Subpixel Registration for Dual-Layer Detectors Based on the Amplitude Response

In this paper, we propose a novel subpixel registration algorithm by maximizing the amplitude response of the convex combination of an image pair, which is obtained from a slant-edge phantom as a fiducial mark. This algorithm does not rely on resampling with interpolation or upsampling, and requires only a small portion of the slant-edge image. We next introduce a modified algorithm, which employs a translation based on interpolation. Here, the amplitude and phase changes due to the interpolation are analyzed and compensation schemes are developed for the modified algorithm. A cyclic-coordinate optimization to estimate the horizontal and vertical translations is formulated. The estimate accuracy and precision of the proposed algorithm are better than the existing registration cases. Through numerical evaluations, the estimate mean absolute error was $0.3021\times 10^{-3}$ pixel. The proposed algorithm was applied to register the lower-layer image with respect to the upper-layer image acquired from the dual-layer detectors in radiography imaging.