JPH04152782A - Picture coding transmitter - Google Patents

Abstract

PURPOSE:To improve the coding efficiency by selecting a prescribed non-flat characteristic for a frequency amplitude characteristic of each filter being a component of a band split section and a synthesis section in response to an frequency amplitude characteristic of a power spectrum in a pass band of an input signal to each filter. CONSTITUTION:A non-flat frequency amplitude characteristic to bring a power spectrum of band split signals 3, 4 being input signals to PCM coders 6, 7 to a white noise level is provided to a split low pass filter 27 and a split high pass filter 28. A non-flat characteristic to a reverse characteristic to the frequency amplitude characteristic of the split low pass filter 27 and the split high pass filter 28 is provided to a synthesis low pass filter 33 and a synthesis high pass filter 34. The band split signals 3, 4 obtained by sub-sampling output signals of the split low pass filter 27 and the split high pass filter 28 respectively by a low frequency subsample circuit 29 and a high frequency subsample circuit 30 have power spectrums subject to white noise level processing. Thus, the coding efficiency is improved more than a conventional system employing PCM coder and decoder.

Description

[Detailed description of the invention]

[0001]

[Industrial application field]

The present invention relates to an image encoding and transmitting apparatus that encodes and transmits an image signal, and particularly relates to an image encoding and transmitting apparatus that uses subband encoding that divides an image signal into bands and encodes each band. [0002]

[Conventional technology]

