Adaptive thresholding for robust iterative image reconstruction from limited views projection data

X-ray computed tomography (CT) is generally obtained through a reconstruction of object attenuation function from its projection data acquired through many angle views. However, in some cases the number of projection views is limited to only a small number that is theoretically not enough for stable reconstruction. The theory of compressed sensing (CS) introduce a new framework for solving this problem. In this paper, we present a simple iterative thresholding image reconstruction algorithm that include a minimization of cost function that include Log-likelihood and ℓ₁ norm distance to a reference image. The resulting thresholding function in the reconstruction algorithm is approximated with adaptive function that can preserve low-contrast regions from incorrect thresholding. Moreover, we present a simple approach to handle registration error between the reconstructed image and the reference. This approach is based on using dynamic reference image that generated from previous estimation of the reference image and the current image estimate. The proposed method proved to be robust in image reconstruction from small number of projection views through a simulation study.