Quantifying image quality in computer graphics

Lead Research Organisation: Bangor University
Department Name: Sch of Computer Science

Abstract

A crucial problem in computer graphics is validating results. It is difficult to prove with all scientific rigor that an image produced by a new algorithm is statistically significantly better than other state-of-the-art results. This project is indented to provide means of comparing computer graphics results in a possibly effective and accurate way. A major deliverable of this project is to develop the first computational image quality metric suitable for use with computer graphics applications. The metric is intended to replace tedious user studies with a computational algorithm capable of estimating quality loss due to approximated solutions. Such a metric will have a significant impact on both academic researchers and industries working in the field. It will enable validation studies to be more robust and far cheaper to carry out.

Planned Impact

The proposed research will be directly beneficial to the imaging industry (cameras, displays, video and image processing, graphics cards), university students and the Research Institute of Visual Computing. The biggest beneficiary of this project will be the imaging industry, and indirectly consumers, through better imaging products. The imaging industry will be able to directly use the metric to design and test new products both cheaper and faster. We attach to this proposal a letter of support from Dolby, who has a strong interest in applying objective quality metrics in their video-related projects. The university students will benefit from the newly established group by taking courses that teach on quality assessment. Quality assessment is mostly missing from curricula of graphics and visualization students, but it is a key knowledge necessary to prove with scientific rigor that one image is better than the other. We plan to enrich the current portfolio of visual computing courses in Bangor to include the fundamental topics of quality assessment. This project will also strengthen the newly established Research Institute of Visual Computing (RIVIC) - http://www.rivic.org.uk, which is a collaborative network of four universities in Wales: Bangor, Aberystwyth, Cardiff and Swansea. Since quality issues are relevant for many problems that the researchers in RIVIC are dealing with, the expertise in this area will be beneficial for the entire institute.

Publications


10 25 50
Cadík M (2013) Learning to Predict Localized Distortions in Rendered Images in Computer Graphics Forum
Eilertsen G (2013) Evaluation of Tone Mapping Operators for HDR-Video in Computer Graphics Forum
Hadziabdic K.K. (2013) Comparison of deghosting algorithms for multi-exposure high dynamic range imaging in Proceedings - SCCG 2013: 29th Spring Conference on Computer Graphics
Mantiuk R (2011) HDR-VDP-2
Mantiuk R (2012) Comparison of Four Subjective Methods for Image Quality Assessment in Computer Graphics Forum
Petit J (2013) Assessment of video tone-mapping: Are cameras' S-shaped tone-curves good enough? in Journal of Visual Communication and Image Representation
Trentacoste M (2012) Unsharp Masking, Countershading and Halos: Enhancements or Artifacts? in Computer Graphics Forum
Trentacoste M (2012) Unsharp Masking, Countershading and Halos: Enhancements or Artifacts? in Computer Graphics Forum
 
Description The developed image quality metric for high dynamic range images (HDR-VDP-2) is widely used in both MPEG and JPEG communities to test new video and image compression methods for high dynamic range content.
First Year Of Impact 2011
Sector Digital/Communication/Information Technologies (including Software)
Impact Types Economic
 
Description West Pomeranian University of Technology
Amount £40,713 (GBP)
Funding ID Nr N N516 508539 
Organisation West Pomeranian University of Technology 
Sector Academic/University
Country Poland, Republic of
Start 11/2010 
End 03/2013
 
Description West Pomeranian University of Technology
Amount £40,713 (GBP)
Funding ID Nr N N516 508539 
Organisation West Pomeranian University of Technology 
Sector Academic/University
Country Poland, Republic of
Start 11/2010 
End 03/2013
 
Title HDR-VDP-2: image quality and visibility metric for high dynamic range images 
Description HDR-VDP is a visual metric that compares a pair of images (a reference and a test image) and predicts: Visibility - what is the probability that the differences between the images are visible for an average observer; Quality - what is the quality degradation with the respect to the reference image, expressed as a mean-opinion-score. The metric can be used for testing fidelity (e.g. how distracting are image compression distortions), or visibility (is the information sufficiently visible). 
Type Of Technology Software 
Year Produced 2011 
Open Source License? Yes  
Impact The developed image quality metric for high dynamic range images (HDR-VDP-2) is widely used in both MPEG and JPEG communities to test new video and image compression methods for high dynamic range content. 
URL http://hdrvdp.sourceforge.net/wiki/