JP2938655B2 - Subband division filter and subband synthesis filter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sub-band dividing filter and a sub-band synthesizing filter used for digitally encoding a video signal for transmission or recording.

[0002]

2. Description of the Related Art In recent years, televisions have been actively improved in image quality. As one of them, research and development of high-definition television have been remarkable (for example, the literature “Special feature high-definition television”, Journal of the Institute of Television Engineers of Japan, Vol. 36, No. 10
No. 1982). However, since a video signal of a high-definition television has a large amount of image information as compared with a current television system, a high-efficiency encoding technique is indispensable for efficient transmission or recording.

[0003] One of the effective high-efficiency coding methods for high-definition television is a sub-band coding method (for example, a document "Sub-band extrapolation / interpretation coding method"; Shu, pp. 457-458,
1989). The sub-band coding method is described in J. D. Johnston's Quadrature Mirror Filter (QMF)
Such as "A Filter Family for Use in Quadrature Mirror Filter Banks"("A Filter Family for Use
in QuadratureFilter Banks "in Proc. 1
980 ICASPSP, pp. 291 to 294), the video signal is divided into a plurality of different frequency bands, and encoding, quantization, and the like suitable for each frequency band are performed. The sub-band division filter is a filter for dividing into a plurality of different frequency bands, and the sub-band synthesis filter is a filter for synthesizing a plurality of different frequency bands.

Hereinafter, an example of the above-described conventional sub-band division filter and sub-band synthesis filter for a video signal will be described with reference to the drawings. FIG. 7 shows a block diagram of a conventional subband division filter and subband synthesis filter.

In the block diagram of the sub-band division filter shown in FIG. 7A, reference numeral 200 denotes an input terminal of a video signal; 210, a horizontal division filter for dividing the image in the spatial frequency domain in the horizontal direction; Switches for down-sampling at a period 2T, which is twice the sampling period T of the input video signal, 220 and 230 are vertical division filters for dividing in the vertical direction, 221, 222, 231,
Reference numeral 232 denotes a switch that performs downsampling at a cycle 2T _H that is twice the cycle T _H of one horizontal line.
Reference numerals 3 and 204 denote output terminals for the divided video signals.

In the block diagram of the sub-band synthesizing filter shown in FIG. 7B, 301, 302, 303, 3
04 is an input terminal of the divided video signal, 310, 320
Is a vertical synthesis filter for synthesizing in the vertical direction of the spatial frequency domain of the image, 311, 312, 321 and 322 have a period T _H
, A horizontal synthesizing filter for synthesizing in the horizontal direction, 331 and 332 a cycle T
A switch 300 for upsampling, and an output terminal 300 for the synthesized video signal.

The conventional sub-band division filter and sub-band synthesis filter configured as described above will be described below.
The operation will be described with reference to FIGS. 7, 8, and 9. First, a conventional sub-band division filter will be described with reference to FIG.

The video signal applied to the input terminal 200 is
First, the image is divided by the horizontal division filter 210 into two bands in the horizontal direction of the spatial frequency region of the image. That is, a horizontal low-frequency component of the image (hereinafter, referred to as an L component)
And a high frequency component (hereinafter, referred to as an H component), and are down-sampled by the switches 211 and 212. After that, the L component and the H component are respectively divided into two bands in the vertical direction of the spatial frequency domain of the image by the vertical division filters 220 and 230, and the switches 221 and
After being downsampled at 22, 231, 232,
The LL component, the LH component, the HL component, and the HH component are output to output terminals 201, 202, 203, and 204, respectively.

FIG. 8 shows a conceptual diagram of the divided bands. In FIG. 8, the horizontal axis is the horizontal frequency in the spatial frequency region of the image, and the vertical axis is the vertical frequency, and the video signal is divided into four regions on this plane. That is, a low-frequency component in the horizontal direction and a low-frequency component in the vertical direction (hereinafter, referred to as an LL component), a low-frequency component in the horizontal direction and a high-frequency component in the vertical direction (hereinafter, referred to as an LH component), At the high frequencies in the horizontal direction,
And a low-frequency component in the vertical direction (hereinafter, referred to as an HL component)
And a horizontal high-frequency component and a vertical high-frequency component (hereinafter, referred to as HH component).

[0010] FIG. 9 shows a video signal divided into bands as described above. In FIG. 9, reference numeral 600 denotes a video field, and 601 denotes a sub-band divided video signal output to the output terminals 201, 202, 203, and 204 of the sub-band division filter of FIG. is there.
The video signal input to the input terminal 200 of the sub-band division filter includes all the pixels at the lattice points in FIG.
The video signal output to each of the output terminals 201, 202, 203, and 204 is down-sampled in the horizontal direction to half the number of pixels in the horizontal direction and down-sampled to half the number of pixels in the vertical direction. I have.

