@article{Dash_Mishra_Wong_2016, title={Deep Quality: A Deep No-reference Quality Assessment System}, volume={2}, url={https://openjournals.uwaterloo.ca/index.php/vsl/article/view/95}, DOI={10.15353/vsnl.v2i1.95}, abstractNote={<p>Image quality assessment (IQA) continues to garner great interest<br />in the research community, particularly given the tremendous<br />rise in consumer video capture and streaming. Despite significant<br />research effort in IQA in the past few decades, the area of noreference<br />image quality assessment remains a great challenge and<br />is largely unsolved. In this paper, we propose a novel no-reference<br />image quality assessment system called Deep Quality, which leverages<br />the power of deep learning to model the complex relationship<br />between visual content and the perceived quality. Deep Quality<br />consists of a novel multi-scale deep convolutional neural network,<br />trained to learn to assess image quality based on training samples<br />consisting of different distortions and degradations such as blur,<br />Gaussian noise, and compression artifacts. Preliminary results using<br />the CSIQ benchmark image quality dataset showed that Deep<br />Quality was able to achieve strong quality prediction performance<br />(89% patch-level and 98% image-level prediction accuracy), being<br />able to achieve similar performance as full-reference IQA methods.</p>}, number={1}, journal={Journal of Computational Vision and Imaging Systems}, author={Dash, Prajna Paramita and Mishra, Akshaya and Wong, Alexander}, year={2016}, month={Oct.} }