One of the image signal encoding methods is to divide the image signal into multiple bands and encode each band divided signal.
A so-called subband encoding method is known. In addition to the feature of high coding efficiency, this method hierarchizes image signals in order of visual importance, so ATM (A
synchronous' Transfer N1od
e) When applied to transmission, it has the advantage of minimizing image quality deterioration due to packet discard. This point will be briefly explained. [0003] FIG. 2 shows a basic block diagram of a subband coding transmission device. Input image signal 1 is divided into N band divided signals 3 to 5 by band dividing section 2 . Each band division signal 3~
5 are encoded by encoders 6 to 8, respectively, and transmitted through transmission line 9. On the receiving side, the transmitted output data from encoders 6 to 8 is transmitted to decoder 1.
0 to 12 to generate decoded band division signals 13 to 15, these signals 13 to 15 are combined by a combining section 16, and the output is a decoded image signal 17.
get. [0004] Here, each band pass characteristic 18 when band dividing the input image signal 1 into each band divided signal 3 to 5 in the band dividing unit 2
~20 has the characteristics shown in Fig. 3, the band division signal 3 on the low frequency side contains the most visually important information, and conversely, the band division signal 5 on the high frequency side is visually the least important. It will contain information. Therefore, when multiple transmission modes with different packet discard rates can be used together in ATM transmission, the output of the encoder 6 is transmitted using the transmission mode with a low packet discard rate, and the output of the encoder 6 is transmitted using a transmission mode with a high packet discard rate. If the output of the encoder 8 is transmitted using
Transmission can be performed with less deterioration in image quality due to packet discard. In addition, when only a single transmission mode can be used and the outputs of encoders 6 to 8 are multiplexed and transmitted in a single transmission mode, the output of encoder 6 has an error correction function with high correction capability. By adding a code and adding an error correction code with low correction ability to the output of the encoder 8, or not adding an error correction code, it is possible to perform transmission with less deterioration in image quality due to packet discard. can. [0005] The band dividing section 2 and the combining section 16 are QMF (Quadrat
A basic dividing section and a basic synthesizing section each containing a band division filter and a synthesis filter (called ure-Mirror Filters) are used as basic components, and these can be used alone or configured by connecting multiple pieces in cascade. is common. Encoders 6 to 8, decoders 10 to
12, each D P CM (Different
ial PCM) encoder and DPCM decoder are generally used. Next, after briefly explaining QMF, the reason for using a DPCM encoder and DPCM decoder and the problems associated with this will be explained. [0006] FIG. 4 shows a block diagram of a basic dividing section and a basic combining section including a QMF. The input signal 21 is divided by the basic dividing section 23 into a low frequency divided signal 25 and a high frequency divided signal 26, and the basic combining section 24 divides the input signal 21 into a low frequency divided signal 25 and a high frequency divided signal 2.
6, a restored signal 22 is obtained. In the basic dividing section 23, the input signal 21 passes through a divided low-pass filter 27, and then is subsampled at a ratio of 2:1 by a low-pass subsampling circuit 29 to become a low-frequency divided signal 25, and at the same time, the input signal 21 is filtered through a divided high-pass filter 27. After passing through 28, the signal is subsampled at a ratio of 2:1 by a high frequency subsampling circuit 30, and becomes a high frequency divided signal 26. In the basic synthesis section 24, the low-frequency divided signal 25 is passed through a low-frequency zero interpolation circuit 31 and a synthesis low-pass filter 33, and the high-frequency division signal 26 is passed through a high-frequency zero interpolation circuit 32 and a synthesis high-pass filter 33. The signal passed through the pass filter 34 is added by an adder circuit 35 to form a restored signal 22. Here, the transfer function of the divided low-pass filter 27 is defined as Hl (
z) The transfer function of the divided high-pass filter 28 is Hu
(z), the transfer function of the synthetic low-pass filter 33 is G 1
(z), the transfer function of the synthetic high-pass filter 34 is G
If u (z), then the following relational expression holds true in QMF. Gl(z)=H1(z)
(1) Gu(z)=Hu(z)
(2) Hu(z)
= Hl (-z)
(3) Hl (expjθ) l +IH
u(expjθ)+2=1 (4)(O≦θkuπ
) [0007] However, when the order of each filter is finite, Equation (4) becomes
This is an equation that approximately holds true. As shown in FIG. 5, equation (3) is based on the frequency amplitude characteristic IH1(expjθ)1 of the divided low-pass filter 27 shown by the curve 36, and the frequency amplitude characteristic IHu(expjθ)1 of the divided high-pass filter 28 shown by the curve 37. e
This corresponds to the fact that Xpjθ)1 has a mirror image relationship with θ=π/2 as the center. Note that when the order of each filter is finite, the frequency amplitude characteristic IH■(eXpjθ)I of the divided low-pass filter 27 is expressed as
/2 cannot be made completely zero, and similarly, the frequency amplitude characteristic I Hu (ex
p jθ)1 cannot be made completely zero at θ<π/2. Therefore, aliasing distortion occurs when subsampling is performed at a ratio of 2:1 in the low-frequency sub-sampling circuit 29 and the high-frequency sub-sampling circuit 3°, but when addition is performed in the adding circuit 35 based on the condition of equation (1), aliasing distortion occurs. is completely removed. [0008] FIG. 12 shows the simplest case in which the number of blocks N of an example of a conventional subband coding transmission device using QMF is 2, and the dividing section 2 and the combining section 16 are each composed of one QMF block. The basic dividing section 42 and the basic combining section 43 are used as the encoders 6 and 7, and DPCM decoders are used as the decoders 10 and 11, respectively. [0009] FIG. 13 shows a block diagram of another example of a conventional subband coding transmission apparatus using QMF. This example is a case where the number of band divisions N is 3 in the subband coding transmission apparatus shown in FIG.
and a second basic dividing section 39, and a combining section 1
6 is a first basic synthesis section 40 and a second basic synthesis section 41
It is made up of. In addition, encoders 6 to 8 have DPC
DPCM decoders are used as the M encoder and decoders 10 to 12, respectively. In the band division section 2, the input image signal 1 is manually input to the first basic division section 38, and the low frequency division signal obtained by this basic division section 38 is further input to the second basic division section 39, and this basic division The low frequency division signal and the high frequency division signal obtained in the section 39 are respectively the band division signal 3.
, 4, and the high frequency division signal obtained by the basic division section 38 becomes the band division signal 5. In the combining section 16, first, the decoded band divided signals 13.14 are inputted to the second basic combining section 41 corresponding to the second basic dividing section 39 as a low frequency divided signal and a high frequency divided signal. Next, the second basic synthesis section 41
The synthesized signal obtained in the above and the band-divided signal 15 are sent to the first basic combining section 40 corresponding to the first basic dividing section 38, respectively.
are input as a low-frequency divided signal and a high-frequency divided signal, and a restored image signal 17 is obtained as an output. [0010] In the example shown in FIG. 13, each of the band dividing section 2 and the combining section 16 is configured by connecting two basic dividing sections and two basic combining sections in cascade, and further includes a large number of basic dividing sections. , by cascading the basic combining units, it is possible to realize a subband coding transmission device in which the number of divisions is even greater. In addition, although FIGS. 12 and 13 show a coding transmission device for one-dimensional signals, it is also possible to connect two-dimensional signals in both the horizontal and vertical directions by cascading one-dimensional basic dividing sections and basic combining sections. By using the present invention, it is also possible to extend it to a subband coding transmission device for two-dimensional signals. [0011] The reason why DPCM encoding/decoding is performed on each band division signal in a conventional subband encoding transmission device is as follows. Generally, in subband encoding, if the characteristics of the band division filter and synthesis filter are ideal, by increasing the number of band divisions sufficiently, PCM encoding/
It is known that highly efficient encoding can be achieved even by using decoding. However, in order to bring the characteristics of the band division filter and the synthesis filter closer to ideal characteristics, the order of the filter must be increased. In order to reduce the number of band divisions, a high-order band division filter that uses many elements,
A large number of synthesis filters must be prepared, which significantly increases the hardware scale. Therefore,
The number of band divisions is not made too large, and the correlation remaining in each divided band signal is D P C.
Since it is advantageous in terms of hardware to perform M-encoding for removal, it is generally widely used to use a method that uses DPCM encoding/decoding in combination. [0012] However, when subband encoding combined with DPCM encoding/decoding is applied to ATM transmission, if a transmission error occurs due to packet loss, the error is
Due to encoding/decoding, the transmission error spreads even to parts where no transmission error has occurred, significantly degrading the image quality. In other words, the advantage of subband encoding that image quality deterioration due to transmission errors can be minimized is not fully utilized by combined use of DPCM encoding/decoding. In addition, in order to suppress the spread of transmission errors caused by DPCM encoding/decoding, a method of resetting errors by performing PCM encoding periodically or aperiodically may be considered, but this has the disadvantage of reducing encoding efficiency. occurs. [0013]

