No-reference image quality assessment using visual codebooks.

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Objective image quality assessment (IQA) models aim to predict perceived image quality automatically.
• No-reference IQA (NR-IQA) is crucial as reference images are often unavailable in practical scenarios.
• Existing NR-IQA methods are frequently distortion-specific, limiting their applicability.

Purpose of the Study:

• To develop a general-purpose NR-IQA approach that does not rely on prior knowledge of distortion types.
• To create a computational model capable of accurately predicting human-perceived image quality from distorted images without a reference.

Main Methods:

• Proposed a novel NR-IQA approach utilizing visual codebooks constructed from Gabor-filter-based local features.
• Extracted local features from image patches to capture complex natural image statistics.
• Encoded image statistics by quantizing feature space and accumulating histograms of patch appearances.

Main Results:

• The proposed method demonstrated consistency between predicted quality scores and human perception.
• Achieved comparable performance to state-of-the-art general-purpose NR-IQA methods.
• Outperformed full-reference metrics like Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) on a specific IQA database.

Conclusions:

• The visual codebook approach offers a robust and generalizable solution for NR-IQA.
• This method provides an effective alternative to distortion-specific models in practical applications.
• The approach shows promise for advancing automated image quality evaluation.