Structural Approaches to Image Quality Assessment
Appears in:
Ch. 8.3 of Handbook of Image and Video Processing, 2nd edition
pages 961--974.
ed. Alan Bovik, Academic Press, 2005.
In this chapter, we describe the basic concepts and algorithms of the
structural approach to image quality assessment. Unlike conventional
bottom-up image quality assessment approaches that attempt to simulate
the functions of the relevant components in the early visual system,
the structural approaches follow a top-down philosophy that is based
on hypothesized functionality of the overall visual system. We
demonstrate that image distortions along different directions in the
image space have different perceptual meanings. The structural
approach attempts to separate the directions associated with
non-structural distortions (those that do not change the structures of
the objects in the visual scene) from the remaining structural
distortions. This separation must be adapted to the local image
content. As a specific example, we discuss in detail the structural
similarity (SSIM) index and demonstrate its effectiveness using an
image synthesis method that maximally differentiates two competing
image quality metrics.
Download:
Preprint (pdf)
| Online Publications