A Novel Image Formation Model for Descattering.

Area of Science:

• Optics
• Computer Vision
• Image Processing

Background:

• Existing image descattering models simplify scattering distribution, assuming uniformity even in homogeneous media.
• This simplification leads to inaccuracies and limits the performance of current restoration techniques.
• Experiments show scattering is non-uniform, even in homogeneous environments.

Purpose of the Study:

• To propose a novel image formation model for image descattering that accounts for non-uniform scattering.
• To develop algorithms for parameter estimation and image restoration using the new model.
• To improve the accuracy and effectiveness of image descattering and related applications.

Main Methods:

• Developed a new image formation model incorporating physical parameters: zenith angle, azimuth angle, scattering phase function, and camera focal length.
• Introduced algorithms for estimating model parameters from degraded images.
• Implemented simultaneous restoration of descattered images using the proposed model and algorithms.

Main Results:

• The proposed model accurately describes light transfer in scattering media.
• Parameter estimation results closely match real values.
• Outperformed state-of-the-art methods in underwater image restoration.
• Enhanced performance of existing image dehazing methods by replacing traditional models.

Conclusions:

• The novel image formation model accurately captures non-uniform scattering phenomena.
• The proposed approach significantly improves image descattering, dehazing, and underwater image restoration.
• This work provides a more physically grounded foundation for scattering media image restoration.