Mutual information based objective model for assessment of visual quality
Bibhuprasad Mohanty, Tamanna Sahoo · Journal of Information and Optimization Sciences · 2022
The humans are the ultimate assessor of image or video quality. Thus, evaluating the quality using human perception is a very crucial and interesting topic in the quality assessment research domain. Even though the subjective assessment is the most reliable one, it is time consuming and costly. On the other hand, the objective model provides the assessment result with much less time and labour, albeit with certain degree of tolerable error. In this piece of work, a new objective model, namely visual attentive information fidelity (VAIF) has been proposed in the spatial domain, instead of transformed domain, for assessment of visual quality in a scene or in a patch of the scene. VAIF relates mutual information (MI) and image fidelity using source and destination model. The proposed objective model metrics has been correlated with subjective measurement metrics like MOS, NAS.