Fig 1.
Diagram of limited-data tomography collected through (a) sparse-view CT scan and (b) limited-angle CT scan
.
Fig 2.
Illustration of zero-level set boundaries and multi-region level set modeling.
(a) The zero-level set curve defines the boundary between the object and the region of interest. (b) Two level sets are used to model three regions, each indicated by a unique sign combination of and
.
Fig 3.
This figure Illustrates the placement of single-scale and multiscale Gaussian Functions at each pixel, represented as circles.
Fig 4.
Synthetic images used in simulation experiments to assess the performance of the proposed method.
Fig 5.
Two letters and two images with three different gray levels were used in the simulation experiments to evaluate the performance of the proposed method.
Table 1.
Comparison of reconstruction accuracy under sparse-view and limited-angle scenarios using different methods. Values represent Dice coefficients (higher is better).
Fig 6.
High-resolution reconstruction using FBP for (a) a carved cheese slice from 360 projections, and (b) a walnut slice from 1200 projections.
Fig 7.
Performance comparison of the proposed method against various reconstruction techniques using data from only four projections for the first two tests and six projections for the last two tests.
Fig 8.
Testing the proposed method for character reconstruction using 6 projections within the angular range .
Fig 9.
Reconstruction results using projection data from 4, 6, 8, and 10 angles under noise levels of n = 0.01, 0.05, and 0.10.
Fig 10.
SSIM vs. Noise Level: Reconstruction quality comparison using 4, 6, 8, and 10 projection angles.
Fig 11.
PSNR vs. Noise Level: Reconstruction quality comparison using 4, 6, 8, and 10 projection angles.
Fig 12.
Performance comparison of the proposed method with Tikhonov, Total Variation (TV), and Dual Problem (DP) reconstructions on a real CT dataset of a carved cheese slice with sparse projections.
The sinogram was acquired from 15 projections over a 360∘ range.
Fig 13.
Performance of the proposed method for multiphase level-set functions.
Fig 14.
Comparison of reconstruction methods on real CT data of a walnut slice using 20 projections (top row) and 10 projections (bottom row).
From left to right: Filtered Back Projection (FBP), Total Variation (TV) regularization, and the proposed method.
Table 2.
Performance metrics (PSNR and SSIM) for different reconstruction methods across different sparse views.
Fig 15.
Performance of various reconstruction methods in a limited-angle test using 8 projections spanning from 0 to .
Fig 16.
Performance of the proposed method across different angular ranges using 8 projections.
Fig 17.
Reconstruction results vs. noise levels using limited projection data from 0 to ,
,
, and
.
Fig 18.
SSIM vs. Noise level: Reconstruction quality comparison using limited projection data from 0 to ,
,
, and
.
Fig 19.
PSNR vs. Noise level: Reconstruction quality comparison using limited projection data from 0 to ,
,
, and
.
Fig 20.
Performance comparison of the proposed method with Tikhonov, Total Variation (TV), and Dual Problem (DP) reconstructions on a real CT dataset of a carved cheese slice under limited-angle tomography.
The sinogram was obtained from 15 projections spanning an angular range from 1 to .
Table 3.
Performance metrics (PSNR and SSIM) for different reconstruction methods under limited-angle views using only 8 projections across various angular ranges.