2pm Tuesday, 11 January 2005:

<CENTER><B>
Reduced-Reference Image Quality Assessment
</B>

<P>
Zhou  Wang <BR>
NYU
</CENTER>

Reduced-reference (RR) image quality measures aim to predict the visual 
quality of distorted images with only partial information about the 
reference images. They are useful in a number of applications, e.g., 
real-time visual communication. I will talk about our recently proposed RR 
method, which is based on a natural image statistic model in the wavelet 
transform domain. Specifically, we use the Kullback-Leibler distance 
between the marginal probability distributions of wavelet coefficients of 
the reference and distorted images as a measure of image distortion. A 
generalized Gaussian model is employed to summarize the marginal 
distribution of wavelet coefficients of the reference image, so that only a 
relatively small number of RR features are needed for the evaluation of 
image quality. We find that many well-known types of image distortions lead 
to significant changes in wavelet coefficient histograms, and thus are 
readily detectable by our measure. Paper and software are available at 
http://www.cns.nyu.edu/~lcv/rriqa/. If time is allowed, I will also talk 
about how this method can be used to create a "quality-aware" image that is 
able to track its own quality while transmitted through distortion channels.