[Problem to be solved by the invention]

As mentioned above, in order to reduce the amount of hardware for band division filters and synthesis filters in subband encoding, the number of band divisions is relatively small, and the resulting decrease in efficiency is reduced by DPCM encoding for each band division signal. /If compensation is performed by decoding, the advantage of subband coding, which is to minimize image quality deterioration due to transmission errors, will be halved due to the ripple effect of transmission errors caused by DPCM encoding/decoding. The result is that it is closed. Furthermore, if PCM encoding is performed without error spread by reducing the number of band divisions, the encoding efficiency will decrease. [0014] An object of the present invention is to provide an image encoding and transmitting apparatus that can reduce the number of band divisions to reduce the amount of hardware and increase encoding efficiency without using DPCM encoding that causes error propagation. do. [0015]

[Means to solve the problem]

In order to solve the above-mentioned problems, the present invention inputs an input image signal to a band division unit including a plurality of band pass filters to generate a plurality of band division signals, and outputs these band division signals to a plurality of PCM encoders. and each PCM
The encoded data output from the encoder is transmitted via a transmission path, and each of the encoded data transmitted from the transmission path is decoded by a plurality of PCM decoders.
In an image coding and transmission device that restores an image signal by inputting a plurality of decoded band-split signals output from an M decoder to a synthesis unit including a plurality of synthesis filters, the power spectrum of the input signal of the PCM encoder is so that it is almost white,
At least one frequency amplitude characteristic of the band division filter has a predetermined non-flat characteristic within the pass band, and correspondingly, at least one frequency amplitude characteristic of the synthesis filter has a predetermined non-flat characteristic within the pass band. It is characterized by [0016] The PCM encoder used in this invention may be a normal PCM encoder, but especially a PC with a quantization noise shaping filter.
M encoders are more preferred. This PCM encoder with a noise shaping filter is composed of, for example, a PCM encoder, a PCM decoder, an adder/subtractor, and a quantization noise shaping filter, and a band division signal which is an input signal to the PCM encoder. The output signal of the quantization noise shaping filter is input to the adder, the output of the adder is input to the PCM encoder, the output of the PCM encoder is used as the output of each encoder, and the output signal of the PCM decoder is input to the adder. The input of the PCM encoder and the output of the PCM decoder are input to a subtracter, and the output of the subtracter is input to a quantization noise shaping filter. [0017]

[Effect]

By making the frequency-amplitude characteristics of the band-splitting filter non-flat as described above within the passband, unlike when using QMF, the frequency-amplitude characteristics of the band-splitting filter are almost flat within the passband. The power spectrum of the band-split signal input to each encoder is shaped into white (flat) or a shape close to white. [0018] In general, a signal whose power spectrum is not white can be directly PCM'd.
Rather than encoding/decoding, the signal is converted into a form whose power spectrum is close to white by the first filter, and then PCM
It is known that so-called DPCM has higher encoding efficiency, in which encoding/decoding is performed by passing the signal through a second filter having characteristics opposite to the first filter after PCM decoding. In this invention, the band division filter exhibits the same effect as the first filter, and the synthesis filter exhibits the same effect as the second filter, so that the encoder/decoder becomes the PCM encoder/decoder. Even with a encoder, it is possible to improve the encoding efficiency. [0019] On the other hand, the same functionality as the DPCM encoder can also be obtained by cascading a configuration of a prefilter whose characteristics are determined by the DPCM prediction function and a PCM encoder with a quantization noise shaping filter. Are known. Regarding this, see, for example, Reference 1: N., 5 Jayant, “Di
Digital Coding of Waveforms
”, Prentice-Hall 1984. A DPCM decoder can similarly be considered as a cascade of a PCM decoder and a post-filter. [0020] In the present invention, By making the frequency amplitude characteristics of the band division filter and the synthesis filter non-flat as described above, the band division section has a pre-filter function in addition to the original band division function, and the synthesis section can perform the original synthesis. In addition to this function, it is possible to provide a post-filter function. Therefore, by using a PCM encoder with a quantization noise shaping filter as an encoder and a PCM decoder as a decoder, DPCM as encoder/decoder
While achieving the same coding efficiency as the conventional method using an encoder/decoder, transmission errors do not spread beyond the impulse response of the synthesis filter, making full use of the features of subband coding. [0021]

