JPH0779251B2 - Subband filter and subband coding method - Google Patents

Description

The present invention relates to a subband coding system used for digital transmission of high-definition image signals, and an input digital signal in a plurality of band signals in the subband coding system. It relates to a subband filter that is divided into.

[Conventional technology]

The subband coding method is a method in which an input signal is divided into a plurality of band signals by a subband filter (band division filter), and each band signal is independently coded by an optimum method and transmitted. The coding method may be, for example, differential PCM (DPCM) or other predictive coding method, or DCT.
(Discrete Cosine Transform) Other transform coding methods are used.

A conventional sub-band filter used for band division in such a sub-band coding method is a quadrature mirror filter (QMF: Quadrarure Mirror
lter) or perfect reconstruction low-order filter (SSKF: S
ymmetric Short Kemel Filter) is used.

FIG. 5 is a block diagram showing a configuration example of a conventional subband filter using a second-order perfect reproduction type filter.

In the figure, an input signal (X (t)) 51 is a digital signal converted by an analog / digital converter (A / D converter) not shown. When the 8-bit A / D converter is used, the input signal (X (t)) 51 takes an integer value in the range of 0 to 255.

The input signal (X (t)) 51 is a multiplier 52 ₁ that multiplies a coefficient of 0.5.
Input to the input signal (X
(T-1)) is input to the multiplier 52 ₂ which multiplies the coefficient 0.5. The outputs of the multipliers 52 ₁ and 52 ₂ are added by the adder 54 ₁ and taken out as a low frequency signal (L (t)) 55. That is, the multipliers 52 ₁ , 52 ₂ , the delay circuit 53, and the adder 54 ₁ output coefficient values (0.5, 0.5) for outputting the low-frequency signal (L (t)) 55.
A 0.5) low pass filter (LPF) 56 is constructed.

Further, the input signal (X (t)) 51 is input to a multiplier 52 ₃ for multiplying the factor of 0.5, further input signal via a delay circuit 53 (X (t-1) ) is multiplied by a coefficient -0.5 multiplication Input to the device 57. The outputs of the multipliers 52 ₃ and 57 are added by the adder 54 ₂ and taken out as a high frequency signal (H (t)) 58. That is, the multipliers 53 ₃ and 57, the delay circuit 53 and the adder 54 ₁ output the coefficient value (0.
5, -0.5) high-pass filter (HPF) 59 is configured.

Such a second-order perfect reproduction type sub-band filter is
Two consecutive input signal sequences X (t) and X (t-
1) From L (t) = 0.5 × {X (t) + X (t−1)} H (t) = 0.5 × {X (t) −X (t−1)} The L (t)) 55 and the high frequency signal (H (t)) 58 are divided and output.

Although not shown, each signal is down-sampled (decimated) to 1/2, independently encoded, and transmitted to the transmission line.

On the reception (decoder) side, restoration processing is performed by the subband synthesis filter corresponding to the transmission (coder) side subband filter.

FIG. 6 is a block diagram showing a configuration example of a conventional subband synthesis filter.

In the figure, a low-frequency signal (L (t)) 61 that has undergone decoding processing corresponding to encoding on the transmission side is represented by multipliers 62 ₁ , 62 with a coefficient of 1.0.
Entered in ₂ . Similarly, the decoded high frequency signal (H
(T) 63 is input to the multiplier 62 ₃ having a coefficient of 1.0 and the multiplier 64 having a coefficient of −1.0. The outputs of multipliers 62 ₁ and 62 ₂ are adders
The outputs of the multipliers 62 ₂ and 64 are added by 65 ₁ and added by the adder 65 ₂ and input to the switch circuit 66. The switch circuit 66 switches at a predetermined timing and multiplexes two signals to obtain an output signal 67.

Such a sub-band synthesis filter is used for high frequency signals (H
(T)) and the low-frequency signal (L (t)) X (t) = L (t) + H (t) X (t-1) = L (t) -H (t) This is a configuration for reproducing X (t) and X (t-1).

[Problems to be Solved by the Invention]

As for the sub-band coding method, the optimum coding method can be selected independently for each band signal, so if the visually important components are concentrated in the low-frequency signal such as an image signal, that characteristic is used. As a result, the coding bit rate of the high frequency signal is reduced and the overall coding bit rate can be reduced.

By the way, in order to realize high-quality coding, it is necessary to reduce the coding distortion of each band signal as much as possible, but if a code code is assigned to each level that the band signal can take, it will be completely distorted in coding. It is known that transmission can be performed without causing

In addition, at this time, a short code to a high frequency level,
If the entropy coding that allocates a long code to a level with low occurrence frequency is applied, the coding bit rate can be reduced.

