Databases Reference
In-Depth Information
13-bit template
10-bit template
A1 O
O
A2
O
O
OO
O
O
O
X
O
O
O
X
O
O
O
OOO
OO
Refined
Bitmap
Reference
Bitmap
Refined
Bitmap
Reference
Bitmap
F I GU R E 7 . 15
Contexts used in the generic refinement decoding procedure.
4 fax standards or a variation of the technique used to encode the lowest-resolution layer in
the JBIG recommendation. The first procedure, MMR algorithm was described earlier in this
chapter (Section 7.7.2). The second procedure, called
typical prediction
, is similar to the
procedure used for JBIG as described earlier.
The adaptive arithmetic coding used in JBIG2 uses the same algorithm as that used in
JBIG with four templates available for generating the context. There is a 16-bit template
which contains four variable locations, a 13-bit template with one variable location, and the
two 10-bit templates used for the encoding of the pixels in the JBIG bottom layer.
The generic refinement decoding procedure assumes the existence of a
reference
layer and
decodes the segment datawith reference to this layer. The standard leaves open the specification
of the reference layer. The decoding of the
refined
layer is done in raster scan order using
both the reference bitmap and the portions of the refined layer that have already been decoded.
The decoding uses context adaptive arithmetic coding where the context consists of already
decoded bits from the refined bitmap as well as bits from the reference bitmap. The two
templates used in the refinement decoding procedure are shown in Figure
7.15
. In Figure
7.15
,
the pixel marked
X
is the pixel being decoded while the shaded pixel in the reference bitmap
is the pixel in the same spatial location as the pixel being decoded. The pixels marked
A
1 and
A
2 are the adaptive pixels. The
A
1 pixel location can be anywhere within 128 pixels of the
pixel being decoded in the vertical and horizontal directions within the set of already decoded
pixels. If we denote the
X
pixel location as (0,0) the
A
2 pixel location can be anywhere between
(-128,-128) to (127,127) in the reference bitmap.
The generic refinement procedure can also use a typical prediction mode. In the generic
refinement procedure a pixel is said to be a typical pixel if the 3
3 block of pixels in the
reference bitmap centered on the location of the pixel being decoded all have the same value
and the pixel being decoded also has that value. If every pixel in a line corresponding to a 3
×
3
block of identical valued pixels in the reference bitmap is a typical pixel, the typical pixel flag
LTP
is set to one. When the typical pixel flag is set to one, the pixels corresponding to 3
×
3
blocks of identically valued pixels are not encoded, and the decoder simply copies that value
from the reference bitmap.
×
