Low-Light Image Enhancement via the Absorption Light Scattering Model.

Area of Science:

• Computer Vision
• Image Processing
• Optics

Background:

• Low light conditions significantly degrade image visibility, impacting computer vision algorithm performance.
• Existing methods struggle to effectively enhance low-light images without amplifying noise.

Purpose of the Study:

• To propose a novel low-light image enhancement method based on a new physical model.
• To improve the visibility and detail reproduction of images captured in low-light environments.

Main Methods:

• Development of the absorption light scattering model (ALSM) to explain low-light imaging.
• Utilizing superpixels to constrain local similarity and prevent noise amplification.
• Introducing a soft jointed mean-standard-deviation (MSD) mechanism for adaptive transmittance estimation.

Main Results:

• The ALSM effectively reproduces hidden details and outlines from low-light images.
• Superpixel constraints prevent noise amplification during the enhancement process.
• The proposed MSD mechanism achieves superior transmittance estimation compared to existing strategies.

Conclusions:

• The developed low-light image enhancement method demonstrates significant advantages over current techniques.
• The ALSM provides a robust framework for understanding and improving low-light image quality.
• The integration of superpixels and MSD mechanism offers an effective approach for detail enhancement and noise reduction.