On the other hand, when the conventional sub-band filter is used, even if the number of levels that can take an input signal is 256 (0 to 255), each band signal (low band signal (L (t)) and high band signal ( The number of levels that H (t) can take is 511 (0 to 255 (0.
5 increments), -127.5 to 127.5 (0.5 increments)), which is double the size.

Therefore, when a code code is assigned to each possible level of each band signal, entropy is large, and even if the coding bit rate is reduced according to the bias of the signal distribution, the effect is canceled out. There was a point.

It is an object of the present invention to provide a subband filter and a subband coding method that reduce the entropy of a band signal to be separated and that can reduce the overall coding bit rate.

[Means for Solving the Problems]

FIG. 1 is a block diagram showing the basic configuration of the present invention.
Incidentally, FIG. 1 (a) is a basic configuration of the subband filter according to claim 1 and claim 2, and FIG.
Is a basic configuration of the subband coding method according to claims 3 and 4.

The invention according to claim 1 has a second-order low-pass filter whose coefficient value is (0.5, 0.5) and a second-order high-pass filter whose coefficient value is (0.5, -0.5). , In a sub-band filter that divides an input digital signal of a predetermined integer value into low-band and high-band signals and outputs, when each band signal has a non-integer value, at least one band signal is A band signal entropy reducing means for converting into a predetermined integer value is provided and configured.

The invention according to claim 2 is the subband filter according to claim 1, wherein the band signal entropy reducing means converts at least one band signal into a predetermined non-integer value when each band signal has an integer value. This is the configuration.

The invention according to claim 3 has a second-order low-pass filter whose coefficient value is (0.5, 0.5) and a second-order high-pass filter whose coefficient value is (0.5, -0.5). , A sub-band filter that divides an input digital signal of a given integer value into low-band and high-band signals and outputs, and each band signal is encoded by a corresponding encoding method, multiplexed, and transmitted Demultiplexing the encoding means and the received signal,
Decoding means for decoding in accordance with the encoding method and outputting as low band and high band signals, and each decoded band signal are taken in and a digital signal is reproduced by processing corresponding to the sub-band filter. In the sub-band coding method with a sub-band synthesis filter, the sub-band filter, when each band signal takes a non-integer value,
The sub-band synthesis filter includes a band signal entropy reducing unit that converts one of the band signals into a predetermined integer value and outputs the band signal to the encoding unit. When it takes an integer value, it comprises band signal restoration means for converting one band signal to the original non-integer value.

According to a fourth aspect of the invention, in the subband coding method according to the third aspect, the band signal entropy reducing means of the subband filter outputs one of the band signals when each band signal has an integer value. It is configured to convert to a predetermined non-integer value and output to the encoding means, and the band signal restoring means of the sub-band synthesis filter, when the other band signal for one non-integerized band signal takes an integer value. In addition, one of the band signals is converted to the original even value.

[Work]

In the subband filter of the present invention, the band signal entropy reducing means converts the non-integer value of at least one of the band signals into a predetermined integer value or the integer value into a predetermined non-integer value.

That is, at least one band signal can take only an integer value or a non-integer value, the number of levels can be reduced by half in encoding, and entropy can be reduced.

By the way, the input signal sequence X (t) of the subband filter
And one of X (t-1) is even and the other is odd, the low-frequency signal (L
(T)) and the high frequency signal (H (t)) are always non-integers.

When the input signal sequences X (t) and X (t-1) are both even or odd, the low-frequency signal (L (t)) and high-frequency signal (L (t)) corresponding to the sum and difference thereof are H
Both (t) are always integers.

In the sub-band coding method of the present invention, the band signal entropy reducing means of the sub-band filter converts all one band signal into an integer value or a non-integer value for coding. In the band signal restoration means of the sub-band synthesis filter, the other band signal transmitted in the same state is monitored, the above-mentioned property is used, and one band which is made into an integer or a non-integer according to its numerical state is used. Perform the operation to return the signal to the original value.

That is, the original band signal can be accurately reproduced in decoding, and one band signal can take only an integer value or a non-integer value, so that the number of levels can be reduced by half in encoding. Entropy can be reduced.

Embodiments Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of an embodiment of the subband filter of the present invention.

In the figure, the configuration of a low-pass filter and a high-pass filter surrounded by a chain line, its input signal (X (t)) 51,
Further, the low-frequency signal (L (t)) 55 and the high-frequency signal (H (t)) 58 that are output are the same as in the conventional configuration shown in FIG.

