It is sometimes desirable to identify a predictive relationship between different classes of images and the subjective impressions of image quality that they generate.  To discover these relationships, we use psychometric functions, linear regression and factor analysis. We also develop, implement and test image quality engineering metrics.  

The image quality engineering metrics that have some predictive value (for particular viewing and judgment tasks) include:

• Flicker metric (based on human temporal sensitivity)
• Color difference metrics (CIELAB DE)
• SCIELAB (spatially-weighted DE)
• MSE weighted by human contrast sensitivity
• Visibility metrics based on a multi-channel vision models