MULF RR-IQA: Multi-feature based reduced reference image quality assessment

This paper introduces a novel reduced reference image quality assessment technique called “MULF RR-IQA.” The technique is based on the assumption that the marginal distribution of wavelet coefficients within each subband can be accurately modeled by a Gaussian distribution. In our proposed method, we first extract features, including the marginal distributions of neighboring coefficients and entropy in the wavelet domain. These extracted features are then combined, and the image quality is predicted using a similarity measure. We validate the performance of the proposed approach on the widely recognized LIVE Image database, demonstrating a strong correlation between the human evaluations and the objective scores obtained from the proposed method.