Page 158 - ITU Journal, ICT Discoveries, Volume 3, No. 1, June 2020 Special issue: The future of video and immersive media

ITU Journal: ICT Discoveries, Vol. 3(1), June 2020

Disadvantages:
– Imaging applications using Recommendation T.81 and/or T.82 are not necessarily compatible;

– the T.80-Series Toolkit is flexible, but very comprehensive;
– the employment of the T.80-Series compression methods in concrete application (including
selection of proper parameters, defining resolutions, color models, interleave structure, pixel
aspect ratio, communication protocols for transmission, etc.) is still a major task.”

2.1 JPEG: An architecture for image compression

Pennebaker and Mitchell [2] describe the common components and the toolbox nature of the JPEG-1 format,
“JPEG is more than an algorithm for compressing images. Rather, it is an architecture for a set of image
compression functions. It contains a rich set of capabilities that make it suitable for a wide range of
applications involving image compression.

In one respect, however, JPEG is not a complete architecture for image exchange. The JPEG data streams
are defined only in terms of what a JPEG decoder needs to decompress the data stream. Major elements are
lacking that are needed to define the meaning and format of the resulting image. The JPEG committee
recognized that these aspects are quite controversial and would probably have delayed the decision-
making process needed to complete JPEG. They decided that, necessary as these parameters and constraints
are, they are more properly the domain of application standards. The committee therefore deliberately did
not include them in JPEG.”

2.2 JPEG baseline and extended systems
JPEG has defined a baseline capability that must be present in all JPEG modes of operation that use DCT, which
is the common core that enables easier interoperability among applications.
To ensure progressive image build up for certain applications, several modes are supported:
In the progressive DCT modes in the spectral selection mode, the DCT coefficients are grouped into spectral
bands, where for all 8 × 8 blocks the lower-frequency bands are sent first and then the higher frequency ones.
In the successive approximation, the information is first sent with lower precision and then refined in later
scans with higher precision data. In the hierarchical mode, the resolution of the image increases with the
progressing stages. The best type of progressivity to use depends entirely on the application (Fig. 11).
While the earlier modes provide lossy images that can be used for many applications and result in higher
compression rates, there are applications that require lossless image compression (e.g. for medical diagnosis
images). The compression rate there is obviously worse.

On the entropy coding level, Huffman coding is mandatory for all images; optionally, arithmetic coding can be
used. Essential characteristics of JPEG coding processes are shown in Fig. 12.

Fig. 11 – JPEG modes of operation [2]

136 © International Telecommunication Union, 2020