Figure 1.
Flow chart of our algorithm.
Figure 2.
Shows restoration procedure of sparse and low-rank decomposition.
From left to right, from top to bottom: the first image of the chest CT image sequence, the sparse component, the low-rank component, the recovery image of the sparse component, the recovery image of the low-rank component and the recovery CT image. We can see that the detailed information and the visual effect of the CT image after restoration are obviously improved.
Figure 3.
Shows comparison of restoration effect of three low-rank models in dealing with the noiseless CT images.
From left to right: the 10th original chest CT image, the result of RPCA model, the result of LADMAP model, the result of GoDec model.We can see that for noiseless CT image, our method can get satisfactory results, but ringing effect of the recovery image based on the RPCA model is much smaller than the other two models, its advantage in visual effect especially shown in the component which is circled with red ellipses in the images.
Table 1.
Comparison of evaluation index values of the recovery images of the three models.
Figure 4.
Shows comparison of restoration effect of three low-rank models in dealing with the CT images that contain large noise.
From left to right: the 10th original stomach CT image, the result of RPCA model, the result of LADMAP model, the result of GoDec model, we can see our method can do some good effect on the medical CT images which contain large noise, the contrast of the images after restoration enhance obviously and the human organ boundaries are clear, there is more detailed information shown out, but the recovery image based on GoDec model shows less noise than the other two images.
Table 2.
Comparison of evaluation index values of the recovery images of the three models.
Figure 5.
Shows comparisons of different methods.
From top to bottom: chest CT experiment, lung CT experiment, stomach CT (1) experiment, stomach CT (2) experiment. From left to right in each row: the original image, the recovery image using Wang's method, the recovery image using Hussien's method, the recovery image using our method. It needs to say, we select the RPCA model in chest CT and lung CT, and select the GoDec model in stomach CT. We can see that using our method to restore the medical CT image sequence can obtain satisfactory results, the recovery images can show high contrast, clear detail information and clear organ boundaries which are especially shown in the components circled with red ellipses in the images.
Table 3.
Comparison of evaluation index values of the chest CT image.
Table 4.
Comparison of evaluation index values of the lung CT image.
Table 5.
Comparison of evaluation index values of the stomach CT image (1).
Table 6.
Comparison of evaluation index values of the stomach CT image (2).