Half of an image is enough for quality assessment
This provides insights into IQA model interpretability for researchers, though it appears incremental as it extends findings to existing CNN-based models.
The study tackled the problem of understanding how deep models assess image quality by introducing a positional masked transformer for Image Quality Assessment (IQA), revealing that half of an image plays a trivial role while the other half is critical, with semantic measures like saliency showing high correlation to region importance.
Deep networks have demonstrated promising results in the field of Image Quality Assessment (IQA). However, there has been limited research on understanding how deep models in IQA work. This study introduces a novel positional masked transformer for IQA and provides insights into the contribution of different regions of an image towards its overall quality. Results indicate that half of an image may play a trivial role in determining image quality, while the other half is critical. This observation is extended to several other CNN-based IQA models, revealing that half of the image regions can significantly impact the overall image quality. To further enhance our understanding, three semantic measures (saliency, frequency, and objectness) were derived and found to have high correlation with the importance of image regions in IQA.