【Example】

FIG. 1 shows a block diagram of an image encoding and transmitting apparatus using subband encoding according to an embodiment of the present invention. On the transmitting side, the human image signal 1 is divided into N=
After being divided into two band division signals, the PCM encoder 6
．． 7, and the resulting encoded data is transmitted to the receiving side through a transmission path 9. On the receiving side, the encoded data transmitted from the transmitting side is processed by a PCM decoder 10.
11, the decoded band-split signal 13.1
4 are generated, the synthesizer 16 synthesizes them to obtain a restored image signal 17. [0022] The band division section 2 is configured by one basic division section 42 in this embodiment. In the basic dividing section 42, the input image signal 1 passes through a divided low-pass filter 27, and then is subsampled at a ratio of 2:1 by a low-pass subsampling circuit 29 to become a low-pass divided signal 25, and a divided high-pass filter 27. Pass filter 28. After passing through, it is subsampled at a ratio of 2:1 by a high frequency sub-sampling circuit 30 to become a high frequency divided signal 26. On the other hand, the combining section 16 is composed of one basic combining section 43 corresponding to the basic dividing section 2. In the basic synthesis section 43, the decoded low-frequency divided signal from the PCM decoder 10 is passed through the low-frequency zero interpolation circuit 31 and the synthesis low-pass filter 3.
3 and the signal obtained by passing the decoded high-frequency divided signal from the PCM decoder 11 through the high-frequency zero interpolation circuit 32 and the synthesis high-pass filter 34 are added by the adding circuit 35, and a restored image is obtained. A signal 17 is obtained. [0023] In the embodiment of FIG.
are respectively replaced from the DPCM decoder to the PCM decoder, and the frequency amplitude characteristics of the divided low-pass filter 27, divided high-pass filter 28, composite low-pass filter 33, and composite high-pass filter 34. This is basically the same as FIG. 12 except that . [0024] Here, it is more preferable that the PCM encoders 6 and 7 are PCM encoders with a quantization noise shaping filter. The configuration of this PCM encoder with a quantization noise shaping filter is shown in FIG. The PCM encoder 6o with quantization noise shaping filter shown in FIG. 6 includes a PCM encoder 61, a local PCM decoder 62, an adder 63, a subtracter 64, and a transfer function H
(z) is composed of a quantization noise shaping filter 65. The input signal and the output signal of the quantization noise shaping filter 65 are input to an adder 63, and the output signal of the adder 63 becomes the input signal of the PCM encoder 61. The output signal of the PCM encoder 61 is a PCM with a quantization noise shaping filter.
The signal becomes the output signal of the encoder 6o and is also input to the local PCM decoder 62. Furthermore, the PCM encoder 61
The input signal and the output signal of the local PCM decoder 62 are input to the subtracter 64, and the output signal of the subtracter 64 becomes the input signal of the quantization noise shaping filter 65. [0025] When the frequency amplitude characteristics of each filter 27, 28, 33, and 34 in FIG. It can have the same characteristics as the conventional example shown in No. 12. This is done by the DPCM encoder/
A description will be given with reference to a block diagram showing an example of the configuration of a decoder. [0026] FIG. 7 is a block diagram of an example of a DPCM encoder/decoder. The DPCM encoder 70 in FIG. 7 includes a subtracter 73, a quantizer (PCM encoder) 71, an inverse quantizer (local PCM decoder) 72, an adder 74, and a transfer function of Hd(z). It is composed of a certain prediction filter 75. A difference signal between the input signal and the output signal of the prediction filter 75 is generated by the subtracter 73 and input to the quantizer 71. The output signal of the quantizer 71 is sent to the D PCM encoder 7o.
is output as the output signal of the inverse quantizer 72.
is input. Furthermore, the output signal of the inverse quantizer 72 and the output signal of the prediction filter 75 are input to an adder 74, and the output signal of the adder 74 is input to the prediction filter 75 as a locally decoded signal. [0027] On the other hand, the DPCMPCM decoding device shown in FIG. 7 includes a quantizer 77, an adder 78, and a prediction filter 79 whose transfer function is Hd(z). Inverse quantizer 77
The output signal of the DPCM encoder 70 is input to the DPCM encoder 70, the output signal of the inverse quantizer 77 and the output signal of the prediction filter 79 are input to the adder 78, and the output signal of the adder 78 is input to the DPCM encoder 70.
This becomes a PCM decoding signal 6 and is input to the prediction filter 79. [0028] FIG. 8 is a block diagram of another example of a DPCM encoder/decoder. The DPCM encoder 80 in FIG. 8 is configured by cascading a prefilter 81 and a PCM encoder 82 with a quantization noise shaping filter. That is,
The input signal is input to a PCM encoder 82 with a quantization noise shaping filter via a prefilter 81, and its output signal becomes the output signal of the DPCM encoder 80. here,
The transfer function of the prefilter 81 is 1-Hd(z), and the transfer function Hn(z) of the quantization noise shaping filter in the PCM encoder 82 with quantization noise shaping filter is Hd(z).
Then, this DPCM encoder 80 is D in FIG.
This is equivalent to the PCM encoder 70. [0029] The DPCM decoder 83 in FIG.
and a post-filter 85 are connected in cascade. That is, the output signal of the DPCM encoder 80 is input to the dequantizer 59, and the output signal of the dequantizer 84 is passed through the post-filter 85 and becomes an output signal of the DPCM decoder 83. Here, the transfer function of the post-filter 85 is 1/
(1-Hd(z)), the DPCM decoder 83
is the DPCMPCM decoding capacity 6 in FIG. [00301 Furthermore, in FIG. 7, the transfer function Hd of the prediction filter 75
If the characteristic given by (z) is the characteristic that maximizes the DPCM encoding efficiency, the characteristic given by the transfer function 1-Hd(z) of the prefilter 81 in FIG. This is a characteristic that flattens (whitens) the power spectrum of . [0031] In this embodiment, the DPCM encoder/
