Extending Camera's Capabilities in Low Light Conditions Based on LIP Enhancement Coupled with CNN Denoising

Area of Science:

• Image processing
• Computer vision
• Computational imaging

Background:

• Variable lighting, especially low-light, degrades sensor image quality.
• Image enhancement and denoising are crucial for accurate information retrieval.
• The Logarithmic Image Processing (LIP) framework, initially for transmission imaging, is adaptable for reflection imaging due to its human visual system compatibility.

Purpose of the Study:

• To explore the limits of image enhancement and denoising in variable lighting conditions.
• To preserve the quality of enhanced images, focusing on noise reduction, visual quality, and color fidelity.
• To evaluate the separate contributions of LIP enhancement and CNN denoising.

Main Methods:

• Application of the Logarithmic Image Processing (LIP) framework for image enhancement.
• Utilizing Convolutional Neural Networks (CNNs) for image denoising.
• Employing rigorous metrics to assess enhancement reliability, including noise reduction, visual image quality, and color preservation.

Main Results:

• Logarithmic Image Processing (LIP) laws accurately simulate exposure time variation for low-light enhancement.
• The combined LIP enhancement and CNN denoising effectively reduces noise and preserves image quality.
• Rigorous metrics confirm significant noise reduction and color preservation capabilities.
• Standard exposure time can be substantially reduced, expanding sensor utility.

Conclusions:

• The LIP framework combined with CNN denoising offers a robust solution for low-light image processing.
• This approach enhances image quality and reliability under challenging lighting conditions.
• Reduced exposure times significantly increase the applicability and versatility of image sensors.