JPH02132988A - Digital y/c separating method - Google Patents

Abstract

PURPOSE:To attain the complete Y/C separation by obtaining an infinitesimal adjoining virtual sampled value from plural sampled values. CONSTITUTION:From three infinitesimal adjoining sampled values to continuously store the sampled continuous sampled value string at a prescribed period and operate a sampling function desirably to the sampled value string and a composite signal constructive mood, a first coefficient string concerning a luminance signal Y and second and third coefficient strings are calculated. These first to third coefficient strings are selectively operated between combinations a1+b1 and a1-b1 of a sampled value string and with computing elements 8, 10 and 14 to multiple and add, a Y/C separating characteristic is obtained.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a digital Y signal that can accurately separate a composite signal consisting of a luminance signal Y and a chrominance signal C in digital quantities from a sample value sequence obtained by sampling a luminance signal Y and a chrominance signal C using a predetermined cloth. /C separation method.

[Conventional technology]

Figures 8 and 9 are examples of conventional digital Y/C separation methods.
This will be explained with a diagram.

FIG. 8 is a diagram showing sampling positions where a composite signal is sampled using a clock having a frequency four times the subcarrier frequency fsc (period T = 1/4 fsc).

When the composite signal is in the NTSC system, the composite signal E. Luminance example Y, color signal C (actually color difference signal C, =RY and 02=
BY), then it is as follows.

E=Y 10C=Y 10C. 10C2, or E = Y
10-'(R-Y)coswsct+-'-(B
-Y) sinwsct1.14
2.03 However, wsc = 2yfsc From this, assuming that the composite signal is the same for both the luminance signal and the chrominance signal, at Wset = 0° or 180°, & = Y+ +Q, Ez = Yt'Q',
Y+ = Y+', CI=Q' When WSC=90° or 270°, E+' = Yz +C2, El' = Y2' C
2' Y2 ``Y2' . C2 = C2'' (
EI E2), C2 = I(El'E2'
) FIG. 9 is a block diagram of the subordinate Y/C separation method using this calculation method. 1 is a delay element, which is a delay amount that makes the phase of the color signal T degrees 180 degrees.

This relationship exists at every 2T in the horizontal scanning direction, at the same position between scanning lines in the vertical direction, and at the same position between frames.

2 and 3 are a subtracter and an adder that perform the above calculation.

This is a digital Y/C separation method according to the prior art.

[Problem to be solved by the invention]

However, in the prior art described above, the assumption Y = Y', C
= C'. This assumption holds true when the density and color are the same in the horizontal and vertical directions.

Also, between frames, this is only for still images. Therefore, the assumption does not hold true for images with density changes, color changes, or movement. As a result, conventional techniques have the problem of producing unsightly images due to color frizz, color mixing, and color blurring.

The present invention is intended to solve these problems, and its purpose is to calculate virtual sample values from a plurality of consecutive sample sequences and derive a relationship that allows it to be assumed that the luminance signal and color signal are the same. The object of the present invention is to provide a method for nearly perfect Y/C separation.

(Means for solving problems) The present invention has the following features.

1. A composite signal consisting of Y (luminance signal) and C (chrominance signal) is in a predetermined phase relationship with the color reference signal {Digital Y/C separation in which the signal is sampled using a unifying clock and Y/C is separated by digital quantity. In the method, the following sequence of sampled values (
all at-b al1 aO+ bO+
bb+-+, b+, 1, and second and third coefficient sequences related to C are calculated.

These first, second, and third coefficient sequences are selectively applied to the sample value sequence combinations (a+ +b+) and (a+ b+) to produce accurate Y/C using calculation means that multiplies and adds them. This is a Y/C separation method that obtains separation characteristics.

2. In the above, the Y/C is calculated by correcting the first, second, and third coefficient sequences so that no significant frequency fluctuation occurs even if the length of the sample value sequence is set to the minimum i-2. This is a separation method.

[Effect]

According to the above configuration of the present invention, since the distance between the three sample values can be made infinitely small, the unknowns of the three equations, Y2Y'=Y"=Y
It can be completely assumed that ``'', C = C' = C'' = C'''.

Therefore, complete Y/C separation is possible.

〔Example〕

To explain the embodiment of the present invention, a relational expression will be derived with reference to FIG.

In Figure 2, [ao, a++ az, '=・an], [
b. , b. . . bo] is a sample value obtained by sampling a complex I; In this example, the clock is 4 of the subcarrier frequency fsc of the NTSC system.
At twice the frequency and period, T = 1/4 fsc.

According to the sampling theorem, band fB≦W =s 4f5C=
The virtual sample value S that can be connected using 2 fsc can be expressed as follows.

sinL S=Σsa(a+,b1), where sa=H, where L is the distance to the virtual sample value to be obtained with the sample value interval (sampling clock interval) as π.

