Block adaptive quantization

Details Block adaptive quantization Block adaptive quantization

1999-10-18

2001-07-03

Lobo, Ian J. (Department: 3662)

Communications: directive radio wave systems and devices (e.g.,

With particular circuit

Digital processing

C342S02500R, C342S194000, C342S195000

Reexamination Certificate

active

06255987

ABSTRACT:

BACKGROUND OF THE INVENTION
This invention relates to block methods and apparatus for the adaptive quantisation of data and in particular but not exclusively to such methods and apparatus for the block adaptive quantisation of radar data.
In a typical pulse compression radar system, radar data is captured, converted to digital form, compressed and downloaded to a ground station for reconstruction and analysis of the data. The data from a pulse compression radar is characterised by having a gaussian noise-like distribution and so normal compression techniques such as the JPEG format are unsuitable. Accordingly a different form of compression is required and Block Floating Power Quantisation (BFPQ) or Block Adaptive Quantisation (BAQ) is often used. It is used to compress the data generated by the reception chains so that the data rate is low enough to be handled by the subsequent logic circuits.
In a typical synthetic aperture radar using BAQ compression, sample pairs are compressed using a block adaptive quantisation algorithm The BAQ algorithm operates on successive blocks of data, of known length The exponent for each block of data is calculated by summing the total of both the samples of the block and determining an exponent value.
The sample data is divided by the exponent value and this is done by using look-up tables. However, in a typical application for the Advanced Synthetic Aperture Radar (ASAR) developed for the European Space Agency, a BAQ device had a block size of 63 sample pairs, and required a large number of look-up tables (256) to produce the exponent and the compressed digital values. Also the operating speed of the device was slow in view of the large number of look-up tables required.
SUMMARY OF THE INVENTION
We have developed a BAQ method and apparatus which employs a novel combination of look-up tables and bit shift for processing of the I and Q samples, which reduces the number of look-up tables required and results in a faster operation. Furthermore, in a conventional BAQ process the code tables are spaced linearly. This means that, on a fractional quantisation error curve, the curves are bunched together at the top end of the graph. In a preferred embodiment we provide a logarithmic distribution of the curves and the code tables, thereby giving a more uniform fractional error.
Accordingly in one aspect this invention provides a method for the block adaptive quantisation of data, which comprises processing a block of input data samples to derive an aggregate or average value, using said aggregate or average value to derive an exponent value (E), determining compressed values of said data samples from a plurality of divide look-up tables and applying a bit shift operation, the look-up table and bit shift each being selected in accordance with the selected exponent value (E), and outputting said compressed data.
Preferably said aggregate or average value is applied to an exponent look-up table to determine said exponent value.
Where the system is required to apply different degrees of compression, the exponent value (E) may be obtained from one of a number of look-up tables, with the look-up table to which the aggregate or average value is applied being selected in accordance with the degree of compression required.
Preferably the exponent value (E) is processed to give a remainder value (R) which is used to select one of said divide look-up tables, and a shift value (S) which is used to apply a variable bit shift operation to the data sample.
The bit shift operation may be applied to the data value output by said selected divide look-up table (after the division operation), or it may be applied to the data sample before application to the selected divide look-up table.
Advantageously, the divide look-up tables and the bit shift values together define a series of code tables which spans at least part of the dynamic range of the input samples, with the series of code tables being made up of a preset number of repeated groups of code tables, and the shift value (S) of the exponent value (E) being used to select one of said groups of code tables and the remainder value (R) being used to select one of said code tables in the group.
The code tables within each group and the groups of code tables themselves are preferably spaced logarithmically with respect to the dynamic range of the input data samples.
In the illustrated embodiment, the groups of code tables are spaced at substantially 6 dB, and each group compresses four code tables spaced at substantially 1.5 dB.
In one particular embodiment having four divide look-up tables, the exponent value is 5 bits wide and of the form (S
2
,S
1
,S
0
,R
1
,R
0
), wherein (S
2
,S
1
,S
0
) represents the shift value (S) and (R
1
,R
0
) represents the remainder value R.
Where the input data is supplied in the form of in phase and quadrature samples (In, Qn) respectively, and the exponent value (E) may be calculated according to the following formulae, according to the level of compression required:
(i) where compression to 2, 3 or 4 bit is required
E
=
INT
[
k
×
log
2
(
1
+
∑
N
=
1
LBAQ
⁢
(
&LeftBracketingBar;
I
n
&RightBracketingBar;
+
&LeftBracketingBar;
Q
n
&RightBracketingBar;
)
LBAQ
)
-
C
]
where k is a constant, LBAQ is the number of sample pairs in a block, and C is a respective constant dependent on the extent of compression required, or
(ii) where compression to 1 bit is required
E
=
INT
(
∑
N
=
1
LBAQ
⁢
(
&LeftBracketingBar;
I
n
&RightBracketingBar;
+
&LeftBracketingBar;
Q
n
&RightBracketingBar;
)
LBAQ
)
with In, Qn and LBAQ as defined above.
In another aspect, this invention provides a method for the block adaptive quantisation data, which comprises:
processing a block of data samples to obtain an aggregate or average value of said block;
using said aggregate or average value to obtain an exponent value;
deriving from said exponent value a first parameter and a second parameter;
applying to said data samples a division operation selected on the basis of one of said parameters, and a variable bit shift operation on the basis of the other of said parameters, to obtain compressed data values, and
outputting said compressed data values and said exponent value.
In yet another aspect this invention provides apparatus for applying block adaptive quantisation to a block of data samples, said apparatus including:
means for processing a block of input data samples to determine an aggregate or average value;
means for deriving from said aggregate or average value an exponent value;
compression means for applying to said data samples a division operation and a variable bit shift operation each selected in accordance with said exponent value, to obtain compressed data, and
output means for outputting said compressed data.
Said compression means preferably includes;
means for processing said exponent value to obtain a first parameter value and a second parameter value,
a plurality of divide look-up tables,
division means for applying said data samples to a divide look-up table selected in accordance with one of said parameters, and
bit shift means for applying a variable bit shift operation to said data in accordance with the other of said parameters.
Whilst the invention has been described above it extends to any of the inventive combination of the features set out above or in the following description.


REFERENCES:
patent: 5666121 (1997-09-01), Fang et al.
Monet et al., “Block Adaptive Quantization of Images,” IEEE Transactions on Communications, vol. 41, No. 2, Feb. 1993.*
Kwok et al., “Block Adaptive Quantization of Magellan SAR Data,” IEEE Transactions on Geoscience and Remote Sensing, vol. 27, No. 4, Apr. 1989.*
Owens et al., “Compression of Synthetic Aperture Radar Phase History Data Using Trellis Coded Quantization Techniques,” Proceedings of the International Conference on Image Processing, 0/1997.

Affiliated with

Also associated with

Rating

Block adaptive quantization does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Block adaptive quantization, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Block adaptive quantization will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2498377