KR-20260065171-A - System for restoring phase images using Gabor holograms based on unsupervised learning and method thereof

KR20260065171AKR 20260065171 AKR20260065171 AKR 20260065171AKR-20260065171-A

Abstract

The present invention relates to an unsupervised learning-based phase image restoration system and a method using a Gabor hologram. According to the present invention, an unsupervised learning-based phase image restoration system using a Gabor hologram may include: a cyclic coherence module that generates pair data in which an inline hologram and a non-optical axial phase image are not paired, by applying the data to a pre-trained unsupervised model; and a noise removal module that generates a final phase image by applying the generated pair data to a pre-trained diffusion model. As such, according to the present invention, a high-quality quantitative phase image can be obtained cost-effectively from a Gabor single-shot digital hologram. In addition, biological samples can be analyzed with high accuracy in a miniaturized holographic setup.

Inventors

문인규
박성환

Assignees

재단법인대구경북과학기술원

Dates

Publication Date: 20260508
Application Date: 20241101

Claims (10)

A cyclic coherence module that generates paired data of inline holograms and non-optical phase images by applying data in which the inline hologram and non-optical phase image are not paired to a pre-trained unsupervised model; and An unsupervised learning-based phase image restoration system using a Gabor hologram, comprising a noise removal module that generates a final phase image by applying the generated pair data to a pre-trained diffusion model.
In paragraph 1, The above unsupervised model is, A first generator that generates a first hologram and a first phase image; and An unsupervised learning-based phase image restoration system using a Gabor hologram, comprising a first discriminator that receives the generated first hologram and the actual hologram as input and determines truth and falsehood through binarization.
In paragraph 1, The above diffusion model is, A second generator that generates a second phase image with noise removed using a phase image including a hologram generated from the above-mentioned cyclic coherence module and noise at time t; and An unsupervised learning-based phase image restoration system using a Gabor hologram, comprising a second discriminator that receives a phase image containing noise at actual time point tk and a third phase image synthesized from the noise at time point tk in the second phase image, and determines truth and falsehood through binarization.
In paragraph 3, The above second generator is, The hologram generated from the above-mentioned cyclic consistency module and the phase image containing noise at time t are used by the fourth generator ( An unsupervised learning-based phase image restoration system using a Gabor hologram that generates a second phase image with noise removed by applying it to ).
In paragraph 3, The above second discriminator is, Distinguish between the actual noisy phase image at time step tk and the synthesized noisy phase image at time step tk, and An unsupervised learning-based phase image restoration system using a Gabor hologram that extracts noise from the result determined by the second generator and removes noise from the third phase image to generate a final phase image.
A step of applying data in which the inline hologram and the non-optical axis phase image are not paired to a pre-trained unsupervised model to generate paired data in which the inline hologram and the non-optical axis phase image are matched; and An unsupervised learning-based phase image restoration method using a Gabor hologram, comprising the step of applying the generated pair data to a pre-trained diffusion model to generate a final phase image.
In paragraph 6, The above unsupervised model is, A first generator that generates a first hologram and a first phase image; and An unsupervised learning-based topological image restoration method using a Gabor hologram, comprising a first discriminator that receives the generated first hologram and the actual hologram as input and determines truth and falsehood through binarization.
In paragraph 6, The above diffusion model is, A second generator that generates a second phase image with noise removed using a phase image including a hologram generated from the step of generating the above pair data and noise at time t; and An unsupervised learning-based phase image restoration method using a Gabor hologram comprising a second discriminator that receives a phase image containing noise at actual time point tk and a third phase image synthesized from the noise at time point tk in the second phase image, and determines truth and falsehood through binarization.
In paragraph 8, The above second generator is, The hologram generated in the step of generating the above pair data and the phase image containing noise at time t are used by the fourth generator ( An unsupervised learning-based phase image restoration method using a Gabor hologram that generates a second phase image with noise removed by applying it to ).
In paragraph 8, The above second discriminator is, Distinguish between the actual noisy phase image at time step tk and the synthesized noisy phase image at time step tk, and An unsupervised learning-based phase image restoration method using a Gabor hologram, which extracts noise from the result determined by the second generator and removes noise from the third phase image to generate a final phase image.