The function of the prefilter 81 in the decoder configuration is
Divided low-pass filter 2 in the basic dividing section 42 in the figure
7. In the divided high-pass filter 28, the function of the post-filter 85 is provided to the synthesis low-pass filter 33 and synthesis high-pass filter 34 in the basic synthesis unit 43 in FIG. 1, and the quantization noise in FIG. PCM with shaping filter
The encoder 82 is assumed to be the encoders 6 and 7 in FIG.
The inverse quantizer 84 in the decoder 10 in FIG.
11. As a result, the same function as the conventional example shown in FIG. 12 using a DPCM encoder/decoder as an encoder/decoder can be realized without directly using a DPCM encoder/decoder. Therefore, it is possible to improve the encoding efficiency and at the same time suppress the spread of errors. [0032] Next, the characteristics of each filter that maximizes the encoding efficiency in FIG. 1 will be described. From the above explanation, it is clear that the divided low-pass filter 27 and the divided high-pass filter 28 have non-flat filters that whiten the power spectra of the band-divided signals 3 and 4, which are the input signals of the PCM encoder 6.7. The frequency amplitude characteristic may be set as follows. Regarding the synthetic low-pass filter 33 and the synthetic high-pass filter 34, the divided low-pass filter 27 and the divided high-pass filter 2
The non-flat characteristic may be an inverse characteristic to the frequency amplitude characteristic of No. 8. Furthermore, regarding the frequency amplitude characteristics of the quantization noise shaping filter in the PCM encoder with quantization noise shaping filter used as the PCM encoder 6.7, the prefilter 81 and the quantization noise shaping filter in FIG. PC
It can be determined according to the relationship with the characteristics of the quantization noise shaping filter in the M encoder 82. [0033] Specifically, since the power spectrum Sx(θ) of the input image signal 1 in FIG. 1 generally has a high-frequency attenuation characteristic as shown by the curve 102 in FIG. In order to whiten the power spectrum of the band-divided signals 3 and 4, the frequency amplitude characteristic of the divided low-pass filter z7 in the band dividing section 2 is made to be a high-frequency emphasis characteristic within its passband, and The frequency amplitude characteristic of the divided high-pass filter 28 may be a low-frequency suppression characteristic within its passband. [0034] On the other hand, in the synthesis section 16, the synthesis low-pass filter 33
The frequency amplitude characteristic of the synthesized high-pass filter 34 is set as a high-frequency suppression characteristic that is an inverse characteristic to the characteristic of the divided low-pass filter 27 within its pass band, and the frequency amplitude characteristic of the synthetic high-pass filter 34 is set as a high-frequency suppressing characteristic that is an inverse characteristic to the characteristic of the divided low-pass filter 27 within its pass band. A low-frequency emphasis characteristic that is an inverse characteristic to the characteristic of the pass filter 28 may be used. A filter group with such characteristics cannot be realized by QMF, but for example, Document 2: M,
Vetterli, “Fiter Banks’All
owing Perfect Reconstruct
tion”, Signal Processing,
vol, 10. No, 3. April, 1
986. This is possible using the filter bank shown in PP219-244. This filter bank will be referred to as PRF (Perfect Reconst) from now on.
This will be called a ruction filter). P
In RF, each filter 27, 2 in the block diagram of the basic dividing section 23 and basic combining section 24 shown in FIG.
The characteristics of 8, 33, and 34 are set as shown in the following equation. G l (z) = Hu (-z)
(5) Gu(z) = −
Hl (-z)
(6) Hl(z) ・Hu(-z) -Hu(z)
・Hl (-z) = Z- / 2 (7)
(k is a constant) [0035] From equation (6), if the frequency amplitude characteristic I Hu (expjθ) of the divided low-pass filter 27 is a high-frequency emphasis (or high-frequency suppression) characteristic within its passband, the synthesized high frequency The frequency amplitude characteristic l Gu(expjθ)1 of the pass-pass filter 34 has a low-frequency emphasis (or low-frequency suppression) characteristic within its passband. Similarly, from equation (5), divided high-pass filter 28
If the frequency amplitude characteristic I Hu (e
1 has high frequency suppression (or high frequency emphasis) characteristics within the passband. Also, equation (7) is the restored signal 2 in FIG.
2 is delayed with respect to the input signal 21, and has exactly the same waveform as the input signal 21. Note that it is possible to completely satisfy equations (5), (6), and (7) even if the orders of each of the filters 27, 28, 33, and 34 are finite. [0036] FIG. 9 shows each filter 27, 28, 33.3 in FIG.
4 shows the frequency amplitude characteristics of No. 4. As shown in the figure, the frequency amplitude characteristic 91 of the divided low-pass filter 27, the frequency amplitude characteristic 92 of the divided high-pass filter 28, the frequency amplitude characteristic 93 of the composite low-pass filter 33, and the frequency of the composite high-pass filter 34. The amplitude characteristics 94 are as described above. As a result, the divided low-pass filter 27
The power spectrum 5xl (θ) of the output signal of the split high-pass filter 28 and the power spectrum 5xu of the output signal of the divided high-pass filter 28
(θ) becomes a spectrum that is whiter than the power spectrum Sx(θ) of the input image signal 1 shown by the curve 102 in each passband as shown by the curves 101 and 103 in FIG. The same applies to the band divided signal 3.4 obtained by subsampling the output signals of the divided low-pass filter 27 and divided high-pass filter 28 in the low-pass sub-sampling circuit 29 and the high-pass sub-sampling circuit 30, respectively. becomes the power spectrum. Therefore, encoder 6
, 7 are PCM encoders, and the decoders 10.11
Even if the PCM decoder is used, the encoding efficiency is improved compared to the conventional method using a PCM encoder/decoder. [0037] Note that the change in the frequency amplitude characteristic added to the input image signal 1 within the passband by the divided low-pass filter 27 and the divided high-pass filter 28 is caused by the change in the frequency amplitude characteristic added to the input image signal 1 by the divided low-pass filter 27 and the divided high-pass filter 28. The restored image signal 1 is compensated by the frequency and amplitude characteristics of the band pass filter 34.
