|
|
|
|
Overview
Superior quality for the same file size
JPEG2000 is a wavelet based image compression system employing DWT (Discrete Wavelet
Transform) and other sophisticated signal processing technologies to achieve its
compression. These 'wavelets' are partially representative of, and applied to, the
whole image resulting in only a slight 'softening' of details, even at very high
compression levels. By contrast, JPEG images are processed in blocks of 8x8 pixels
and compressed using a DCT (Discrete Cosine Transform). At high compression levels,
signal degradation is such that the DCT blocks no longer blend into one another,
and a mosaic effect can be observed in the decompressed image - sometimes referred
to as "JPEG artefacts". These artefacts are unacceptable for quality sensitive applications
such as pre-press, medical imaging, and biometrics severely hindering JPEG's suitability
for high level applications. JPEG2000 instead allows for visually lossless compression
at ratios often far surpassing those of JPEG - the very best of all worlds.
Smaller files at comparable image quality
The compression ratios that are achievable with JPEG2000 are nothing short of incredible.
This is due partly to the inherent efficiency of state-of-the-art wavelet image
compressors, and partly to the excellent pre and post processing applied by the
JPEG2000 standard before and after the wavelet transform itself.
These are 100% actual size crops from a larger image (856kb) |
Lossless compression mode
For applications demanding
perfect image reproduction on decompression, JPEG2000 supports lossless compression.
Compression ratios of up to 1:12 are still achievable, and as such, JPEG2000 offers
one of the best lossless compression systems in the industry, outperforming the
best known classical compression schemes such as LZW. Lossless compression is also
exceptionally powerful when used in conjunction with Regions Of Interest, as detailed
below.
User definable ROIs (Regions Of Interest)
Regions Of Interest are one of the most exciting features of the JPEG2000 specification.
They allow areas of an image containing high detail to be compressed either losslessly
or at a higher quality setting than those of less detail. For instance, an image
depicting a Rose bloom in sharp focus may be compressed using JPEG2000 so the bloom
is 'visually lossless', while the background which is out of focus and devoid of
detail is compressed much more heavily. ROIs may be rectangular, circular, or indeed
any closed-path complex shape. During progressive rendering of streamed JPEG2000
data (covered later in this document), image detail present in ROIs is provided
first, and rendered before, all other image information. This has the effect of
ensuring significant details in the image appear first, maintaining the viewers
interest while the remainder of the image is streamed and rendered.
Full support for ICC and CIE Lab colour profiles and non-RGB images
JPEG was designed only to support RGB (3 component) images. This obviously makes
it unsuitable for any application employing non-RGB images, with the printing industry
requiring storage of 4 colour component colour separated images in CMYK format being
a prime example. JPEG2000 solves this dilemma by allowing for an almost unlimited
number of components within a single image file, each of which may have its own
standard or proprietary definition within the context of the image it resides in.
Components may in turn be associated with ICC or CIE Lab colour profiles, allowing
for the first time accurate and predictable colour matched results across platforms
and applications.
Layered code stream format
The JPEG2000 code stream is a layered format allowing tremendous flexibility in
the way an image is decompressed from source. For instance, an image corresponding
to 1:10 compression may be extracted from a losslessly compressed image on a server
by only transmitting and decompressing 1/10th of the data. This allows images to
be accessed at a quality and complexity best suited to the user's individual requirements.
In addition, the compression process allows lossy image data to be ordered by resolution,
quality, colour and spatial location within the image. This determines progressive
rendering behaviour during transmission.
Greater resilience to data transmission errors
JPEG2000 provides extensive error resilience as a fundamental part of the image
code stream specification. Special "packet markers" may be emitted by encoders,
allowing for excellent error recovery by decoders when parsing code streams containing
random bit errors. In contrast, the majority of inconsistencies that may be encountered
in a JPEG code stream are unrecoverable and effectively destroy all remaining image
data following the error.
Extensive support for embeddable metadata within images
Finally, JPEG2000 makes extensive provision for rich metadata support within images
stored in the JP2 file format. The following metadata types may be stored within
the image file:
 IP (Intellectual Property) - Image information such as image
creator information and copyright details
IPTC
(International Press Telecommunications Council) - Image
information conforming to the IPTC standard
UUID (Universal Unique Identifier) - Vendor specific image information. · XML (eXtensible
Markup Language) - Application defined XML metadata
These metadata capabilities may be utilized by digital archives, image banks, and
proprietary applications alike, allowing image files to store far more than simply
raster data.
|
|
|
|
 |
|