The band signal entropy reducing means provided in the subband filter of the present embodiment, when the low frequency signal (L (t)) 55 and the high frequency signal (H (t)) 58 are non-integer, It is characterized by adding 0.5 or −0.5 to L (t) 55 to make it an integer.

By the way, when one of the input signal sequences X (t) and X (t−1) is an even number and the other is an odd number, the low-frequency signal (L
Both (t)) 55 and high frequency signal (H (t)) 58 are integers. Therefore, here, the non-integer value identification circuit 21 detects whether or not the high frequency signal (H (t)) 58 is a non-integer, and outputs 0.5 or -0.5 when it becomes a non-integer. , 0 is output when it becomes an integer. The adder 23 adds the output value of the non-integer value discriminating circuit 21 to the non-integer value of the low frequency signal (L (t)) 55 to obtain the low frequency signal (L '(t)) 25 which takes only the integer value. Convert.

Note that either 0.5 or −0.5 may be added when converting the low-frequency signal (L (t)) 55, which takes a non-integer value, into an integer. Either one may be added in advance to match the subband synthesis filter. Set it.

As described above, by providing the band signal entropy reducing means in the sub-band filter, the low band signal (L (t)) of each band signal can be only an integer value in this embodiment. Therefore, the number of levels that the low frequency signal (L (t)) can take is 512/2 = 256, which can be reduced to about half of the conventional level number 511 of the low frequency signal (L (t)). , It is possible to significantly reduce the entropy.

By the way, although the above-described embodiment is a configuration example in which the low-frequency signal (L (t)) is converted into an integer, the same applies when the high-frequency signal (H (t)) is converted into an integer. In the case of an image signal, the effect of converting the high frequency signal (H (t)) into an integer is smaller than that of converting the low frequency signal (L (t)) into an integer because of its characteristics. (L (t)) is the target of integer conversion.

On the contrary, even if the integer value of one of the band signals is made a non-integer, the entropy reduction is the same, and the effect of the present invention can be brought out.

FIG. 3 is a block diagram showing an embodiment configuration of the subband synthesis filter.

In the figure, the basic components of the sub-band synthesis filter surrounded by the alternate long and short dash line, and the low-frequency signal (L (t)) input thereto
61, the high frequency signal (H (t)) 63, and the output signal 67 are the same as those in the conventional configuration shown in FIG.

The band signal restoring means provided in the sub-band synthesis filter of the present embodiment, when the high-frequency signal (H (t)) 63 transmitted in that state is a non-integer, the low-frequency signal (L
(T)) is also a non-integer, and is characterized in that it is returned to the original non-integer value by adding 0.5 or -0.5 to the low-frequency signal (L '(t)) 31 taking an integer value.

That is, the non-integer value identification circuit 33 determines that the high frequency signal (H
(T)) It is detected whether 63 is a non-integer, and when it is a non-integer, 0.5 or -0.5 is output corresponding to the non-integer value identification circuit 21 of the subband filter, When it becomes an integer, 0 shall be output. Adder 35
Adds the output value of the non-integer value discriminating circuit 33 to the low-frequency signal (L '(t)) 31 and when the high-frequency signal (H (t)) 63 takes a non-integer value, the A low frequency signal (L (t)) 61 converted into an integer value is obtained.

As described above, in the subband coding method that uses pairs in the subband filter and the subband synthesis filter, one of the band signals is converted into an integer or a non-integer on the transmission (coder) side and is directly received on the reception (decoder) side. From the state of the other band signal to be transmitted, it is possible to convert one of the band signals that have been made into an integer or non-integer into the original numerical value.

That is, the sub-band coding method of the present invention can accurately reproduce the original band signal and can significantly reduce the coding bit rate when coding one band signal, and It is possible to reduce the overall coding bit rate.

By the way, in the above-described embodiment, only the low-frequency signal (L (t)) taking a non-integer value (integer value) is set to 1 or -1 which is the minimum value of the numerical value for making it an integer (non-integer). However, if the error can be tolerated on the receiving side, it is not always necessary to add the band signal restoring means as shown in FIG. 3 to the subband synthesis filter. Further, in such a case, the subband filter does not need to limit the band signal to be integerized (non-integerized) to one, and the low band signal (L (t)) and the high band signal (H (t)). )
It is possible to make both of them into integers (non-integers) and further reduce the coding bit rate.

Further, when converting one band signal into an integer or non-integer with the subband filter, it is necessary to determine whether the band signal has an integer value or a non-integer value.
In the configuration shown, the input signal sequences X (t) and X (t
-1) When one is an even number and the other is an odd number, the fact that each band signal is a non-integer is used to take advantage of the high band signal (H
(T)) is being monitored.

