issue contents
March 2025 issue

Cover illustration: Deep learning-based reconstruction of high-resolution and high-speed X-ray videos for high-speed imaging: (left) selected low-resolution, high-speed video frames of 4 times larger pixel size, (middle) selected high-resolution, low-speed video frames of 20 times larger frame separation, (right) selected high-resolution, high-speed video frame reconstruction (see Tang, Bicer, Sun, Fezzaa and Clark, pages 432–441). The results show that the new method can significantly improve the reconstruction, achieving an average peak signal-to-noise ratio of more than 35 dB on the two representative X-ray image sequences.
facility information
research papers



















A deep learning-based algorithm is developed and evaluated that demonstrates the potential to reconstruct simultaneously high-resolution high-frame-rate X-ray image sequences with high fidelity through spatio-temporal fusion. Experimental evaluation shows that the method can significantly improve the accuracy of the reconstruction, achieving an average peak signal-to-noise ratio (PSNR) of more than 35 dB on two representative X-ray image sequences with input data streams of four times lower spatial resolution and 20 times lower frame rate, respectively.

beamlines




computer programs
