Image Quality Assessment: Integrating Model-Centric and Data-Centric Approaches
This addresses the incremental improvement of IQA methods for researchers by integrating model and data approaches to reduce overfitting and enhance dataset creation.
The paper tackles the problem of isolated model-centric and data-centric approaches in image quality assessment (IQA), which impedes progress, by proposing a computational framework that integrates both; experimental results show that a sampling-worthiness module successfully identifies diverse failures of blind IQA models for inclusion in next-generation datasets.
Learning-based image quality assessment (IQA) has made remarkable progress in the past decade, but nearly all consider the two key components -- model and data -- in isolation. Specifically, model-centric IQA focuses on developing ``better'' objective quality methods on fixed and extensively reused datasets, with a great danger of overfitting. Data-centric IQA involves conducting psychophysical experiments to construct ``better'' human-annotated datasets, which unfortunately ignores current IQA models during dataset creation. In this paper, we first design a series of experiments to probe computationally that such isolation of model and data impedes further progress of IQA. We then describe a computational framework that integrates model-centric and data-centric IQA. As a specific example, we design computational modules to quantify the sampling-worthiness of candidate images. Experimental results show that the proposed sampling-worthiness module successfully spots diverse failures of the examined blind IQA models, which are indeed worthy samples to be included in next-generation datasets.