Individual Contrast Preferences in Natural Images

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Image Processing

Background:

• Image quality assessment (IQA) traditionally uses objective metrics.
• Personalized IQA acknowledges individual observer differences.
• Understanding user preferences for image attributes like contrast is crucial for personalized IQA.

Purpose of the Study:

• To investigate individual differences in contrast preferences for natural images.
• To identify distinct groups of observers based on their contrast preferences.
• To create a publicly available database of contrast preferences for future research.

Main Methods:

• An in-lab experiment was conducted with 22 observers.
• Observers assessed 499 natural images, providing contrast level preferences.
• A three-alternative forced choice method with a modified adaptive staircase algorithm was employed.

Main Results:

• Observers were clustered into three distinct groups: low contrast, natural contrast, and high contrast preference.
• Significant individual variations in contrast preferences were observed.
• A database of 10,978 contrast preference values was compiled.

Conclusions:

• Individual contrast preferences exist and can be categorized.
• These findings support the development of personalized image quality assessment systems.
• The released database will enable further research into viewer-specific image perception.