Color demosaicking via fully directional estimation.

Area of Science:

• Computer Vision
• Image Processing
• Digital Signal Processing

Background:

• Single-sensor imaging requires demosaicking to reconstruct full color images.
• Existing demosaicking algorithms often struggle with perceptual quality and artifacts.

Purpose of the Study:

• To develop an effective demosaicking algorithm that improves perceptual quality.
• To reduce interpolation artifacts in reconstructed color images.

Main Methods:

• Utilizes fully directional estimation based on the Bayer color filter array pattern.
• Employs multiscale gradients in green channels for diagonal information.
• Introduces combinational weights from horizontal, vertical, diagonal, and anti-diagonal directions.
• Divides color difference calculations into two guided parts to minimize artifacts.

Main Results:

• Outperforms traditional horizontal and vertical estimation methods in perceptual quality.
• Demonstrates efficiency in both objective and subjective image quality measures.
• Effectively reduces artifacts commonly seen in interpolation procedures.

Conclusions:

• The proposed adaptive demosaicking algorithm offers superior performance.
• The method provides a significant improvement in full color image reconstruction.
• This approach is efficient and yields high-quality results with fewer artifacts.