Next, a conventional subband synthesis filter will be described with reference to FIG. The LL component, LH component, HL component, and HH component of the band-divided video signal are input terminals 301, 302, 303, 304, respectively.
Given to. The LL component and the LH component are connected to the switch 31
After up-sampling at 1 and 312, the image is synthesized by the vertical synthesis filter 310 in the vertical direction of the spatial frequency domain of the image, and the L component is obtained. HL component and HH component
After being up-sampled by the switches 321 and 322, they are synthesized by the vertical synthesis filter 320 to obtain an H component. Next, the L component and the H component are supplied to the switch 331,
After up-sampling at 332, the image is synthesized in the horizontal direction by the horizontal synthesis filter 330, and a synthesized video signal is obtained at the output terminal 300.

[0012]

However, in the above-described conventional configuration, the down-sampling is performed after the filtering in the divided filter, and the up-sampling is performed before the filtering in the synthesis filter. Will be needed. The horizontal division filter 210 needs to end the filtering operation for all the input samples before the next sample is input. Similarly, the vertical division filters 220 and 23
In the case of 0, downsampling is performed after the horizontal division filter 210, and the filtering operation needs to be completed before the next sample is input for a video signal having a half number of samples. That is, in the sub-band division filter, the filtering operation required for the horizontal division must be completed within the sampling period of the input video signal, and the filtering operation required for the vertical division must be completed within twice the sampling period. Need to complete. Similarly, the subband synthesis filter needs to be completed within twice the sampling period for vertical synthesis and within the sampling period for horizontal synthesis.

Particularly, in a sub-band splitting filter and a sub-band synthesizing filter for a video signal of a high-definition television, since the sampling period is short, very high-speed signal processing is required for each filtering operation. Therefore, in order to realize this high-speed signal processing, high-speed logic elements that are generally unsuitable for IC because of high power consumption must be used, and reduction of circuits using these high-speed logic elements has been a problem. .

In view of the above problems, the present invention provides a sub-band splitting filter and a sub-band synthesizing filter that realize the same signal processing with a low-speed signal processing circuit without using a circuit for performing high-speed signal processing. It is.

[0015]

In order to solve the above-mentioned problems, a sub-band division filter according to the present invention uses a spatial frequency of an image by using n (n ≧ 2) sample points in the horizontal direction of a video signal as a processing unit. A first band division filter that divides the region into a plurality of frequency bands in the horizontal direction and outputs n sample points in parallel for each frequency band, and m (m ≧ 2) video signals
A second band division filter that divides the sample line into a plurality of frequency bands in the vertical direction of the spatial frequency domain of the image as a processing unit, and rearranges the n sample points in the horizontal direction into m sample lines Means.

The subband synthesizing filter of the present invention receives a video signal divided into a plurality of frequency bands in the spatial frequency domain of an image, and processes m (m ≧ 2) sample lines of the video signal as a processing unit. A first band synthesis filter that performs band synthesis in the vertical direction of the spatial frequency domain of the image and n (n ≧ 2) sample points in the horizontal direction of the video signal are input in parallel, and the n sample points are The processing unit includes a second band synthesis filter that performs band synthesis in the horizontal direction of the spatial frequency domain of the image, and a reordering unit that rearranges m sample lines into n sample points in the horizontal direction.

[0017]

With the above arrangement, in the sub-band splitting filter of the present invention, the video signal output from the band splitting filter that splits the band in the horizontal direction of the image to the reordering unit has a half rate for downsampling. In addition to this, since this band division filter outputs n filtering results in the horizontal direction of the image in parallel, the original 2
The rate becomes 1 / n. That is, the time is 2nT, which is 2n times the sampling period T of the video signal. Then, for the n horizontal video signals of the image input in parallel to the rearranging means every time 2nT, m vertical images of the image are output in parallel every 2mT, and the bandwidth is set in the vertical direction of the image. This is input to the band division filter to be divided.

Similarly, in the sub-band synthesizing filter of the present invention, the video signals output from the band synthesizing filter for performing band synthesis in the vertical direction of the image to the rearranging means are m video signals in the vertical direction of the image. , Are output in parallel every 2 mT. Then, the video signals input to the rearrangement means are output in parallel in the horizontal direction of the image every 2nT, and are input to the band synthesis filter that performs band synthesis in the horizontal direction of the image.

