Method and device for compressing data to be transmitted at constant bit rate
View Patent ↗The invention concerns a method of compressing and transmitting digital data, in particular images, at a constant bit rate, the method comprising the following steps: splitting input data into blocks of data, applying an orthogonal or bi-orthogonal transform ( 12 ) to each block, quantizing ( 14 ) the coefficients obtained by applying the transform, entropically coding ( 20 ) the quantized coefficients, temporarily saving the data resulting from coding in a buffer memory ( 18 ), removing a particular number of bits from the memory for each block of digital data, and adjusting the quantizing step by applying a control function ( 50 ) to obtain, after coding, a number of bits close to the number of bits removed from the memory for each block of data. The control function ( 50 ) takes account of the data bit rate measured for at least the preceding block.
1. A method of compressing and transmitting digital data, in particular images, at a constant bit rate, the method comprising:
Splitting input data into blocks of data,
Applying an orthogonal or bi-orthogonal transform to a current block,
Quantizing coefficients obtained by applying the transform,
Entropically coding the quantized coefficients,
Temporarily saving the data resulting from coding in buffer memory,
Removing a particular number of bits from the memory for each block of digital data, and adjusting the quantizing step by applying a control function to obtain, for the current block, after coding, a number of bits close to the number of bits removed from the memory for each block of data, said control function being based on measurement of the bit rate produced by the entropic coding, and said control function taking account of the data bit rate measured for at least a preceding block,
Wherein the coefficient obtained by applying the orthogonal or bi-orthogonal transform are ordered in order of decreasing moduli and quantizing step is obtained from the modulus of the coefficient whose rank is equal to the integer nearest the ratio of the target bit rate and a parameter which is estimated for each block based on at least one preceding block.
2. A method according to claim 1 , wherein the control function takes account of the bit rate measured for a plurality of successive blocks preceding the current block.
3. A method according to claim 1 , wherein said parameter referred to as β e (n), is estimated from the values β m (m) measured for a plurality b of preceding data blocks, said parameter β e (n) being calculated using the following equation:
β
e
(
n
)
=
∑
i
=
1
b
-
1
α
(
i
)
β
m
(
n
-
r
-
i
)
,
in which:
β e (n) is the estimate of the parameter β for the current data block,
β m (n) is the value of β measured for the block n,
r is the regulation delay (r≧1), and
α(i) is a weighting coefficient such that:
∑
i
=
0
b
-
1
α
(
i
)
=
1.
4. A method according to claim 3 , wherein the values β m (n) are calculated using the following equation:
β
m
=
8
P
(
n
)
M
(
n
)
,
in which P(n) is the number of bytes obtained after entropically coding the block of data of index n, and
M(n) is the rank of the coefficient from which the quantizing step is deduced in the set of ordered coefficients supplied by the orthogonal or bi-orthogonal transform for the block n, said rank M(n) being obtained using the following equation:
M
(
n
)
=
NINT
(
8
Rc
′
(
n
)
β
e
(
n
)
)
in which Rc′(n) is the number of bytes assigned to the block n and NINT is the function that rounds up or down to the nearest integer.
5. A method according to claim 4 , wherein the bit rate Rc′(n) assigned to the block n is calculated as a function of the last measurable level in the buffer memory.
6. A method according to claim 5 , wherein the bit rate Rc′(n) is determined using the following equation:
Rc
′
(
n
)
=
Rc
+
N
O
-
N
b
(
n
-
r
)
T
,
in which Nb(n) is the level of the buffer memory after coding the data block n,
Rc is the number of bytes to be removed from the memory for each block,
N 0 is the setpoint level and the initial level of the buffer memory,
T is a time constant, and
r is a positive integer corresponding to a regulation delay.
7. A method according to claim 6 , wherein the constant T is from 1 to 10.
8. A method according to claim 3 , wherein the number b and/or the values α(i) are fixed from one block to another.
9. A method according to claim 3 , wherein the number b and/or the values α(i) vary from one block to another as a function of statistical properties calculated from known values of β m .
10. A method according to claim 9 , wherein the number b and/or the coefficients α(i) vary as a function of the correlation coefficient between known values of β m .
11. A method according to claim 1 , wherein the digital data consists of image data and each block of data corresponds to an entire image or to a particular number of lines of said image, for example eight lines.
12. The method according to claim 11 , wherein the input data comprises images of terrestrial areas acquired continuously by sensors on board an Earth observation satellite.