.png){kind=icon}
Return
Frequency domain generative diffusion model for temporal compressive coherent diffraction imaging
.png){kind=icon}
PRE
AI
DOI：10.1016/j.optlaseng.2024.108424.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Temporal compressive CDI utilizes single-exposure snapshot compression sampling in the frequency domain to visualize the dynamic process of CDI. Due to the lack of effective frequency domain prior information constraint, the traditional temporal compressive CDI method has limited imaging quality. In order to achieve high-quality dynamic reconstruction at high frame rates, a score-based diffusion model in frequency domain is used to extract the prior gradient distribution of the target in the complex frequency domain. Subsequently, this information is harmoniously woven into the physics-driven iterative phase recovery reconstruction process to restore multiframe high-quality images. Simulation and experimental outcomes have demonstrated that the proposed technique surpasses conventional methods, delivering enhanced structural details and a reduction in artifact. The highest achieved PSNR and SSIM values are 37.25 dB and 0.997, respectively. The proposed method has equally well performance on targets across datasets, demonstrating the strong generalization of the method.
Keywords:
Coherent diffraction imaging
High -quality imaging
Image reconstruction
Generative diffusion model
Journal
IF:
3.7
Papers:
7.3K
Citations:
1.7W
Organization
Cited Papers
ShareSaveShareSaveChemical composition mapping with nanometre resolution by soft X-ray microscopy.png)
NATURE PHOTONICS
NATURE PHOTONICSIF32.9
PREAI
ShareSaveShareSavePhysics-driven deep learning enables temporal compressive coherent diffraction imagingOPTICA
OAAI
OPTICAIF8.5
OAAIShareSaveShareSave
View more