JP2871149B2 - Band division coding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a band division coding method used in a coding apparatus and a decoding apparatus for performing band compression on signals such as image signals and audio signals.

[0002]

2. Description of the Related Art A band division (sub-band) coding method is known as one of band compression or high-efficiency coding techniques for image signals, audio signals, and the like. A conventional band division encoding method when an image signal is input will be described below.

FIG. 6 is a block diagram of a conventional band division coding apparatus. In FIG. 6, a is an encoder block, b
Is a decoder block, 400 is a bandpass filter, 4
01, 402, 403, and 404 are the first, second,
Third and fourth encoders, 405 is a multiplexer, 40
6 is a communication path, 407, 408, 409 and 410 are first, second, third and fourth band-divided signals, respectively.
1 is a demultiplexer, 412, 413, 414, 41
5 is a first, second, third and fourth decoder, respectively, 41
6, 417, 418, and 419 are first, second, third, and fourth decoded band signals, respectively, and 420 is a band-pass filter.

[0004] The operation of the band division coding apparatus configured as described above will be described below. In FIG.
The input image signal is divided into four bands 407, 408, 409, and 410 by a band-pass filter 400, and encoders 401, 402, 403, and 40 are provided for each band.
4 is encoded. As shown in FIG. 5, each of the encoders includes an encoder 300, a quantizer 301,
It comprises a variable length encoder 302. The encoder 300 used in each band is configured by a combination of various encoding methods such as predictive encoding, orthogonal transform encoding, vector quantization, intra-frame encoding, and inter-frame encoding. The encoded signals are combined by the multiplexer 405 and sent out to the communication path 406. On the decoder side, the signal sent from the communication path 406 is demultiplexed by the demultiplexer 41.
1 for each original band signal, and a decoder 41 for each band.
2, 413, 414, and 415, pass through a band-pass filter 420, and are combined. Since the transfer rate is constant in the communication channel 406, the encoder determines the bit rate of each band in advance according to the transfer rate, and determines the compression ratio (for example, “Digital Coding of Waveform” (N S. Jay
"DigitalCo" by ant-Peter Noll
Ding of Waveforms "pp.486-
509)).

[0005]

However, in the configuration shown in FIG. 6, the bit allocation to each band is unique, and encoding is not performed according to the characteristics of the input signal which changes every moment. Since it does not cope with the change in the information amount of the signal, the coding quality cannot be improved.

The present invention has been made in view of the above problems, and enables control of a bit generation amount by encoding according to input signal characteristics.
Further, it is an object of the present invention to provide a band division coding method for improving the coding quality by keeping the distribution balance of distortion among the divided bands constant.

[0007]

In order to solve the above-mentioned problems, a band division encoding method of the present invention divides a band of an input signal, encodes each in-band signal for each divided band,
In a band division coding method for multiplexing and transmitting coded signals of each band obtained by coding, a buffer is provided for controlling a transfer rate, and the internal capacity of the buffer is set to a predetermined time interval Δt. And a band corresponding to the lowest frequency band among the divided input signal bands according to the change amount of the buffer capacity from the buffer capacity before the time interval Δt to the current buffer capacity. The quantization width of the in-band signal is determined, and the lowest frequency is set so that the coding distortion of the band corresponding to the lowest frequency band and the coding distortion of the other frequency bands have a predetermined ratio to each other. Characterizing and controlling the quantization width of the frequency band other than the band corresponding to the band,
In the control of the transfer speed of each in-band signal, the quantization width of the signal of at least one band other than the band corresponding to the lowest frequency band is reduced only when the amount of change in the buffer capacity is monotonically decreasing. It is characterized by. Further, in the control of the ratio of the coding distortion between the respective band signals, the quantization of at least one in-band signal other than the band corresponding to the lowest frequency band is performed only when the amount of change in the buffer capacity is not monotonically increasing. Characterized in that the width is reduced, and in the control of the ratio of the coding distortion of each band signal, the coding distortion of the signal of the band corresponding to the lowest frequency band and the coding distortion of the other band. In ratio
If the coding distortion of the signal of the band corresponding to the lowest frequency band is larger, the signal of at least one band other than the band corresponding to the lowest frequency band is satisfied so as to satisfy the predetermined ratio. Is characterized by increasing the quantization width.

