Quaternion wavelet transform based full reference image quality assessment for multiply distorted images

Area of Science:

• Computer Vision
• Signal Processing
• Image Processing

Background:

• Real-world image distortions often involve multiple degradation types simultaneously, posing challenges for traditional single-distortion assessment methods.
• Existing full-reference image quality assessment (FR IQA) metrics may not adequately capture the complex degradations present in multiply distorted images.

Purpose of the Study:

• To develop a novel FR IQA metric specifically designed for images affected by multiple distortions.
• To leverage the power of Quaternion Wavelet Transform (QWT) for a more comprehensive analysis of image quality.

Main Methods:

• Reference and distorted images were decomposed using QWT.
• Local similarity of both phase and magnitude was computed for each QWT subband.
• A weighting method incorporating Human Visual System (HVS) characteristics was employed to construct the IQA metric.
• Quality scores from individual subbands were averaged to yield the final image quality score.

Main Results:

• The proposed QWT-based FR IQA metric demonstrated superior performance compared to state-of-the-art methods on multiply distorted images.
• The joint analysis of phase and magnitude information via QWT proved effective in capturing image quality degradation.

Conclusions:

• The developed QWT-based FR IQA metric offers a robust solution for assessing the quality of multiply distorted images.
• Incorporating HVS characteristics and QWT provides a more accurate and reliable image quality evaluation framework.