The aesthetic and technical quality of images is an important factor in providing users with a better visual experience. Image Quality Assessment (IQA) allows models to build a relationship between an image and the user’s perception of its quality. Many IQA approaches have achieved success using the idea of supercurrent neural networks (CNNs) . In addition, CNN-based IQA models are limited by fixed-size input requirements in batch learning, that is, input images that need to be cropped or resized to a fixed-size form.
To solve this problem, researchers from Google presented a ” Multiscale Image Quality Interpreter (MUSIQ)” to bypass CNN’s fixed-size input restrictions in order to predict effective image quality on images with the original resolution.
The paper was published at ICCV 2021 , where the model supports processing input full-size images with different resolutions and aspect ratios. It will also allow the extraction of multiscale features to capture image quality at different levels of detail.
“We apply MUSIQ to four large-scale IQA datasets, demonstrating consistent state-of-the-art results for three technical quality datasets (PaQ-2-PiQ, KonIQ-10k and SPAQ) and performance comparable to that of state-of-the-art -models in the AVA aesthetic quality dataset,” the study says.