[0008]

According to the above-described configuration of the band division encoding method of the present invention, in encoding the band-divided in-band signal of the input signal, the amount of bits generated by the encoding is given priority over the quantization step width of the lowest band. , The bit generation amount is controlled by coarse control, and at the same time, the quantization step width of each in-band signal is set so that the encoding distortion of each in-band signal has a predetermined balance. By fine-tuning and updating, the bit allocation for each in-band signal is made variable, and the signal quality at the time of decoding is improved.

[0009]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a band division coding method according to an embodiment of the present invention.

FIG. 1 is a block diagram of a band division encoder showing one embodiment of the band division encoding method of the present invention. FIG.
, 100 is a band division filter, 101 is a first encoder, 102 is a second encoder, 103 is a third encoder, 104 is a fourth encoder, 105 is a distortion detection block, 106 Is a transfer rate control buffer, 107 is a buffer bit amount detection block, 108 is an encoding parameter control block, 109 is a communication path, 110 is a signal of the first divided band, and 111 is a signal of the second divided band. A signal 112 is a signal of the divided third band, and a signal 113 is a signal of the divided fourth band.

FIG. 2 to FIG. 4 are control flowcharts of the band division encoder using the band division encoding method of the present invention. FIG. 5 is a diagram showing a configuration in the encoder block.
In FIG. 5, 300 is an encoder, 301 is a quantizer, 3
02 is a variable length encoder.

First, the operation of the band division encoder using the band division encoding method according to one embodiment of the present invention will be described with reference to FIG. In FIG. 1, an input image signal is first, second, and
Third and fourth four bands 110, 111, 112, 11
The signals are divided into three signals and input to first, second, third, and fourth encoders 101, 102, 103, and 104, respectively, and are encoded.

Here, the number of bands to be divided is not limited to this. These encoders are, as shown in FIG. 5, an encoder 300 and a quantizer 30 corresponding to each band.
1, a variable length encoder 302. The encoder 300 used in each band can be configured by a combination of various encoding methods such as predictive encoding, orthogonal transform encoding, vector quantization, intra-frame encoding, and inter-frame encoding. In the present invention, since the encoding means is not particularly limited, the signal output variable-length coded in the first, second, third, and fourth encoders 101, 102, 103, and 104, which is omitted, has a transfer rate of The data is sent to the control buffer 106 and accumulated. On the other hand, the bit rate determined by the transfer rate of the communication path 109 is sent out from the transfer rate control buffer 106 to the communication path 109 every hour.

Since the information amount of the input signal changes every moment, the bit amount generated when encoding each band also changes accordingly. Also, according to the spectrum of the input signal, the first, second, third and fourth four bands 110,
Since the information amounts of the signals 111, 112, and 113 also vary from time to time, the amount of bits generated when encoding each band also varies accordingly.

However, since the transfer rate of the communication path 109 is constant, the first, second, third, and fourth four bands 110,
It is necessary to adjust the total amount of bits generated when encoding the signals 111, 112, and 113 to the transfer rate determined by the communication path. At the same time, the first, second, third, and fourth bands 110, 111, 112, 113
It is necessary to adaptively adjust the balance of the respective bit generation amounts according to the characteristics of the input signal.

Although various methods are conceivable for this, the present invention is characterized in that adaptive processing is performed by controlling the coding distortion of each band.

The method will be described below. First, the distortions D1, D2, D3, and D4 caused by the respective encoders 101, 102, 103, and 104 are encoded by a distortion detection block 105.
Is detected by The information on the amount of accumulated bits in the transfer rate control buffer 106 is measured by the in-buffer bit amount detection block 107 for each unit of time Δt, and the measured information is sent to the encoding parameter control block 108.
The encoding parameter control block 108 determines the quantization step width of the encoder in each band according to the flowcharts shown in FIGS. By controlling the quantization step width, bit generation amount control and distortion control are performed.

Next, the bit generation amount control and distortion control of the band division encoder using the band division encoding method of the present invention will be described in detail with reference to the flowcharts of FIGS. FIG. 2 shows the entire bit generation amount control and distortion control.
Is shown in the flowchart of FIG.

First, the coding parameter control block 10
8 measures the bit accumulation amount R in the buffer every time Δt,
The quantization step width (Qstep1) of the first band is determined. Normally, the energy of an image signal is concentrated in the lowest band (corresponding to the first band), the amount of information contained in the first band is the largest, and the amount of all bits generated and the code for the image quality after decoding are high. The influence of the distortion is dominated by the first band.

