Which are the color
transforms that JPEG2000 define?
What do we mean by
scalability?
What do we mean by resolution
scalability?
What do we mean by distortion
scalability?
Which are the differences among
the different types of ROI coding?
How does the Maxshift method work?
Which are the color transforms that JPEG2000
define?
They are only two: RCT and ICT.
What is the RCT?
RCT means ‘Reversible Color Transform’.
It is a transform that maps an integer number domain into an integer
number codomain and it is used for Lossless compression.
It is divided into Forward RCT e Inverse RCT.
|
Forward RCT
|
Inverse RCT
|
What is the ICT?
ICT means ‘Irreversible Color Transform’; it is the same transform as
the conventional YCBCR used for image and video
representation.
Y=0.299(R-G)+G+0.144(B-G)
CB=0.654(B-Y)
CR=0.173(R-Y)
What do we
mean by scalability?
The meaning of scalability is
not easy to define without saying what scalability concerns.
As far as JPEG2000 is
concerned, it is possible to have scalability both in resolution and distortion
because of the used coding algorithms. This means that we can obtain a lower
memory requirement either by reducing the image resolution or by increasing the
distortion; but all this has to be done a posteriori, that is after the
generation of the bit-stream.
What do we mean by resolution
scalability?
A compressed bit-stream is scalable in resolution if it has some
identifiable elements that represent the subbands at every resolution
level, without any dependence on higher resolution levels.
What do we mean by distortion
scalability?
In order to have a distortion scalable bit-stream, inside it there must
be identifiable elements highlighting the quality of the compressed
representation in a progressive way and that allow to establish the aimed
distortion level a posteriori, thus allowing to re-organize the
bit-stream in a new embedded bit-stream.
What are the ROIs?
ROI stands for Region Of Interest.
The basic concept deals with the identification of the regions of the
image that are more important than the other ones, so they will be encoded with
a better quality.
There are two main methods to code the ROI into a JPEG2000 image:
·
Block based mode
·
Priorization by scaling
Which are the differences among the different types of ROI
coding?
‘Block based’ method consists of the possibility of increasing the
quality of some blocks that correspond to the most important region of interests.
This approach is not optimal for static ROI solutions, while it is more
useful for dynamic ones.
‘Priorization by scaling’ method consists of an increase of the priority
of the pixels of the region, so that they can own the most significant bits of
the code.
How does
the
Maxshift method work?
JPEG2000
is based on a technology that is able to code only the most important areas of
an image (ROI), which can be in arbitrary form. This system is called 'Maxshift method' and is able to achieve high levels
of compression but with the drawback of a loss in the quality of the
background.
Unlike other image coding
methods, JPEG2000 uses the DWT; if we scale appropriately the wavelet
coefficients of the ROI, so that the bits associated to it belong to a
bit-plane that is more significant compared to the background bit-planes (fig.
1 and 2), the information associated to it is sent for first. Therefore, by
changing the quantization and truncating the obtained code it is possible to
get different quality levels for the ROI and the background (thus sacrificing
the background image in order to obtain higher compression ratios without
compromising the image quality).
Fig. 1 Quantized wavelet
coefficients
Fig.2 Wavelet coefficients being
quantized after being scaled