However, as shown in FIG. 4 (a), one of the band signal itself which is converted into an integer or a non-integer may be monitored. Further, as shown in FIG. 4 (b), two input signal sequences X (t) and X (t−) in which the non-integer value identification circuit 41 is continuous.
The same applies to a configuration in which 1) is directly monitored and the band signal takes a non-integer value when only one of them is an odd number. In the latter case, to detect that the band signal has an integer value, two consecutive input signal sequences X (t) and X (t-1) are both even or both odd. Determine that.

In any case, the configuration of the subband synthesis filter is the same as that shown in FIG. 3, and can be handled by the band signal restoring means according to the subband filter.

〔The invention's effect〕

As described above, the subband filter of the present invention can significantly reduce the entropy of at least one band signal.

That is, by using this subband filter, it is possible to reduce the overall coding bit rate.

Further, the subband coding method of the present invention is made integer (non-integerized) by adding band signal entropy reducing means to the conventional subband filter and adding band signal restoring means to the subband synthesis filter. It is possible to accurately reproduce one band signal and significantly reduce the entropy of one band signal.

That is, it is possible to reduce the overall coding bit rate while maintaining high quality characteristics.

[Brief description of drawings]

FIG. 1 is a block diagram showing the basic configuration of the present invention. FIG. 2 is a block diagram showing the configuration of an embodiment of the subband filter of the present invention. FIG. 3 is a block diagram showing an embodiment configuration of a subband synthesis filter. FIG. 4 is a block diagram showing the configuration of another embodiment of the subband filter of the present invention. FIG. 5 is a block diagram showing a configuration example of a conventional subband filter using a second-order perfect reproduction type filter. FIG. 6 is a block diagram showing a configuration example of a conventional subband synthesis filter. 21, 33, 41 …… Non-integer value identification circuit, 23, 35 …… Adder,
25, 31 ... Low-frequency signal (L '(t)), 51 ... Input signal (X (t)), 52, 57 ... Multiplier, 53 ... Delay circuit, 54
…… Adder, 55 …… Low-pass signal (L (t)), 56 …… Low-pass filter (LPF), 58 …… High-pass signal (H (t)), 5
9 …… High-pass filter (HPF), 61 …… Low-pass signal (L
(T)), 62, 64 ... Multiplier, 63 ... High frequency signal (H
(T)), 65 ... Adder, 66 ... Switch circuit, 67 ...
Output signal.

Claims (4)

[Claims] 1. A second-order low-pass filter having a coefficient value of (0.5, 0.5) and a second-order high-pass filter having a coefficient value of (0.5, −0.5) are input. In a sub-band filter that divides a predetermined integer digital signal into low-band and high-band signals and outputs them, when at least one of the band signals has a non-integer value, at least one band signal is adjusted to a predetermined value. A subband filter characterized by comprising a band signal entropy reducing means for converting it into a numerical value. 2. The subband filter according to claim 1, wherein the band signal entropy reducing means converts at least one band signal into a predetermined non-integer value when each band signal has an integer value. A sub-band filter characterized by being. 3. A second-order low-pass filter having a coefficient value of (0.5, 0.5) and a second-order high-pass filter having a coefficient value of (0.5, −0.5) are input. A sub-band filter that divides a predetermined integer digital signal into low-band and high-band band signals and outputs the signals, and coding means that codes each band signal by a corresponding coding method, multiplexes the signals, and transmits the signals. Decoding means for demultiplexing the received signal, decoding corresponding to the above-mentioned coding method, and outputting as respective band signals of the low band and the high band; In a sub-band coding method including a sub-band synthesis filter that reproduces the digital signal by a process corresponding to the filter, the sub-band filter has one of the following when each band signal has a non-integer value. The band signal includes a band signal entropy reducing unit for converting the band signal into a predetermined integer value and outputting the band signal to the encoding unit, wherein the sub-band synthesis filter is configured such that the other band signal with respect to the one band signal that is an integer is not A sub-band coding method, comprising band signal restoration means for converting the one band signal into an original non-integer value when taking an integer value. 4. The subband coding system according to claim 3, wherein the band signal entropy reducing means of the subband filter sets one band signal to a predetermined non-uniformity when each band signal has an integer value. The band signal restoring means of the sub-band synthesis filter is configured to convert to a numerical value and output to the encoding means, when the other band signal with respect to the one non-integerized one band signal has an integer value, A subband coding method characterized in that one band signal is converted into the original coefficient value.