As described above, the processing time of the video signal is time 2
By setting every nT or 2mT, a parallel processing filtering algorithm can be adopted, and a subband division filter or a subband synthesis filter can be realized by a low-speed signal processing circuit.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A subband division filter and a subband synthesis filter according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a block diagram of a sub-band division filter and a sub-band synthesis filter of the present embodiment.

In the block diagram of the sub-band division filter shown in FIG. 1A, reference numeral 10 denotes an input terminal of a video signal;
1 is a horizontal division filter, 2 is a means for rearranging video signals,
Reference numerals 3 and 4 denote vertical division filters, and reference numerals 11, 12, 13 and 14 denote output terminals for the divided video signals. In the block diagram of the sub-band synthesis filter of FIG. 1B, 21, 22, 23, and 24 are input terminals of band-divided video signals, 5 and 6 are vertical synthesis filters, and 7 is an array of video signals. Replacement means, 8 is a horizontal synthesizing filter, and 20 is an output terminal for a synthesized video signal.

Regarding the sub-band division filter and the sub-band synthesis filter of the present embodiment configured as described above,
Hereinafter, the operation will be described using (FIG. 1), (FIG. 2), (FIG. 3), (FIG. 4), (FIG. 5), and (FIG. 6). First, the sub-band division filter of FIG. 1A will be described.

First, the video signal supplied to the input terminal 10 is first divided by the horizontal division filter 1 into two bands in the horizontal direction of the spatial frequency domain of the image.
The components are obtained. Here, to explain the configuration of the horizontal division filter 1, the algorithm will be described using equations. When the video signal is x and the number of taps N of the horizontal division filter used for horizontal filtering is N = 4N ′ (N ′ = 1, 2, 3,...)
.), The filter coefficients are h ₀ , h ₁ ,..., H _N−1 , and the number of horizontal pixels P in the coding region of the image is P = 4P ′ (P ′ = 1, 2, 3,.
..), The outputs L and H of the low-pass filter and the high-pass filter are given by the following (Equation 1) and (Equation 2) in consideration of the properties of the FIR filter.

[0024]

(Equation 1)

[0025]

(Equation 2)

Here, the suffix j represents the pixel number in the horizontal direction, and j = 0, 1,..., P / 2-1. The number of pixels is halved due to downsampling. When (Equation 1) and (Equation 2) are decomposed into two terms, respectively, the following (Equation 3) and (Equation 4) are obtained.

[0027]

(Equation 3)

[0028]

(Equation 4)

Next, (Equation 3) and (Equation 4) are respectively given by j =
2j ′ (j ′ = 0, 1,..., P / 4−1), that is, the case of an even-numbered sample (hereinafter, referred to as an even-numbered sample) and j = 2j ′ + 1, that is, the odd-numbered sample (hereinafter, , An odd sample), and when divided into four terms, the following (Equation 5), (Equation 6), (Equation 7), and (Equation 8) are obtained.

[0030]

(Equation 5)

[0031]

(Equation 6)

[0032]

(Equation 7)

[0033]

(Equation 8)

That is, in (Equation 5), (Equation 6), (Equation 7), and (Equation 8), eight different sum-of-product terms appear, and each sum-of-product term is (N / 4) term The result is the addition.

Filter coefficients are calculated by using four consecutive video signals x for calculating L and H of even samples, and four video signals shifted by two pixels from the calculation for calculating L and H of odd samples. The result of filtering is obtained by calculating the sum of products with
Here, considering that two components for two samples after 2: 1 downsampling are obtained, the calculation of the eight different terms necessary for calculating the L component and the H component is performed by eight processing units. When the calculation is performed independently by the (filter), the product-sum calculation for each (N / 4) term may be completed within 4T. Here, the time T is a sampling period of the input video signal.

The above (Equation 5), (Equation 6), (Equation 7),
FIG. 2 shows a block diagram of an example of a circuit for realizing the filtering of (Expression 8). In FIG. 2, reference numeral 30 denotes an input terminal of a video signal, 40 denotes a shift register, and 41 and 4.
2, 43, 44, 45, 46, 47, and 48 are switches that open and close simultaneously with a period of 4T, 49, 50, 51, 52, 5
3, 54, 55, 56 are filters, 57, 58, 59,
60, 61, 62, 63, and 64 are arithmetic units, 31, 32,
33 and 34 are output terminals for the divided video signals.