The frequency amplitude characteristics of input image signal 7 are the same as those of input image signal 1. [0038] By the way, when using PRF, the frequency amplitude characteristic of the divided low-pass filter 27 (the frequency amplitude characteristic of the synthetic high-pass filter 34) and the frequency amplitude characteristic of the divided high-pass filter 28 are determined from equation (7). (frequency amplitude characteristics of the synthetic low-pass filter 33) cannot be set independently. Therefore, it is not always possible to set the characteristics of each filter so that the encoding efficiency of any of the band-divided signals 3.4 is maximized. Taking these things into consideration, each filter 27, 28 is actually designed so that the total encoding efficiency including the encoding efficiency for the band-divided signal 3.4 is maximized.
．． It is desirable to set frequency amplitude characteristics of 33 and 34. [0039] FIG. 11 shows a block diagram of an image signal encoding and transmitting apparatus according to another embodiment of the present invention. The configuration of this embodiment is such that the encoder 6, 7.8 is changed from a DPCM encoder to a PCM encoder (preferably a PCM encoder with a quantization noise shaping filter), and the decoder 10, 11.12 This is basically the same as FIG. 13 except that the DPCM decoder is replaced with a PCM decoder and the characteristics of each filter are different, and a detailed explanation of the configuration will be omitted. In FIG. 11, the characteristics of each filter constituting the first basic dividing section 38 and the first basic combining section 40 are set so that the encoding efficiency is maximized for the band-divided signal 5, Basic division part 3
9 and the characteristics of each filter constituting the second basic synthesis section 41, the composite characteristic of the characteristics of each filter constituting the first basic division section 38 and the first basic synthesis section 40 is the band-divided signal 3. , 4 is set so that the total efficiency of the encoding efficiency is maximized. [0040] Specifically, the frequency amplitude characteristics of the divided low-pass filter and the divided high-pass filter in the first basic dividing section 38 are high-frequency emphasis characteristics and low-frequency suppression characteristics, respectively, within the passband, and The frequency amplitude characteristics of the synthesis low-pass filter and the synthesis high-pass filter in the first basic synthesis section 41 may be high-frequency suppression characteristics and low-frequency emphasis characteristics, respectively, within the passband. On the other hand, the second basic dividing section 39 is similar to the first basic dividing section 38.
Since the power spectrum of the input signal input from the input signal has a high-frequency increasing characteristic, the frequency amplitude characteristics of the divided low-pass filter and the divided high-pass filter in the second basic dividing section 39 have a high-frequency increasing characteristic within the passband. With suppression characteristics and low frequency emphasis characteristics,
Further, a synthesis low-pass filter in the second basic synthesis section 40,
It is desirable that the frequency amplitude characteristics of the synthetic high-pass filter have high-frequency emphasis characteristics and low-frequency suppression characteristics, respectively, within the passband. [0041] Similarly, in order to further improve coding efficiency, even in a configuration using a larger number of basic division units and basic combination units, the power spectrum of the input signal input to each basic division unit is high. Since there may be cases where the filter has a specific frequency amplitude characteristic such as a frequency rise characteristic, the frequency amplitude characteristic of the filter may be selected accordingly. [0042] In each of the above embodiments, a PCM encoder with a quantization noise shaping filter was used as the PCM encoders 6 to 8, but a normal PCM encoder that does not include a quantization noise shaping filter may be used. Even in the case where the input signal of the encoder (bandwidth It is possible to whiten the power spectrum of the split signal) and increase the coding efficiency. However, when using a PCM encoder with a quantization noise shaping filter, it is preferable to completely whiten the power spectrum of the input signal to the encoder in order to maximize encoding efficiency; When using a PCM encoder that does not have a shaping filter, shaping the power spectrum of the input signal to the encoder into a shape that is slightly different from white, that is, close to white, is the best way to maximize encoding efficiency. Optimal. [0043] Furthermore, in the above embodiment, the frequency amplitude characteristic of the divided low-pass filter is set to a high-frequency emphasis characteristic (or high-frequency suppression characteristic) within the passband, and the frequency amplitude characteristic of the divided high-pass filter is set to be low within the passband. The frequency and amplitude characteristics of the synthetic low-pass filter are set as high-frequency suppression characteristics (or high-frequency emphasis characteristics) within the passband, and the frequency and amplitude characteristics of the synthetic high-pass filter are set as Although the low-frequency emphasis characteristics (or low-frequency suppression characteristics) are described above, these frequency amplitude characteristics are merely examples, and are not limited thereto. In short, the frequency and amplitude characteristics of the band division filter and the synthesis filter need only be determined so that the power spectrum of the input signal to the PCM encoder is shaped into white or nearly white. The non-flat characteristic may be set according to the power spectrum within the passband of the input signal. In addition, the present invention can be implemented with various modifications without departing from the scope of the invention. [0044]

