Fast and High Quality Highlight Removal From a Single Image.

Area of Science:

• Computer Vision
• Image Processing
• Computational Photography

Background:

• Specular reflection in photography causes color deviation.
• Existing highlight removal methods struggle with diverse natural scenes.

Purpose of the Study:

• To develop a fast and high-quality highlight removal technique for single nature images.
• To address the challenge of wide applicability in diverse natural scenes.

Main Methods:

• Utilizes an L2 chromaticity definition and a dichromatic model.
• Derives a normalized dichromatic model with a unit circle equation for projection coefficients.
• Employs robust clustering in an illumination-orthogonal subspace and a pure diffuse pixels distribution rule in an illumination-parallel subspace.

Main Results:

• Achieves robust highlight removal by adaptively determining cluster numbers.
• Effectively maps specular-influenced pixels to their diffuse components.
• Demonstrates superior performance across various challenging highlight removal scenarios.

Conclusions:

• The proposed analytic solution offers an efficient and effective method for highlight removal.
• The technique is applicable to high-resolution images due to minimal complex calculations.
• This approach significantly improves color accuracy by accurately removing specular reflections.