The video signals supplied to the input terminal 30 are sequentially input to the shift register 40,
For the output of 0, the values of the past six pixels are obtained. This is because the oldest four pixels are used to calculate the filtering results L and H of the even samples, and the new four pixels are used to calculate the L and H of the odd samples. Then, the values of these six pixels are determined by the switches 41, 42, 43, 4
4, 45, 46, 47, 48 perform 4: 1 downsampling, and filters 49, 50, 51, 52, 5
3, 54, 55 and 56. Here, the filter 49 is in the first term of (Equation 5), the filter 50 is in the second term of (Equation 6), the filter 51 is in the third term of (Equation 5), and the filter 52 is in the (Equation 6). In the fourth term, the filter 53 is in the first term of (Equation 7), the filter 54 is in the second term of (Equation 8), the filter 55 is in the third term of (Equation 7), and the filter 56 is
Corresponds to the fourth term of (Equation 8).

Next, the results of the filters 49 and 51 are output by an operator 57, the results of the filters 50 and 52 are output by an operator 58, the results of the filters 53 and 55 are output by an operator 59, and the results of the filters 54 and 56 are output. Are added by the arithmetic unit 60. Also,
The results of the calculators 57 and 58 are added in the calculator 61, and this is a low-frequency output (hereinafter, referred to as “Les”) of even-numbered samples in the horizontal direction. The results of the computing units 57 and 58 are subtracted by the computing unit 62, which becomes a high-frequency output (hereinafter, Hes) of even-numbered samples in the horizontal direction.

Similarly, the results of the computing units 59 and 60 are added by the computing unit 63, and the result is a low-frequency output (hereinafter, referred to as Los) of an odd sample in the horizontal direction. Arithmetic units 59, 60
Is subtracted by the arithmetic unit 64, and the result is a high-frequency output (hereinafter referred to as Hos) of odd-numbered samples in the horizontal direction. The above computing units 57, 58, 59, 60, 61, 62, 6
The processing by 3 and 64 corresponds to the operation of each of the product-sum terms of (Equation 5), (Equation 6), (Equation 7), and (Equation 8). As described above, Les and Los are obtained as L components, and He and Hos are obtained as H components.

Next, returning to (A) of FIG. 1, the description of the sub-band division filter will be continued. Horizontal split filter 1
The L component and the H component obtained by the above are rearranged in the video signal by the rearranging means 2 so that the later vertical filtering can be easily performed. The operation of the sorting means 2 is
This will be described with reference to FIG.

In FIG. 4, reference numeral 100 denotes a video field. The reordering means 2 alternately outputs a pair (Les, Los) of the horizontal filtering results of the even-numbered sample and the odd-numbered sample, which are simultaneously obtained as a time series in the horizontal direction, a pair of Les and a pair of Los for two lines. Sort as you do. For example, in FIG. 4, if the horizontal filtering result of the 2kth line is written so as to be represented by a subscript 2k, (L _{2k, 0} , L _{2k, 1} ), (L _{2k, 2} , L _{2k, 3} ),
..., ( _{L2k, P / 2-4} , _{L2k, P / 2-3} ), ( _{L2k, P / 2-2} ,
L _{2k, P / 2-1} ), (L _{2k + 1,0} , L _{2k + 1,1} ), ...
..., the _{(L 2k + 1, P /} 2-2, L 2k + 1, P / 2-1) Result pair of filtering input during the time 2T _H so on,
( _{L2k, 0} , _{L2k + 1,0} ), ( _{L2k, 1} , _{L2k + 1,1} ), ...
・, (L _{2k, P / 2-2} , L _{2k + 1, P / 2-2} ), (L _{2k, P / 2-1} , L
_{2k + 1, P / 2-1)} over time 2T _H so on output.
Although the above description has been made for the L component, the same applies to the H component. The pairs (Hes, Hos) are rearranged so that Hes pairs and Hos pairs for two lines are output alternately. Thereafter, even lines L _{2k, i} (i = 0,1,..., P / 2-1) of the low-pass filtering result after rearrangement are Lel,
The odd line L _{2 k + 1, i} is written as Lol. Similarly, the even-numbered lines resulting from the high-pass filtering are written as Hel, and the odd-numbered lines are written as Hol.

As described above, a video signal input as a pair of an even sample and an odd sample in each band is output as a pair of an even line and an odd line. Since the input video data is raster-scanned, the vertical filter changes the video data to be filtered every 2T, unlike the horizontal filter. However, by rearranging the video data described above, the video data to be filtered changes every time 4T.

