Towards a Full-Reference Quality Assessment for Color Images Using Directional Statistics

Area of Science:

• Computer Vision
• Image Processing
• Perceptual Quality Assessment

Background:

• Accurate quantification of color image perceptual quality is crucial for image processing.
• Existing metrics often struggle with diverse chromatic and achromatic distortions.
• Subjective evaluations are the gold standard but are time-consuming and costly.

Purpose of the Study:

• To develop a novel computational model for quantifying color image perceptual quality.
• To ensure the model aligns with subjective human evaluations.
• To create a robust metric effective across various image distortions.

Main Methods:

• A full-reference color metric based on directional statistics.
• Utilizes local color descriptors from hue, chroma, and lightness channels.
• Employs directional statistics for periodic hue data and weighting mechanisms for score aggregation.

Main Results:

• The proposed metric demonstrates consistent performance across numerous chromatic and achromatic distortions.
• Extensive experiments on large-scale databases validate its effectiveness.
• The metric shows superior performance compared to existing methods in predicting visual quality.

Conclusions:

• The directional statistics-based color similarity index offers a reliable method for color image quality assessment.
• This novel metric is well-suited for evaluating and optimizing color image processing algorithms.
• The model provides a computationally efficient and accurate alternative to subjective evaluations.