US 11,836,895 B2
Deep-learning based structure reconstruction method and apparatus
Xin Gao, Thuwal (SA); Yu Li, Thuwal (SA); and Renmin Han, Thuwal (SA)
Assigned to KING ABDULLAH UNIVERSITY OF SCIENCE AND TECHNOLOGY, Thuwal (SA)
Filed by KING ABDULLAH UNIVERSITY OF SCIENCE AND TECHNOLOGY, Thuwal (SA)
Filed on Jun. 24, 2022, as Appl. No. 17/848,780.
Application 17/848,780 is a division of application No. 16/961,376, granted, now 11,403,735, previously published as PCT/IB2018/059636, filed on Dec. 4, 2018.
Claims priority of provisional application 62/621,642, filed on Jan. 25, 2018.
Prior Publication US 2022/0343465 A1, Oct. 27, 2022
Int. Cl. G06T 3/40 (2006.01); G01N 21/64 (2006.01); G06N 3/08 (2023.01); G06T 5/00 (2006.01); G06T 5/20 (2006.01); G06T 5/50 (2006.01)
CPC G06T 3/4076 (2013.01) [G01N 21/6458 (2013.01); G06N 3/08 (2013.01); G06T 3/4046 (2013.01); G06T 5/002 (2013.01); G06T 5/20 (2013.01); G06T 5/50 (2013.01); G06T 2207/10056 (2013.01); G06T 2207/10064 (2013.01); G06T 2207/20084 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A method for generating a super-resolution image, the method comprising:
receiving a time-series of fluorescent images having a first resolution;
processing the time-series of fluorescent images with a residual network module to generate denoised images; and
multiscale upsampling the denoised images with a multiscale upsampling component for generating the super-resolution image, having a second resolution,
wherein the second resolution is larger than the first resolution,
wherein the second resolution is beyond a diffraction limit of light, and
wherein the step of processing comprises:
applying a Monte Carlo dropout layer to the time-series of fluorescent images; and
applying an output of the Monte Carlo dropout layer simultaneously to (1) a residual block and (2) a denoise shortcut layer.