EDDMF: An Efficient Deep Discrepancy Measuring Framework for Full-Reference Light Field Image Quality Assessment.

Area of Science:

• Computer Vision
• Image Processing
• Deep Learning

Background:

• Growing demand for immersive experiences drives research in Light Field Image (LFI) quality assessment.
• Existing methods may face challenges in efficiently evaluating LFI quality degradation.

Purpose of the Study:

• To propose an efficient deep discrepancy measuring framework for full-reference LFI quality assessment.
• To develop a novel metric that accurately evaluates quality degradation in distorted LFIs.

Main Methods:

• A patch generation module extracts spatio-angular and sub-aperture patches to reduce computational cost.
• A hierarchical discrepancy network using CNNs extracts features from spatio-angular patches.
• Local discrepancy features from sub-aperture patches are used as complementary information.
• Angular-dominant and spatial-dominant features are combined for patch quality evaluation.

Main Results:

• The proposed framework achieves superior performance compared to state-of-the-art metrics.
• Demonstrates lower computational complexity in LFI quality assessment.
• Validated on four representative LFI datasets.

Conclusions:

• The developed framework is the first patch-based, full-reference LFI quality assessment metric utilizing deep learning.
• Offers an efficient and effective solution for LFI quality evaluation.