Abstract
A perceptual image compression codec exploits the characteristics of the
human senses to minimize the perceivable quality loss of digital images
under compression. Such a codec has an even higher value if the
resulting codestreams are compatible to an existing standard, and are
thus decodable by all-day, existing applications. This work will first
describe three basic mechanisms perceptual coding is based on today,
followed by strategies how to implement them in standardized
environments, namely JPEG, JPEG 2000 and JPEG-XR. Following that,
strategies to evaluate the success of perceptive coding are discussed,
namely subjective measurements and objective quality metrics. Finally,
the circle is closed back to compression codecs by showing on the
example of JPEG 2000 and SSIM that quality metrics can also be used to
drive the rate-allocation of a codec, and hence explore the quality
judgment of a metric directly.
Users
Please
log in to take part in the discussion (add own reviews or comments).