A Metric for Light Field Reconstruction, Compression, and Display Quality Evaluation

Area of Science:

• Computer Vision
• Image Processing
• Multimedia Systems

Background:

• Light fields capture rich visual information, enabling advanced applications.
• Light field processing techniques, while beneficial, often introduce visual quality degradation.
• Existing quality metrics may not adequately address the unique distortions in processed light fields.

Purpose of the Study:

• To propose a novel full-reference light field quality metric.
• To accurately quantify visual quality loss in processed light fields.
• To address distortions arising from sparse sampling, compression, and reconstruction.

Main Methods:

• Developed a full-reference light field quality metric.
• Evaluated global spatial quality using view structure matching.
• Assessed local spatial quality via near-edge mean square error.
• Quantified angular quality through multi-view quality analysis.

Main Results:

• The proposed metric effectively captures common light field distortions, including blur, blocking, ghosting, stretching, flickering, and sampling artifacts.
• Experimental results demonstrate high accuracy in estimating light field visual quality.
• The metric outperforms existing state-of-the-art quality assessment methods for light fields.

Conclusions:

• The developed light field quality metric provides an accurate and comprehensive assessment of visual quality.
• It effectively addresses the specific distortions encountered in typical light field processing pipelines.
• This metric is valuable for optimizing light field processing and ensuring high-quality visual experiences.