Performance optimization of three down-sampling imaging strategies and their comparison with the conventional Fourier telescope

Abstract

In order to obtain a new imaging strategy of the Fourier telescope (FT) with a better imaging quality and a less imaging time, we optimize and compare three down-sampling imaging strategies in this paper: the compressed sensing method (CS), the low-frequency full sampling method (LF) and the variable-density random sampling method (VD), which are different from the traditional Fourier telescope in both of the image quality and the imaging time. The analytical methods are as follows: based on the target’s spectral data obtained from the field experiment of traditional FT, three down-sampling methods (LF, VD and CS) are used to reconstruct the target’s images according to their own sampling modes and reconstruction methods, respectively; the differences between the three down-sampling methods and the traditional FT regarding the image quality are compared by the instinctive observation and the Strehl ratio; based on the analysis of the imaging time, the differences between the three down-sampling methods and the traditional FT regarding the imaging time are preliminarily compared. The analysis shows that: 1) the image quality of the compressed sensing method is better than that of the other two down-sampling methods (LF and VD), slightly lower than that of the traditional imaging; 2) although the image quality of the compressed sensing method is slightly lower than that of the traditional FT, its imaging time is much lower than that of the traditional FT; 3) the field data used in the analysis contain noises, which means that the reconstruction methods of the above three down-sampling strategies have a better robustness to the noises. Based on the above results, it can be seen that the Fourier telescope based on compressed sensing (CS-FT) is an excellent imaging strategy which can greatly reduce the imaging time in the condition with actual noises.

Motion artifact reduction for magnetic resonance imaging with deep learning and k-space analysis

“…Among the different CS algorithms, the split Bregman algorithm [ 3 ] is superior to the other algorithms in terms of reconstruction speed. Compared with full-sampled images, images reconstructed with CS normally still lose some image details, and are slightly blurrier [ 4 ], however, CS has already been widely adapted in commercial MRI scanners because of meeting the requirement of clinical use from under-sampled k-space data.…”