Near-Infrared Coloring via a Contrast-Preserving Mapping Model

Area of Science:

• Computer Vision
• Image Processing
• Digital Photography

Background:

• Near-infrared (NIR) gray images paired with visible color images offer benefits for image restoration and classification.
• Existing naive coloring methods struggle with discrepancies in brightness and structure between NIR and visible images, leading to unrealistic results.

Purpose of the Study:

• To develop a new, realistic coloring method for near-infrared gray images.
• To address the limitations of naive coloring techniques by resolving discrepancies between NIR and visible image data.

Main Methods:

• A novel contrast-preserving mapping model was developed to adjust NIR images, matching their luminance to visible images while retaining original contrast and details.
• A method was created to derive and transfer realistic colors from visible images to the adjusted NIR images using the proposed mapping model.

Main Results:

• The proposed method successfully preserves local contrast and image details from the original NIR images.
• Realistic colors were effectively transferred from visible images to the processed NIR images.
• The NIR coloring technique demonstrated efficacy in noise and haze removal and local contrast enhancement.

Conclusions:

• The developed contrast-preserving mapping model and coloring method provide a significant improvement for NIR image processing.
• This technique enables realistic colorization of NIR images, enhancing their utility for various applications including image restoration and classification.