Light scattering control with the two-step focusing method based on neural networks and multi-pixel coding.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning Applications

Background:

• Focusing light through scattering media is crucial for advanced optical imaging and deep tissue penetration.
• Existing methods often face limitations in achieving high-quality focus and significant light enhancement.

Purpose of the Study:

• To develop a novel two-step focusing technique utilizing neural networks (NNs) and multi-pixel coding.
• To achieve high-quality light focusing with near-theoretical maximum enhancement in scattering media.

Main Methods:

• A single-layer neural network (SLNN) generates an initial focusing mask.
• Multi-pixel coding encodes this initial mask, and a second SLNN is trained using coded masks and speckle patterns to obtain the final high-quality focusing mask.
• Experimental validation using a 16x16 modulation unit mask with 8 pixels per unit.

Main Results:

• Achieved a light focusing enhancement of 40.3, closely approaching the theoretical maximum.
• Demonstrated significant improvements over single-step methods, with 220% and 24% higher enhancement compared to using only the initial mask or direct multi-pixel encoding, respectively.
• Required 3000 images (1000 for the first step, 2000 for the second) for training.

Conclusions:

• The proposed two-step NN-based method effectively enhances light focusing through scattering media.
• This approach offers a promising new strategy for improving optical imaging resolution and penetration depth.
• The method provides a significant enhancement over conventional techniques.