Edge Preserving Total Variation Regularization for Dual Energy CT Images
Dual-energy computed tomography (CT) offers the potential to recognize material properties by decomposing sinograms into Compton and photoelectric bases and subsequently reconstructing the basis images. However, the presence of high-density materials such as metal can distort the reconstructed images, leading to inaccurate material characterization. We proposed a novel reconstruction technique to reduce noise and metal artifacts in dual-energy CT images by exploiting 1) statistical correlation between measurements and decomposed sinograms, 2) intra-image correlation between decomposed images and 3) inter-image sparsity.
Related publications: Electronic Imaging 2019
GPU Accelerated 3D Millimeter Wave Image Reconstruction
SAR 3D image reconstruction is a computationally complex, ill-conditioned inverse problem. Approximation methods such as matched filter (MF) has a limited resolution. We proposed a partitioned inverse algorithm, which is robust to noise, faster and has better resolution than MF.
This scene is a reconstructed image of a metal-coated toy balloon dog (4,914,100 voxels). The reconstruction took 50 s with the proposed method and 3.5 hrs with MF (with GPU). The proposed method was 250 times faster than MF in this example. Furthermore proposed method has improvement of 24% in resolution over MF even with 0 dB SNR (simulation results).
Related lighting talks: UW EE Graduate Research Showcase, IMS 3MTT
Related publications: ICASSP 2017, IMS 2017
Accelerated Enhanced Resolution 3-D SAR Imaging With Dynamic Metamaterial Antennas
Beam steering of Dynamic Metamaterial Antennas (DMAs) can be used to improve SAR imaging. We proposed the first Enhanced Resolution Strip-map Mode (ERSM) SAR approach for 3-D imaging with DMAs, which achieved an improved resolution in all dimensions. GPU-accelerated partitioned inverse (PI) algorithms were extended to allow for the fast 3-D ERSM SAR image reconstruction.
Related publications: TMTT 2017