method in order to conduct a correlation-based registration considering the intensity difference between images in the presence of outlier objects. In the proposed method, the parameters are decomposed into two blocks, one of which is for the spatial registration and the other for the intensity compensation. The two blocks are sequentially updated by the Gauss-Newton update and the polynomial regression, respectively. Because of the separated blocks, we can perform a joint optimization with low computational complexity and high implementation flexibility. For example, we apply separately appropriate scaling techniques to the parameter blocks for a stable and fast convergence of the algorithm. Furthermore, we apply the constrained monotone regression with a robust outlier detection scheme for the intensity compensation block. From numerical results, it is shown that the proposed algorithm more effectively performs a correlation-based registration considering the intensity difference alleviating the influence of the outlier objects compared to the traditional registration algorithms that perform the joint optimization.

Noise power spectrum (NPS) at zero frequency can efficiently represent noise properties of imaging systems. However, estimations of zero-frequency NPS are susceptible to distortion due to several factors leading to low-estimate precisions. In this letter, a high-precision estimate algorithm for zero-frequency NPS is proposed. The proposed estimate is a compensated sum of signal variances and covariances. The accuracy and precision of the proposed approach were first observed with synthetic data. A flat panel radiography detector based on CsI(Tl)-scintillator was then used to experimentally observe the improved estimate precision of the proposed algorithm. Experiment results showed that the estimation precision was improved to 14 times compared to that of the conventional method.