Description

System for restoring phase images using Gabor holograms based on unsupervised learning and method thereof The present invention relates to an unsupervised learning-based phase image restoration system and method using a Gabor hologram, and more specifically, to an unsupervised learning-based phase image restoration system and method using a Gabor hologram that restores a phase image by applying an unsupervised diffusion model to a Gabor single-shot digital hologram. Cells are transparent or translucent, so imaging is performed using off-axis quantitative phase digital holography. Off-axis quantitative phase digital holography stores unlabeled holographic images of living cells at low radiation, and single holograms are processed to reconstruct a 3D image corresponding to the phase distribution and volume of the cells. There is a limitation in that precise control of optical elements is required because the reference wave and the object wave must be aligned before capturing a hologram of a biological sample, such as a cell. To compensate for this, Garbor holographic microscopy is used, but there is a problem where in-focus images and out-of-focus images overlap. In addition, reconstructing phase images from Gabor hologram images through supervised learning is labor-intensive and costly because a large number of pair datasets must be acquired, and reconstructing phase images from Gabor holograms through unsupervised learning is difficult to judge the accuracy of the results because there are no labels to verify. Therefore, there is a need for a technology to reconstruct phase images using unsupervised learning-based Gabor single-shot digital holograms. The technology forming the background of the present invention is described in Korean Published Patent No. 10-2024-0119854 (published August 6, 2024). FIG. 1 is a drawing illustrating a phase image restoration system according to one embodiment of the present invention. FIG. 2 is a flowchart of an unsupervised learning-based phase image restoration method using a Gabor hologram according to another embodiment of the present invention. FIG. 3 is a diagram illustrating the process of a cyclic consistency module according to another embodiment of the present invention. FIG. 4 is a diagram illustrating the process of a noise removal module according to another embodiment of the present invention. FIG. 5 is a diagram illustrating the entire process of a learned second generator according to another embodiment of the present invention generating a final phase image from which noise has been removed. FIG. 6 is a drawing illustrating an example of photographing red blood cells and cancer cells using an inline holographic microscope and a non-optical holographic microscope according to another embodiment of the present invention. FIG. 7 is a diagram illustrating the result of generating a final phase image using a second generator learned according to another embodiment of the present invention. FIG. 8 is a diagram illustrating quantitative cell results of a final phase image generated according to another embodiment of the present invention. FIG. 9 is a diagram comparing a case where a topological image restoration method is learned using supervised learning and a case where a topological image restoration method is learned using unsupervised learning according to another embodiment of the present invention. Preferred embodiments according to the present invention will be described in detail below with reference to the attached drawings. In this process, the thickness of lines or the size of components shown in the drawings may be exaggerated for clarity and convenience of explanation. Furthermore, the terms described below are defined in consideration of their functions within the present invention, and these may vary depending on the intent or practice of the user or operator. Therefore, the definitions of these terms should be based on the content throughout this specification. In the embodiments described below, an unsupervised learning-based phase image restoration system (100) using a Gabor hologram can be performed by a computing device. FIG. 1 is a drawing illustrating a phase image restoration system according to one embodiment of the present invention. As illustrated in FIG. 1, the phase image restoration system (100) may include a cycle consistency module (110) and a denoising module (120). First, the cyclic coherence module (110) can generate paired data by applying unpaired data, in which an in-line hologram and an off-axis phase image are not paired, to a pre-trained unsupervised model (e.g., Generative adversarial networks (GAN)). Here, the unsupervised model may be a model trained to generate paired data by receiving multiple in-line holograms and off-axis phase images as input. Specifically, the unsupervised model may include a first generator (111) that generates a first hologram and a first phase image, and a first discriminator (112) that receives t