Returning to (A) of FIG. 1 again, the description of the sub-band division filter will be continued. The L component and the H component rearranged by the rearranging means 2 are respectively divided into two bands in the vertical direction of the spatial frequency domain of the image by the vertical division filters 3 and 4, respectively, and the LL component, the LH component,
HL component and HH component are obtained. Here, in order to explain the configuration of the vertical division filters 3 and 4, a filtering algorithm will be described using equations. The video signal is x, the number of taps M of the vertical division filter used for the vertical filtering process is M = 2M ′ (M ′ = 1, 2, 3,...), And the filter coefficients are h ′ ₀ , h ′ ₁ ,. , H ′ _M−1 , and the number of vertical pixels Q in the coding area of the image is represented by Q = 2Q ′ (Q ′ = 1,
(2, 3,...), The outputs L and H of the low-pass filter and the high-pass filter are (Equation 9) and (Equation 10), respectively, in consideration of the properties of the FIR filter.

[0044]

(Equation 9)

[0045]

(Equation 10)

Here, the subscript j represents the pixel number in the vertical direction, and j = 0, 1,..., Q / 2-1. The number of pixels is halved due to downsampling. When QMF is used for the vertical division filter, the symmetry of the filter coefficient of QMF, that is,

[0047]

[Equation 11]

From Equations 9 and 10, the following equations are obtained.

[0049]

(Equation 12)

[0050]

(Equation 13)

Is as follows. That is, (Equation 12) and (Equation 1)
In 3), a product-sum calculation of each M / 2 term appears.
In (Equation 12), the sum of the video signals appears, and in (Equation 13), the term of the difference between the video signals appears. However, these two video signals x are always a pair of an even-numbered line and an odd-numbered line.

FIG. 5 shows a block diagram of an example of a circuit for realizing the filtering of (Expression 12) and (Expression 13). In FIG. 5, 150 and 151 are input terminals for video signals, 160, 161, 162, 163, and 16 respectively.
4,165,166,167,168,169, the delay element of time 2T _H, 180,181,182,183,
184, 185, 186, 187, 188, 189 are multipliers, 170, 171, 172, 173, 174, 17
5, 176, 177, 178, 179 are arithmetic units, 19
0, 191, 192, 193, 194, 195, 19
6, 197 are adders; and 152, 153 are output terminals of the filter.

The input terminal 150 is supplied with the video signal of the even-numbered line Lol, and the input terminal 151 is supplied with the video signal of the odd-numbered line Lel, and these and the delay elements 160, 161, 162, 16
3, 164, 165, 166, 167, 168, 169
The value of the past (M-1) line obtained by
70, 171, 172, 173, and 174 add, and arithmetic units 175, 176, 177, 178, and 179 subtract. The addition or subtraction corresponds to the addition or subtraction of the video signals of (Equation 12) and (Equation 13), respectively. The pair of values to be added or subtracted is the current line, the oldest line, and 1 The previous line, the second oldest line, and so on.

The operation units 170, 171, 172, 1
The results obtained at 73 and 174 are output to multipliers 180 and 18
Multiplied by 1, 182, 183, 184, the results are all added by arithmetic units 190, 191, 192, 193 and output to output terminal 152 as a result of a low-pass filter.

On the other hand, arithmetic units 175, 176, 177, 1
The results obtained at 78 and 179 are output to multipliers 185 and 18
6, 187, 188, and 189, the results of which are all added by the arithmetic units 194, 195, 196, and 197, and output to the output terminal 153 as a result of the high-pass filter. As described above, a low-frequency component and a high-frequency component in the vertical direction are obtained.

Returning to FIG. 1A, the description of the sub-band division filter will be continued. The low band video signal obtained by the vertical division filter 3 corresponds to the LL component, and the high band corresponds to the LH component. The high-frequency video signal obtained by the vertical division filter 4 corresponds to the LH component, and the high-frequency component corresponds to the HH component.
These LL component, LH component, HL component, and HH component are output to output terminals 11, 12, 13, and 14, respectively.

Next, the sub-band synthesis filter of FIG. 1B will be described. First, the input terminal 21,
The LL component and LH component of the video signal respectively given to 22 are input to the vertical synthesizing filter 5, and the even line Lel and the odd line Lol of the L component are obtained. Further, the HL component and the HH component respectively supplied to the input terminals 23 and 24 are input to the vertical synthesizing filter 6, and an even line Hel and an odd line Hol of the H component are obtained.

Here, in order to explain the configuration of the vertical division filters 5 and 6, a filtering algorithm will be described using equations. Since the synthesis of the L component and the synthesis of the H component are the same procedure, only the L component, that is, the vertical synthesis filter 5 will be described below. Generally, the same coefficients as those used in the vertical division filter are used for the low-pass filter and the high-pass filter used for the vertical synthesis filter. Therefore, the video signal is x, and the number of taps M of the vertical synthesis filter used for the vertical filtering process is M = 2M ′ (M ′ =
1, 2, 3,...), And filter coefficients are h ′ ₀ , h ′ ₁ ,.
·, H ′ _M−1 , the number of vertical pixels Q of the image coding area
= 2Q ′ (Q ′ = 1, 2, 3,...).