【Effect of the invention】

According to the present invention, the frequency and amplitude characteristics of each of the filters constituting the band dividing section and the combining section are determined as the frequency and amplitude characteristics of the power spectrum within the passband of the input signal to each filter. By shaping the power spectrum of each band-divided signal into a white color or a shape close to white as a predetermined non-flat characteristic according to the conventional QMF, the encoder/decoder for the band-divided signal PCM encoder/
While using a decoder, it is possible to increase encoding efficiency compared to subband encoding using a conventional PCM encoder/decoder. The PCM encoder/decoder has the advantage of not causing error ripple phenomenon unlike the DPCM encoder/decoder. Therefore, according to the present invention, in ATM transmission involving packet discard, it is possible to suppress image quality deterioration due to packet discard and realize an image encoding and transmitting apparatus using subband encoding with high encoding efficiency. [0045] Further, according to the present invention, by setting the frequency amplitude characteristics of each filter as described above and using a PCM encoder with a quantization noise shaping filter as an encoder, DPCM encoding/decoding is possible. It is possible to achieve the same functionality as conventional subband coding using subbands. Therefore, high coding efficiency can be achieved, and at the same time, the spread of transmission errors, which was a drawback when using a DPCM encoder/decoder, can be suppressed to within the impulse response length of the synthesis filter.
It becomes possible to more effectively suppress image quality deterioration and increase encoding efficiency.

