Enhancement of human visual perception-based image quality analyzer for assessment of contrast enhancement methods

Abstract EN

Prior to this work, Human Visual Perception (HVP) -based Image Quality Analyzer (IQA) has been proposed.

The HVP-based IQA correlates with human judgment better than the existing IQAs which are commonly used for the assessment of contrast enhancement techniques.

This paper highlights the shortcomings of the HVP-based IQA such as high computational complexity, excessive (six) threshold parameter tuning and high performance sensitivity to the change in the threshold parameters’ value .

In order to overcome the aforementioned problems, this paper proposes several enhancements such as replacement of local entropy with edge magnitude in sub-image texture analysis, down-sampling of image spatial resolution, removal of luminance masking and incorporation of famous Weber-Fechner Law on human perception.

The enhanced HVPbased IQA requires far less computation (>189 times lesser) while still showing excellent correlation (Pearson Correlation Coefficient, PCC > 0.90, Root Mean Square Error, RMSE<0.3410) with human judgment.

Besides, it requires fewer (two) threshold parameter tuning while maintaining consistent performance across wide range of threshold parameters’ value, making it feasible for real-time video processing.