Gradient Magnitude Similarity Deviation: A Highly Efficient Perceptual Image Quality Index

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Accurate image quality assessment (IQA) is crucial for applications like image compression and restoration.
• Existing IQA models need to balance prediction accuracy with computational efficiency, especially with increasing visual data volumes.
• Image gradients are sensitive to distortions, and different image regions degrade unevenly.

Purpose of the Study:

• To develop a novel, effective, and computationally efficient IQA model.
• To leverage gradient information for predicting perceptual image quality.

Main Methods:

• Introduced the Gradient Magnitude Similarity Deviation (GMSD) model.
• Utilized pixel-wise gradient magnitude similarity (GMS) between reference and distorted images.
• Employed a novel pooling strategy: the standard deviation of the GMS map for overall quality prediction.

Main Results:

• GMSD achieves highly competitive prediction accuracy for perceptual image quality.
• The GMSD algorithm demonstrates significantly faster computation compared to state-of-the-art IQA methods.
• The standard deviation of the GMS map effectively captures global variations in local image quality.

Conclusions:

• GMSD offers a promising solution for efficient and accurate image quality assessment.
• The method's speed and accuracy make it suitable for high-volume visual data processing.
• Gradient magnitude similarity deviation provides a robust metric for perceptual image quality evaluation.