The result L of the filtering by the low-pass filter is as follows: when U is the LL component and V is the LH component, at the even line (Lel) and at the odd line (Lel),

[0060]

[Equation 14]

[0061]

(Equation 15)

Is obtained. Here, the subscript j of L = 2n, 2n
+1 represents a pixel number in the vertical direction, and j = 0, 1,.
, Q / 2-1, and element 0 on the right side is a value inserted at the time of upsampling.

From (Equation 14) and (Equation 15),

[0064]

(Equation 16)

[0065]

[Equation 17]

Is obtained. Similarly, the result H of the filtering by the high-pass filter is as follows at the even line (Hel) and at the odd line (Hol).

[0067]

(Equation 18)

[0068]

[Equation 19]

Is obtained. Therefore, the reference "Application of Quadrature Mirror Filters to Split Band Voice Coding Schemes"
s To Split Band Voice CodingSchemes ", in Pro
c. 1977 ICASSP, pp. 191 to 195)
Than,

[0070]

(Equation 20)

[0071]

(Equation 21)

There is a relationship as follows. Here, x is a synthesized video signal, and subscripts j = 2n, 2n + 1 of x represent pixel numbers in the vertical direction, and j = 0, 1,..., Q-1.

From (Equation 20) and (Equation 21), a video signal can be obtained by doubling each right side. That is,
The values of the even lines and the odd lines of the video signal are obtained from the sum and difference of the filtering results L and H. The above (Equation 2)
FIG. 6 shows a block diagram of an example of a circuit for realizing the filtering of (0) and (Equation 21). (FIG. 6)
700 and 701 are input terminals for video signals, 710 and 71
1, 712, 713, 714, 715, 716 are time 2
T _H delay elements, 720, 721, 722, 723, 7
24, 725, 726, and 727 are operation units;
1,732,733,734,735,736,73
7, 750, 751 are multipliers, 740, 741, 74
2, 743, 744, 745 are adders, 702, 703
Is the output terminal of the filter.

The input terminals 700 and 701 are respectively LL
Component, LH component, and these, and a delay element 710,
The value of the past (M-1) line obtained by 711, 712, 713, 714, 715, 716 is
At 0, 721, 722, and 723, addition, an arithmetic unit 724,
In 725, 726, and 727, the value is subtracted. This addition or subtraction corresponds to addition or subtraction of (U and V) between the components of (Equation 20) and (Equation 21), respectively, and the pair of values to be added or subtracted is the current line and the previous one. The lines are ...

The results obtained by the arithmetic units 720, 721, 722, and 723 are multiplied by multipliers 730, 731, 732, and 73.
3 and the result is added to adders 740, 741, 74
2 are all added, and the multiplier 750 multiplies by 2 times, and outputs the low band Lel of the even line to the output terminal 703. On the other hand, computing units 724, 725, 726, 727
Are obtained by the multipliers 734, 735, 736, 7
37, and the result is added to adders 743, 744, 7
45, all are added, and the multiplier 751 multiplies by two, and the low band Lol of the odd line is output to the output terminal 702. As described above, the synthesis in the vertical direction is performed, and Lel and Lol of the L component are obtained. With the same configuration, Hel and Hol of the H component are obtained.

The L obtained by the vertical synthesis filters 5 and 6
The component and the H component are rearranged in the video signal by the rearranging means 7 so that horizontal filtering can be performed later.
The operation of the rearranging means 7 is just opposite to that of the rearranging means 5, and will be described with reference to FIG.

The rearranging means 7 converts pairs of video signals (Lel, Lol) of two lines of even-numbered lines and odd-numbered lines in the vertical direction which are simultaneously obtained as a result of vertical filtering.
Rearrangement is performed so that a pair (Les, Los) of even-numbered samples and odd-numbered samples, which is a time series in the horizontal direction, can be obtained at the same time. For example, in FIG. 5, (L _{2k, 0} , L _{2k + 1,0} ),
(L _{2k, 1} , L _{2k + 1,1} ),..., (L
_{2k, P / 2-2} , _{L2k + 1, P / 2-2} ), ( _{L2k, P / 2-1} , L
_{2k + 1, P / 2-1)} so on, a video signal to be inputted during the time _{2T H, (L 2k, 0} , L 2k, 1), (L 2k, 2, L
_{2k, 3} ), ..., ( _{L2k, P / 2-4} , _{L2k, P / 2-3} ),
(L _{2k, P / 2-2} , L _{2k, P / 2-1} ), (L _{2k + 1,0} ,
L _{2k + 1,1} ), ..., (L _{2k + 1, P / 2-2} , L
_{2k + 1, P / 2-1)} over time 2T _H so on output.
Although the above description has been made with respect to the L component, the same applies to the H component, and the data is rearranged so that a pair (Hes, Hos) can be simultaneously obtained from a pair (Hel, Hol).

