GcsDecolor: Gradient Correlation Similarity for Efficient Contrast Preserving Decolorization.

Area of Science:

• Computer Vision
• Image Processing

Background:

• Conventional image decolorization often relies on gradient error-norm measures, which may not fully preserve original image appearance.
• Existing methods face challenges in balancing fidelity and computational efficiency.

Purpose of the Study:

• To propose a novel gradient correlation similarity (Gcs) measure-based decolorization model.
• To develop efficient algorithms for solving the proposed model, ensuring faithful preservation of original color image appearance.

Main Methods:

• A new Gcs measure is defined, calculating gradient correlation between color image channels and the transformed grayscale image.
• Two solvers are developed: an augmented Lagrangian and alternating direction method for the approximated linear parametric model, and a discrete searching solver for real-time computation.

Main Results:

• The augmented Lagrangian and alternating direction method solver shows excellent iterative convergence and superior performance.
• The discrete searching solver offers simplicity, speed (real-time computational speed), and robustness with respect to parameters and candidates.

Conclusions:

• The proposed Gcs measure-based decolorization model and its associated algorithms demonstrate significant potential.
• Extensive experiments and comparisons confirm the effectiveness and advantages over state-of-the-art methods.