Underwater image quality assessment method based on color space multi-feature fusion

Area of Science:

• Computer Vision
• Image Processing
• Digital Signal Processing

Background:

• Underwater environments cause significant image degradation, impacting visual quality.
• Existing image quality assessment (IQA) methods are insufficient for underwater imagery.
• Accurate measurement of underwater image quality is crucial for subsequent processing.

Purpose of the Study:

• To develop an effective Underwater Image Quality Assessment (UIQA) method tailored for underwater image characteristics.
• To improve the performance of IQA methods in assessing degraded underwater images.

Main Methods:

• Proposed a UIQA method utilizing multi-feature fusion within the CIELab color space.
• Extracted histogram, morphological, and moment statistics from luminance and color components.
• Employed Support Vector Regression (SVR) to build a quality prediction model from fused features.

Main Results:

• The proposed UIQA method demonstrated strong performance on the SAUD and UIED datasets.
• Cross-dataset evaluations on LIVE, TID2013, LIVEMD, LIVEC, and SIQAD datasets confirmed the method's applicability.
• The multi-feature fusion approach effectively quantifies underwater image quality degradation.

Conclusions:

• The proposed CIELab color space multi-feature fusion method offers a robust solution for underwater image quality assessment.
• This approach enhances the accuracy and reliability of evaluating visual quality in challenging underwater conditions.
• The method shows broad applicability across various image quality assessment datasets.