A new image quality metric based on MIx-Scale transform

In the study of image quality assessment (IQA), multi-scale methods are often used because they can supply more flexibility than single-scale ones by incorporating the variations of viewing conditions. However, some recent work demonstrated that single-scale methods based on scale transformation also performed well in terms of the correlation between the quality predictions and the subjective scores. Both of the single-scale based IQA metrics conform to a two-step model: image transform using proper scale coefficient and the SSIM metric, but the influence of different components in SSIM (luminance, contrast and structural similarity) on the chosen scale coefficient is not taken into account. In our re-search, it was found that the scale transform coefficients for different components should be different. This paper accordingly proposes a new MIx-Scale (MIS) based IQA algorithm. Experimental results on six publicly available databases (LIVE, TID2008, CSIQ, Toyama, IVC and LIVE Multiply Distortion) confirm that the proposed IQA metric can often enhance predictive accuracy for image quality assessment comparing to traditional scale-based methods.