A Spatial and Geometry Feature-Based Quality Assessment Model for the Light Field Images

Area of Science:

• Computer Vision
• Image Processing
• Human-Computer Interaction

Background:

• Light field (LF) images capture intricate spatial and geometric scene information.
• Accurate perceptual quality assessment is crucial for LF image processing and display.

Purpose of the Study:

• To propose a novel full-reference image quality assessment (IQA) model for LF images.
• To develop a model that effectively evaluates perceptual quality by considering both spatial and geometric characteristics.

Main Methods:

• Extracted spatial features from sub-aperture images (SAIs) using contourlet transform.
• Extracted geometry features across adjacent SAIs using 3D-Gabor filters.
• Combined spatial and geometry features to compute a final IQA score.

Main Results:

• The proposed spatial and geometry feature-based model (SGFM) demonstrated superior performance.
• SGFM showed higher consistency with human visual system (HVS) perception compared to existing IQA models.
• Experimental validation on three LF IQA datasets confirmed the model's effectiveness.

Conclusions:

• The SGFM provides a robust method for LF image quality assessment.
• The model's design, based on human visual perception, enhances its accuracy.
• SGFM represents a significant advancement in LF image quality evaluation.