Image analysis – Image compression or coding – Pyramid – hierarchy – or tree structure
Reexamination Certificate
1998-11-27
2001-12-18
Johnson, Timothy M. (Department: 2623)
Image analysis
Image compression or coding
Pyramid, hierarchy, or tree structure
C375S240190, C382S261000
Reexamination Certificate
active
06332043
ABSTRACT:
TECHNICAL FIELD
This invention relates to a data processing apparatus for processing digital data using wavelet transform/inverse wavelet transform. More particularly, it relates to a data encoding method and apparatus, a data decoding method and apparatus and a recording medium which may be applied to an acoustic video equipment, communication equipment or a database apparatus for encoding aimed at data compression of speech or pictures or decoding the wavelet encoding.
BACKGROUND ART
FIG. 21
shows a basic structure for band splitting by wavelet transform and synthesis by inverse wavelet transform. An input here is a one-dimensional signal x[i].
Referring to
FIG. 21
, a wavelet transform unit
2
splits the one-dimensional signal x[i] entering an input terminal
1
into signals of plural frequency bands (sub-band signals). An inverse wavelet transform unit
4
synthesizes the signals xa′[j], xb′[j], xc′[j], . . . , split into sub-bands, in order to restore the input signal x[j]. A signal processor
3
processes data split into plural frequency bands in a pre-set manner. For example, if the signal processor
3
executes the encoding, it executes quantization, variable length encoding, transmission, variable length decoding and dequantization.
That is, in the wavelet transform unit
2
, each of analysis filters
11
,
12
,
13
, . . . fed with the one-dimensional signal x[i] splits the one-dimensional signal x[i] entering the input terminal
1
into plural frequency bands. The analysis by these analysis filters
11
,
12
,
13
, . . . gives data strings of respective frequency bands xa[i], xb[i], xc[i], . . . , which are sent to respective downsampling units
21
,
22
,
23
, . . . . These downsampling units
21
,
22
,
23
execute downsampling of saving data of the data strings of respective frequency bands xa[i], xb[i], xc[i], . . . associated with the sampling intervals Dt (t=
1
,
2
,
3
, . . . ) while thinning out the remaining data. The data xa′[i], xb′[i], xc′[i], . . . of the respective frequency bands, obtained on downsampling by the downsampling units
21
,
22
,
23
, . . . are sent to the signal processor
3
for predetermined signal processing.
The data strings of the respective frequency bands, processed in a predetermined manner by the signal processor
3
, are sent to the inverse wavelet transform unit
4
which then sends the input data strings of the input respective frequency bands to upsampling units
41
,
42
,
43
, . . . associated with the respective frequency bands. These upsampling units
41
,
42
,
43
, . . . pad the interval between two temporally neighboring data with a suitable number of zeros. The number of inserted zeros is equal to the number of samples (Dt-
1
) of data thinned out by the downsampling units
21
,
22
,
23
, . . . associated with these upsampling units
41
,
42
,
43
, . . . . This substantially restores the data strings xa[i], xb[i], xc[i], . . . of the respective frequency bands. The data strings xa[i], xb[i], xc[i], . . . of the respective frequency bands, resulting from these upsampling operations, are sent to associated synthesis filters
51
,
52
,
53
, . These synthesis filters
51
,
52
,
53
, . . . perform interpolation processing on the supplied data strings xa[i], xb[i], xc[i], . . .. Output data of the synthesis filters
51
,
52
,
53
, . . . are sent to an adder
6
which then sums these output data to restore the one-dimensional signal x[i] as a synthesized output signal X″[i] which is outputted at an output terminal
5
.
An embodiment for splitting the input one-dimensional signal into two sub-bands is explained specifically. Meanwhile, there are provided two each of the analysis filters and downsampling units of the wavelet transform unit and two each of the upsampling units and synthesis filters of the inverse wavelet transform unit.
In this case, the two analysis filters
11
,
12
of the wavelet transform unit are a low-pass filter and a high-pass filter. The two analysis filters, that is the analysis low-pass and high-pass filters
11
,
12
, band-split the input one-dimensional signal x[i] into a data string XL[i] of the low frequency band and a data string XH[i] of the high frequency band, respectively. Also, the downsampling units
21
,
22
decimate the band-split data strings XL[i] and XH[i] every other ample in order to find down-sampled data strings XL[j] and XH[j] of the two frequency bands, as shown by the following equations (1) and (2):
XL[j]=XL[i], where j=i/2 . . . (1)
XH[j]=XH[i], where j=i/2 . . . (2)
In the inverse wavelet transform unit, the signal-processed data strings of the two frequency bands have their sample intervals expanded by two by upsampling units
41
,
42
. In addition, zero-valued samples are inserted at the center positions of the data strings of the two bands. This upsampling operation may be represented by the following equations (3) and (4):
XL
[
i
]
=
{
XL
[
j
]
i
=
2
×
j
0
i
=
2
×
j
+
1
(
3
)
XH
[
i
]
=
{
XH
[
j
]
i
=
2
×
j
0
i
=
2
×
j
+
1
(
4
)
The data strings XL[i], XH[i] of the respective frequency bands, obtained on upsampling by the upsampling units
41
,
42
, are sent to a low-pass filter for synthesis
51
and a high-pass filter for synthesis
52
, associated with the low-pass filter for analysis
11
and the high-pass filter for analysis
12
, respectively. The low-pass filter for synthesis
51
and the high-pass filter for synthesis
52
interpolate the output data strings of the upsampling units
41
,
42
. The data strings XL[i], XH[i] of the respective frequency bands are summed together by an adder
6
(specifically an adder summing two data strings together) in order to restore the input signal x[i] as the synthesis output signal X″[i].
The low-pass filter for analysis
11
and the high-pass filter for analysis
12
on the side of the wavelet transform unit
2
and the low-pass filter for synthesis
51
and the high-pass filter for synthesis
52
on the side of the inverse wavelet transform unit
4
are configured for completely or approximately satisfying the relation of the following equations (5) and (6):
H
0
(-z)F
0
(z)+H
1
(-z)F
1
(z)=0 (5)
H
0
(z)F
0
(z)+H
1
(z)F
1
(z)=2z-L . . . (6)
In the above equations (5) and (6), H
0
(z), H
1
(z), F
0
(z) and F
1
(z) are transfer functions of the low-pass filter for analysis
1
1
, high-pass filter for analysis
12
, low-pass filter for synthesis
51
and the high-pass filter for synthesis
52
, respectively, L being an optional integer. By these constraint conditions, it can be assured that, if the input data is infinitely long, the output signal X″[i] of the adder
6
in the inverse wavelet transform unit
4
can completely or approximately coincide with the input signal x[i].
FIG. 22
shows an example of filter coefficients of the low-pass filter for analysis
11
, high-pass filter for analysis
12
, low-pass filter for synthesis
51
and the high-pass filter for synthesis
52
, associated with one another.
If the above-mentioned wavelet splitting/synthesis is used for encoding, encoding/decoding is carried out between the downsampling units
21
,
22
and the upsampling units
41
,
42
of FIG.
21
.
FIGS. 23 and 24
show a conventional structure of an encoding device and a decoding device for a one-dimensional data string employing wavelet transform.
In an encoding device
60
, shown in
FIG. 23
, an input signal x[i] at an
Chang Dexter T.
Frommer William S.
Frommer Lawrence & Haug LLP.
Johnson Timothy M.
Sony Corporation
LandOfFree
Data encoding method and apparatus, data decoding method and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data encoding method and apparatus, data decoding method and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data encoding method and apparatus, data decoding method and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2559664