Returning to FIG. 2 again, the description of the subband synthesis filter will be continued. The L component and the H component rearranged by the rearranging means 7 are synthesized by the horizontal synthesis filter 8 in the horizontal direction of the spatial frequency domain of the image, and the synthesized video signal is output to the output terminal 20. Here, to explain the configuration of the horizontal synthesis filter 8, the algorithm will be described using mathematical expressions. In general, the same filter used in the horizontal division is used as the horizontal synthesis filter.

Then, the video signal is represented by x, and the number of taps N of the horizontal synthesis filter used for the horizontal filtering processing is represented by N =
4N ′ (N ′ = 1, 2, 3,...)
h ₀ , h ₁ ,..., h _N−1 , and the number of horizontal pixels P in the coding area of the image are P = 4P ′ (P ′ = 1, 2, 3,...). Result L of filtering by low-pass filter
Is

[0080]

(Equation 22)

[0081]

(Equation 23)

[0082]

(Equation 24)

[0083]

(Equation 25)

Is obtained. Here, element 0 is a value inserted at the time of upsampling. Similarly, the result H of the filtering by the high-pass filter is

[0085]

(Equation 26)

[0086]

[Equation 27]

[0087]

[Equation 28]

[0088]

(Equation 29)

## EQU11 ## Therefore,

[0090]

[Equation 30]

[0091]

(Equation 31)

[0092]

(Equation 32)

[0093]

[Equation 33]

Is obtained. Here, (Equation 30), (Equation 3)
1), (Equation 32), and (Equation 33) are equal to their left sides, as in the case of the vertical synthesis filter.

From the sum and difference between the L component and the H component, four consecutive time-series decoded video signals can be obtained. Here, (Equation 30), (Equation 31), (Equation 32), (Equation 33)
, L appearing in addition and subtraction in eight different product-sum terms
The pixel numbers of the component (U) and the H component (V) are the same, each being four consecutive components. Here, considering that the video data of 4 samples after the 1: 4 up-sampling is obtained, the calculation of the eight different terms necessary for the calculation of the video data of 4 samples is performed by eight processings. When calculating independently in units (filters), it is only necessary to complete the product-sum calculation of (N / 4) terms within time 4T. Here, the time T is a sampling period of the output video signal.

(Equation 30), (Equation 31), (Equation 3)
FIG. 3 shows a block diagram of an example of a circuit for implementing the filtering of (2) and (Equation 33). (FIG. 3)
70, 71, 72, 73 are input terminals for video signals, 79,
80, 81 are delay elements of time 4T , 82, 83, 84,
85, 86, 87, 88, 89 are filters, 75, 7
6, 77, 78, 90, 91, 92, 93 are arithmetic units, 9
Reference numeral 4 denotes a shift register, and 74 denotes an output terminal for the synthesized video signal.

The video signals Les and Hes applied to the input terminals 70 and 71 are respectively added and subtracted by arithmetic units 75 and 76, and the video signals Los and Hos provided to the input terminals 72 and 73 are respectively applied to the arithmetic unit 77. At 78, they are respectively added and subtracted. The outputs of the calculators 75, 76, 77, 78, along with their values and the values delayed by the time 4T by the delay elements 79, 80, 81, together with the filters 82, 83, 84,
85, 86, 87, 88, 89. here,
The first term of (Equation 30) is to the filter 82, the second term is to the filter 86, the first term of (Equation 31) is to the filter 88,
The term is the filter 84, and the first term of (Equation 32) is the filter 8
7, the second term corresponds to the filter 83, the first term in (Equation 33) corresponds to the filter 85, and the second term corresponds to the filter 89.

Next, the results of the filters 82 and 86 are output by an operator 90, the results of the filters 84 and 88 are output by an operator 91, the results of the filters 83 and 87 are output by an operator 92, and the filter 8
The results of 5 and 89 are added in an arithmetic unit 93 and multiplication is performed twice. The adders 90, 91, 92, 93
The processing by (Equation 30), (Equation 31), (Equation 32),
This corresponds to the operation of each product sum term in (Equation 33). The results of the arithmetic units 90, 91, 92, and 93 are input to the shift register 94, and are output from the output terminal 74 in this order. Thus, a combined video signal is obtained.

