Quantifying Visual Image Quality: A Bayesian View

Area of Science:

• Computer Vision
• Image Processing
• Vision Science

Background:

• Image quality assessment (IQA) models quantify the relationship between visual images and human perception.
• IQA research has grown significantly, bridging vision science and engineering applications.
• IQA serves as a test-bed for vision theories and computational biovision models.

Purpose of the Study:

• To provide an overview of IQA methods from a Bayesian perspective.
• To unify diverse IQA approaches under a common framework.
• To offer references for vision scientists and image processing practitioners.

Main Methods:

• Review of existing image quality assessment methods.
• Application of a Bayesian framework for unifying IQA approaches.
• Analysis of IQA methods' implications for biological and artificial vision.

Main Results:

• A unified framework for understanding a wide spectrum of IQA methods.
• Discussion on the successes and limitations of modern IQA techniques.
• Insights into the interplay between biological vision and artificial vision system design.

Conclusions:

• The Bayesian perspective offers a unifying framework for IQA.
• Understanding IQA's impact on biological vision can inform future artificial vision systems.
• Further research can leverage vision science to enhance artificial vision technologies.