Data-driven superresolution imaging in disordered media

Area of Science:

• Wave propagation and imaging in complex media.
• Seismic imaging and geophysical exploration.
• Acoustic and elastic wave phenomena.

Background:

• Estimating Green's functions is crucial for seismic imaging.
• Strongly scattering media pose significant challenges for conventional imaging techniques.
• Superresolution imaging offers enhanced resolution beyond the diffraction limit.

Purpose of the Study:

• To develop a robust methodology for estimating Green's functions in strongly scattering media.
• To achieve superresolution imaging by leveraging ambient scattering.
• To demonstrate the effectiveness of the proposed method using optimization or neural networks.

Main Methods:

• Exploiting large and diverse datasets to estimate ambient medium's Green's functions.
• Utilizing conventional optimization methods for imaging.
• Employing neural networks for superresolution imaging.
• Analyzing the enlargement of the physical imaging aperture due to ambient scattering.

Main Results:

• Accurate estimation of Green's functions in strongly scattering media.
• Achieved excellent imaging results with resolution superior to homogeneous media.
• Demonstrated superresolution phenomenon enabled by ambient scattering.
• Validated the methodology using both optimization and neural network approaches.

Conclusions:

• The proposed methodology enables accurate Green's function estimation and superresolution imaging in complex media.
• Ambient scattering effectively enlarges the imaging aperture, leading to enhanced resolution.
• This approach overcomes limitations of traditional time reversal methods for imaging.