As described above, by performing two processes in the horizontal direction and two processes in the vertical direction in parallel, the video data can be processed every 4T. In the above, the processing in the horizontal direction and the processing in the vertical direction are performed in parallel, but may be performed in two or more.

[0100]

As described above, the sub-band division filter of the present invention provides n filtering results for each frequency band obtained by dividing the image in the horizontal direction using n sample points in the horizontal direction as a processing unit. , A band dividing filter that outputs m sample points in the vertical direction of the image as a processing unit, and a band dividing filter that outputs n sample points in the horizontal direction of the image in the vertical direction. By providing a reordering means for reordering the sample points, the processing time of the video signal becomes every 2 nT or 2 mT, and by adopting the algorithm of the parallel processing filtering,
The sub-band division filter can be realized by a low-speed signal processing circuit.

The sub-band synthesizing filter according to the present invention includes a band synthesizing filter for performing band synthesis in the vertical direction using m sample points in the vertical direction of the image as a processing unit, and n sample points in the horizontal direction of the image. By providing a band synthesis filter that performs band synthesis in the horizontal direction as a processing unit and a rearranging unit that rearranges m sample points in the vertical direction of the image into n sample points in the horizontal direction, the processing time of the video signal is increased. Is time 2
It becomes every nT or every 2 mT, and by adopting the algorithm of the parallel processing filtering, the subband synthesis filter can be realized by a low-speed signal processing circuit.

[Brief description of the drawings]

FIG. 1A is a block diagram illustrating a subband division filter according to an embodiment of the present invention. FIG. 1B is a block diagram illustrating a subband synthesis filter according to an embodiment of the present invention.

FIG. 2 is a block diagram showing details of a horizontal division filter in FIG. 1;

FIG. 3 is a block diagram showing details of a horizontal synthesis filter in FIG. 1;

FIG. 4 is an explanatory diagram of the operation of the sorting means in FIG. 1;

FIG. 5 is a block diagram showing details of a vertical division filter in FIG. 1;

FIG. 6 is a block diagram showing details of a vertical synthesis filter in FIG. 1;

FIG. 7A is a block diagram showing a configuration of a conventional subband division filter. FIG. 7B is a block diagram showing a configuration of a conventional subband synthesis filter.

FIG. 8 is a schematic diagram illustrating frequency division of a conventional subband division filter.

FIG. 9 is a schematic diagram of a video signal output from a conventional subband division filter.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Horizontal division filter 2 Rearrangement means 3 Vertical division filter 4 Vertical division filter 5 Vertical synthesis filter 6 Vertical synthesis filter 7 Rearrangement means 8 Horizontal synthesis filter 10 Input terminal 11 Output terminal 12 Output terminal 13 Output terminal 14 Output terminal 20 Output terminal 21 input terminals 22 input terminals 23 input terminals 24 input terminals

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Kazunari Irie 1-1-6 Uchisaiwaicho, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Ryozo Kishimoto 1-1-6 Uchisaiwaicho, Chiyoda-ku, Tokyo No. Nippon Telegraph and Telephone Corporation (58) Fields surveyed (Int. Cl. ⁶ , DB name) H04N 1/41-1/419 H04N 7/24-7/68

Claims (2)

(57) [Claims] An image signal is input, and the image signal is divided into a plurality of frequency bands in a horizontal direction of a spatial frequency domain of an image by using n (n ≧ 2) sample points in a horizontal direction of the image signal as a processing unit. First band dividing means for dividing and outputting n sample points in parallel for each frequency band; and converting the video signal into an image by using m (m ≧ 2) sample lines of the video signal as a processing unit. A second band splitting means for splitting into a plurality of frequency bands in the vertical direction of the spatial frequency domain;
A rearrangement unit for rearranging the sample points into m sample lines, an output of the first band division unit is connected to an input of the rearrangement unit, and an output of the rearrangement unit is the second A sub-band division filter connected to an input of the band division means. 2. An image signal divided into a plurality of frequency bands in a spatial frequency domain of an image is input, and m
First band synthesizing means for performing band synthesis of the video signal in the vertical direction of the spatial frequency domain of the image by using (m ≧ 2) sample lines as a processing unit; and n (n ≧ 2) of the video signals in the horizontal direction. ) Are input in parallel, a second band synthesizing unit for band-synthesizing the video signal in the horizontal direction of the spatial frequency domain of the image using the n sample points as a processing unit, and m sample lines. A rearrangement unit for rearranging the data into n sample points in the horizontal direction, wherein an output of the first band synthesis unit is connected to an input of the rearrangement unit, and an output of the rearrangement unit is the second band A subband synthesis filter connected to the input of the synthesis means.