Therefore, first, the quantization step width (Qstep1) of the first band is determined to determine the coding distortion in the first band, and at the same time, coarse control is applied to the bit generation amount.

Next, the coding distortion of each band is controlled. This can be achieved by adjusting and setting the quantization step width of each band so that the coding distortion between the bands has a predetermined ratio. At this time, in order to reduce the influence of the bit amount generated by encoding the information amounts of the second, third, and fourth bands to the aforementioned coarse control on the bit generation amount, It is necessary to keep the variation width of the quantization step width of the third and fourth bands smaller than the quantization step width of the first band.

The above processing is repeated at the above-mentioned measurement time intervals Δt. As Δt is smaller, the quantization step width is determined following the actual information amount, and the bit generation amount can be made substantially constant. However, the quantization step width follows the actual information amount. Due to the fluctuation, the coding quality changes from moment to moment, and may interfere with the image quality.

On the other hand, if Δt is large, the response of the quantization step width and the bit generation amount to the actual information amount is delayed, so that the system is relatively stable. Correctly, unless a buffer having a large bit accumulation amount is prepared for the transfer speed regulation, an abnormal state of the system due to a buffer overflow occurs. In view of these, it is necessary to select an appropriate Δt.

Next, a method of determining the quantization step width (Qstep1) of the first band and a method of controlling the bit generation amount will be described in detail with reference to the flowchart of FIG. When the amount of accumulated bits in the buffer in the transfer rate control buffer 106 approaches the maximum buffer capacity (in the case of the flowchart of FIG. 3, this corresponds to a case where the accumulated amount of bits in the buffer is 80% or more of the maximum buffer capacity). , The quantization step width of the fourth band (Qstep3, Qstep
p4) Forcibly add +1 or +2.

This is to prevent the buffer from overflowing. Next, Qstep1 is determined according to the bit accumulation amount R in the buffer. Although various methods are conceivable, for example, it can be determined as follows from the bit accumulation amount R in the buffer at the current time t. Where I
NT is a round-down operation to an integer, and C1 is a constant. However, the decision method is not limited to this.

Qstep1 = INT [R / C1] When determining this quantization step width, Qstep before time Δt
In order to reduce temporal and spatial fluctuations in the decoded image quality due to a large change width from the value of step1 to the value of Qstep1 just determined, if the fluctuation width of the quantization step width is 2 or more, the fluctuation of Qstep1 The change width is suppressed by -1.

Next, the quantization step width (Qstep4, Qstep3, Qstep2) other than the first band is adjusted according to the bit accumulation amount R in the buffer at the present time t and the bit accumulation amount Rb in the buffer before the time Δt. In this embodiment, the quantization step widths (Qstep4, Qstep3) of the fourth band and the third band are adjusted. The adjustment method calculates, for example, the bit accumulation amount change rate B in the buffer by the following equation. Correctly, the formula and the way of changing are not limited to this.

B = (R−Rb) / C2 C2: constant According to the calculated value of B, when B is positive (when bit generation is monotonically increasing), Qstep4, Qstep3, Qs
At least one of step 2 is increased to reduce the bit generation amount. Conversely, when B is negative (bit generation is monotonically decreasing), Qstep 4, Qstep 3, Qstep
The bit generation amount is increased by decreasing at least one of the two.

In the present invention, Qstep4 and Qstep3 are changed according to the value of B as shown in the flowchart of FIG. However, the way of changing is not limited to this. By the above operation, the fluctuation of the bit generation amount can be suppressed. In this case, the quantization step width of the high-frequency band is preferentially increased. However, even if the high-frequency component is preferentially discarded, the image quality deterioration is not noticeably detected by human eyes. is there.

Next, details of the above-described distortion control method for each band will be described with reference to the flowchart of FIG.

This is a method of controlling the ratio of the coding distortion between the respective bands so that the ratio becomes a predetermined value. First, quantization distortions D1, D2, D3, and D4 in each band are measured by the distortion detection block 105 for each time Δt. This may be the measurement of the energy of the quantization distortion, or simply the sum of the absolute values of the quantization distortion as shown in the following equation. Here, X is an input signal, and Xq is a quantized signal. However, the measurement of the encoding distortion is not limited to this.