[Brief explanation of the drawing]

[Block diagram showing the configuration of an image coding transmission device according to an embodiment of the present invention [Fig. 2] Block diagram showing the basic configuration of an image coding transmission device using subband coding

[Figure 3] Diagram showing each band pass characteristic in the band dividing section

[Figure 4] Block diagram showing the basic configuration of the basic dividing section and basic combining section

[
Figure 5: Diagram showing the frequency amplitude characteristics of the divided low-pass filter and divided high-pass filter in the basic division section in the conventional technology

[Figure 6] Block diagram showing a configuration example of a PCM encoder with a quantization noise shaping filter used in this example

[Figure 7] Block diagram showing an example of a configuration of a DPCM encoder/decoder

[FIG. 8] Block diagram showing another configuration example of the D P CM encoder/decoder

[Figure 9] Diagram showing the frequency amplitude characteristics of each filter in Figure 1

[Figure 101] A diagram showing the frequency amplitude characteristics of the power spectra of the input and output signals of the divided low-pass filter and divided high-pass filter in Figure 1. [Figure 11] Configuration of an image encoding device according to another embodiment of the present invention. Block diagram showing

[Figure 12] Block diagram showing an example of an image encoding device using conventional subband encoding using QMF

FIG. 13 is a block diagram showing another example of an image encoding device using a conventional subband encoding method using QMF.

[Explanation of symbols]

1... Input image signal, 2... Band division section, 3, 4.
5...band division signal, 6,7.8...encoder, 9
...transmission line, 10. 11. 12... decoder,
13, 14.15...Decoded band division signal, 16...
・Composition unit, 17...decoded image signal, 23, 38, 3
9゜42...basic division part, 24,41,40.43・
...Basic synthesis unit, 21...Input signal 1.22...Restored signal, 25...Low frequency divided signal, 26...High frequency divided signal, 27...Divided low pass filter, 28 ...Divided high-pass filter, 29...Low frequency sub-sampling circuit, 3o...High frequency sub-sampling circuit, 31...Low frequency zero interpolation circuit, 32...High frequency zero interpolation circuit , 33... Synthetic low-pass filter, 34... Synthetic high-pass filter,
35...Addition circuit, 60-,. CM decoder, 63... Adder, 64... Subtractor, 65... Quantization noise shaping filter.

【Document name】

drawing

[Figure 1]

[Figure 2]

[Figure 3]

[Figure 4]

[Figure 5]

[Figure 6]

[Figure 7]

[Figure 8]

[Figure 9]

[Figure 10]

[Figure 11] (a)

[Figure 12]

[Figure 13]

Claims (3)

[Claims] Claim 1: An input image signal is input to a band division unit including a plurality of band pass filters to generate a plurality of band division signals, and each of these band division signals is encoded by a plurality of PCM encoders. The encoded data output from the PCM encoder is transmitted via a transmission path, the encoded data transmitted from the transmission path is respectively decoded by a plurality of PCM decoders, and the encoded data is output from each PCM decoder. In an image encoding and transmitting apparatus that restores an image signal by inputting a plurality of decoded band-divided signals to a synthesis section including a plurality of synthesis filters, the power spectrum of the input signal of the PCM encoder is almost white. At least one frequency amplitude characteristic of the band division filter has a predetermined non-flat characteristic within the pass band, and at least one frequency amplitude characteristic of the synthesis filter has a predetermined non-flat characteristic within the pass band. Features of image coding and transmission device. 2. An input image signal is input to a band division unit including a plurality of band pass filters to generate a plurality of band division signals, and each of these band division signals is encoded by a plurality of PCM encoders. The encoded data output from the PCM encoder is transmitted via a transmission path, the encoded data transmitted from the transmission path is respectively decoded by a plurality of PCM decoders, and the encoded data is output from each PCM decoder. In an image encoding and transmitting apparatus that restores an image signal by inputting a plurality of decoded band-divided signals to a synthesis section including a plurality of synthesis filters, the power spectrum of the input signal of the PCM encoder is almost white. At least one frequency amplitude characteristic of the band division filter has a predetermined non-flat characteristic within the pass band, and at least one frequency amplitude characteristic of the synthesis filter has a predetermined non-flat characteristic within the pass band, and further, An image encoding and transmitting device characterized in that a PCM encoder with a quantization noise shaping filter is used as a PCM encoder. 3. The PCM encoder with a quantization noise shaping filter comprises: an adder that receives the band-divided signal as one input; a PCM encoder that receives the output signal of the adder;
A local PCM decoder that decodes the output signal of the M encoder, a subtracter that subtracts the output signal of the local PCM decoder and the output signal of the adder, and an output signal of the subtracter. 3. The image encoding and transmitting apparatus according to claim 2, further comprising a quantization noise shaping filter which serves as an input and supplies its output to the other input of the adder.