EVALUATION OF MODEL BASED PARAMETRIC IMAGE ESTIMATION IN MIR

Abstract

Multiple-image radiography (MIR), a new X-ray phase sensitive imaging method, can simultaneously produce parametric images of absorption, refraction, and ultra-small-angle scatter. In the past it has been shown that these three parametric images, when compared to conventional X-ray methods (e.g. mammograms), have excellent diagnostic abilities, thanks to an X-ray phase-contrast mechanism and higher-angle scatter rejections. In this paper we present an evaluation of a newly proposed parametric image estimation method based on a physical model of the image formation in MIR. The proposed method employs the fact that the angular intensity profile (AIP) of the measured X-ray, after interaction with the object, can be modeled as a convolution of a Gaussian curve and the intrinsic imaging system AIP. Further, the new estimation method utilizes an iterative conjugate gradients algorithm that, in combination with the model, offers improvement in parametric image estimation accuracy even in the presence of the imaging noise.