D = Σ | X−Xq | From the quantization distortion measured for each band, D2 / D1, D3 of the distortion of each band with respect to the quantization distortion D1 of the first band
/ D1, D4 / D1 are calculated. The ratio of the distortion of each band to the quantization distortion of the first band is a predetermined value, for example, D2 / D1 = 1, D3 / D1 = 2, D4 /
The quantization step width of each band is adjusted so that D1 = 4. At this time, in order not to disturb the bit generation amount control described above, the quantization step width is reduced only when the change in the accumulation amount in the buffer from time t-Δt to time t is negative.

When the quantization distortion of the lower band is larger than the quantization distortion of the higher band, the quantization step width of the higher band is forcibly increased to suppress the amount of bit generation in the lower band. And to minimize image quality degradation. In this case, the amount of change in the quantization step width is set to ± 1, so as to avoid a sharp change in image quality.

[0034]

As described above, according to the band division coding method of the present invention, in the coding of the band-divided in-band signal of the input signal, the bit generation amount due to the coding is reduced by the quantization of the lowest band. By giving priority to the step width,
At the same time, the quantization step width of each in-band signal is fine-tuned and updated so that the coding distortion of each in-band signal has a predetermined balance so that the bit generation amount is controlled roughly. By doing so, there is an effect that the bit allocation for each in-band signal can be made variable and the quality of the decoded signal can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a band division encoder showing one embodiment of a band division encoding method of the present invention.

FIG. 2 is a control flowchart of a band division encoder using the band division encoding method of the present invention.

FIG. 3 is a flowchart showing a method for determining the quantization step width and a method for controlling the bit generation amount.

FIG. 4 is a flowchart showing a distortion control method for each band.

FIG. 5 is a diagram showing a configuration in an encoder block.

FIG. 6 is a block diagram of a conventional band division encoder.

[Description of Code] 100 Band Division Filter 101 First Encoder 102 Second Encoder 103 Third Encoder 104 Fourth Encoder 105 Distortion Detection Block 106 Transfer Rate Control Buffer 107 Bit in Buffer Amount detection block 108 Coding parameter control block 109 Communication channel 110 Divided first band signal 111 Divided second band signal 112 Divided third band signal 113 Divided fourth band Signal

Continuation of the front page (72) Inventor Ryozo Kishimoto 1-6, Uchisaiwaicho, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (56) References JP-A-3-62789 (JP, A) JP-A-3-3-89 191618 (JP, A) (58) Field surveyed (Int. Cl. ⁶ , DB name) H04N 1/41 -1/419 H04N 7/24-7/68

Claims (4)

(57) [Claims] 1. Band division coding for dividing a band of an input signal, encoding a signal in each band for each of the divided bands, and multiplexing and transmitting a coded signal of each band obtained by encoding. In the method, a buffer is provided for controlling a transfer rate, and the content of the buffer is monitored at a predetermined time interval Δt, and a buffer from the content of the buffer before the time interval Δt to the current content of the buffer is provided. The quantization width of the in-band signal of the band corresponding to the lowest frequency band among the divided input signal bands is determined according to the amount of change in the internal capacity, and the code of the band corresponding to the lowest frequency band is determined. The quantization width of a frequency band other than the band corresponding to the lowest frequency band is determined and controlled so that the coding distortion and the coding distortion of other frequency bands have a predetermined ratio. Band division coding method for. 2. The quantization width of a signal of at least one band other than the band corresponding to the lowest frequency band only when the amount of change in the buffer capacity is monotonically decreasing in the control of the transfer speed of each band signal. 2. The band division encoding method according to claim 1, wherein 3. The at least one in-band signal other than the band corresponding to the lowest frequency band only when the amount of change in the buffer capacity is not monotonically increasing in the control of the ratio of the coding distortion between the respective band signals. 2. The band division encoding method according to claim 1, wherein the quantization width is reduced. 4. The control of the ratio of the mutual encoding distortion of each band signal, wherein the ratio of the encoding distortion of the signal of the band corresponding to the lowest frequency band to the encoding distortion of the other band is the lowest frequency. If the coding distortion of the signal of the band corresponding to the band is larger, the quantization of at least one band of the signal other than the band corresponding to the lowest frequency band is performed so as to satisfy the predetermined ratio. 2. The band division encoding method according to claim 1, wherein the width is increased.