Amount Awarded: $20,000
The last few decades have seen an increase in studies focused on image and video quality assessment resulting in the introduction of metrics that show a high correlation with the perceptual judgement of the human observers. Nevertheless, keeping in mind that unlike other computer vision and image processing tasks the use of Convolutional Neural Networks (CNNs) has been limited in this field of research there exists a huge room for improvement. This lack of CNN based image and video quality metrics is mostly linked to the absence of any large-scale subjective dataset in the field. To answer this issue, this project will focus on launching the collection of a large-scale subjective dataset for image and video quality assessment. During the project we plan to organise workshops focused on large-scale subjective data collection, human perception, and the use of CNNs in image and video quality assessment both at UC Berkeley and the Norwegian University of Science and Technology. This project will also be a great opportunity for Dr. Amirshahi to advance his career and be mentored by Professor Olshausen.