Empirical noise power spectrum based on the image subtraction in radiography imaging

Abstract

In order to measure the noise power spectrum (NPS) of a radiography detector from acquired images without any object, a subtraction-based algorithm, in which the input image is subtracted from another image acquired at the same exposure condition, is usually used to alleviate the fixed pattern noise influence. However, depending on variations in the x-ray tube and the readout circuits in the detector, the acquired images show nonuniform statistics and thus selecting an appropriate image pair from a given image set is important for a stable measurement of NPS. In order to select a good image pair for measuring an accurate normalized NPS (NNPS), a measurement method is proposed based on selecting an image pair from comparing the signal-to-noise ratios, which are obtained from the subtracted image. For the comparison, a commutative relationship as a sufficient condition for the accurate NNPS is considered. For a given image pair, the proposed method can provide a degree of accuracy of the empirical NNPS. For real x-ray images acquired from an indirect radiography detector, the proposed method is tested and shows stable measuring results for a set of images.