Best Paper Image Quality and Sysem Performance XIX
(Electronic Imaging Symposium 2022)
Patch-based CNN Model for 360 Image Quality Assessment with Adaptive Pooling Strategies,
Abderrezzaq Sendjasn, University of Poitiers and Norwegian University of Science and Technology (NTNU)
360° image quality assessment using deep neural networks is usually designed using a multi-channel paradigm exploiting possible viewports, mainly due to the high resolution of such images and the unavailability of ground truth labels (subjective quality scores) for individual viewports. The multi-channel model is trained to predict the score of the whole 360° image. However, this comes with a high complexity cost as multi neural networks run in parallel. This talk discusses patch-based training. To account for the non-uniformity of quality distribution of a scene, a weighted pooling of patches’ scores is applied. The latter relies on natural scene statistics in addition to perceptual properties related to immersive environments.
Abderrezzaq Sendjasn is currently a third year PhD candidate in signal and image processing at the University of Poitiers, France, and NTNU, Norway. He received his MSc in computer science from the University of Oran 1 in Algeria in 2017. His research interests include image processing, human visual perception, image quality assessment, 360-degree images, immersive application, and deep learning.