In FIG. 2, three virtual sample values L, , l. ，，
l. Then, the sample value a. , Q P between bo, 2
．． 2 It is 3 points of 10 points. P represents infinitesimal. From "1" L, I. , I+ is L = Io −k+
+'p(ao -bQ) +k+'p(at -bl)
+ +k+'(-1)'''lp(a1b+) +
+ 10k. (k o”1 ↓) p2 (ao + bo) +
- Since +k is infinitesimal, it can be expanded into a series, so (k+"2) p2(a+ +bv 1+1+"
lo +ko”p(ao bo) k12p(a
t b+) 10- +k12(1)l+lp(a
+ b+) + ten+ko (ko” -8)2p2
(ao tenbo) + − +k(k+"z')
p(a, +b1) becomes 10+.

The horizontal axis at the bottom of FIG. 2 shows the phase relationship between the composite signal and the color reference signal. In terms of the sampling function sa, this is a half movement.

ko (ao tenbo) +k+Cat +b+) +
-" +k+(a1+b1) +Here, the composite signal E=Y+ R-YB-Y. -coswsct +--s+nwsct wsct
=45°=L, 1.14 2.03
4R-Y; r B-Y. y Y 10-cos-+ -sm-1.14 4 2.03 41. ......■. l1 and 1+ are 1.

When organized using holds true.

Y k o (8 k o” 3) (ao+bo)+k+(8k12-3)(a+bl
)+ − ・− + 10k+(8 kI23) (at +b
l) + ・+ ・・・・■ RY= . )+... Y+α(R-Y) 1β(B-Y) I3 -Y = ko" (ao t)o) k+"(a b+) +kz'(a2 b2)+... Y) =1. Solving from ■', ■', and ■', however, R (a b,) = ku' (ao bo) + k+' (a b1) Furthermore, L, I. , If we rearrange Lya by proxy, we get .

As described above, the two color difference signals, the luminance signal Y and the color signal C, can be calculated using known quantities and are completely Y/C separated.

Next, the phase is determined at 135°. In this case, rewrite the name of the sample value etc. by viewing it at a 135° point. The basic equation is as follows.

B-Y=-(β R(a1・b1)-Q(a1+bl)) ・・・・・・
... [Phase] Similarly, at phase 225°, Y ``ko (8ku'' 3) (ao + bo) + k+
(8k+' 3) (a+ +b1) 10...+...
・・・・・・・・・・・・・■l RY= (R(at bI) α Q (a+ +bl) )+...10...0Y) =
1 Y −a (R − Y) + β (B − Y) = 1. phase 31
At 5°, Y”ka (8ko” 3) (ao ten bo) + k + (8
kI2 3) (a++b+)+・・+ ・・・・・
......0R ・Y = L( ・R(a.
・b・) +Q(a+ +b1)) ・・・・・・・
0α Y) =1. Solving from these and rearranging using the same method for the case of 45 degrees, Y = ko (8 ko' ・3) (ao ten bo)
10k. (8 k,” ・3) (a+ 10bl) 10...@? Above, any phase can be expressed in the same format.

With color difference signals R-Y and B-Y, R(a1-bl) and Q(a
The actual calculation is carried out keeping in mind that the addition signs of ++b1) are different in phase.

To describe the relationship between R and Q, when the density is constant and the color is the same, R
o20+3,=R2■s ”Q:us,Qts=Rl3
There is a characteristic that s-Qr2s=R315. this is,
The calculation can be confirmed by specifically substituting numerical values into the expression of the composite signal E.

Next, on the color signal axis shown in Fig. 6, 0'', 90'
, 180'. Sample value a at 270°. ” Calculate the relational expression using virtual sample values on both sides including Iu.

3) Shiju+ +k+'p2(at +b+)
k2"p2(az +bl) +k:+'p2(a3
+b3) 10...10.

Similarly, when l+ is calculated, 1. =(1 +kop2)ao -kIp(a+-b+
)-k2p(a2-b2)-kj(a3-b3)”・
"- - +k+'p2 (a1+b,) -kz"p2
(a2+bz) +k:+'p' (a3+b3)・+・
・Becomes ten.

Here, an equation is created at 0° on the color signal axis.

From these, 1- =Woao +Wi'a+ +Wia2+WJ+
a3+ − ・” +Wib+ + Wzb2+W:+
b3++...+(1 +kop2)ao +k+p(
1 +kIp)a+ +kzp(1 k2p)a2
+k. +p(1 +k3p)a3+ten+ -k+p(1
-k+p)b+ -kzp(1 +k2p)b2
k]p(1 −k:+p)b, (1 +kop2)ao +k+p(at −b+)
+k2p(a2-b2) +k3p(a3-bSimilarly to the above, using p→0, Y
) 10 = Io [phase] + , @r, ■', R -y = -Qo -y) α 1l However, α = πn, β = 藷.

[Phase], [Phase], L obtained before O, I. , l+ is rearranged as follows.

Y = ko(ao +ao) +8 k,”(at +
b+) -8k2'(a2 +b2) +8k:+'C
as+b. +) ・+8k+'( ・1)IN(a
, +b+) + 110...@)'R ・Y =L(
a.・Y)・・・・・・・・・・・・[Phase]′α b3) 2 k+(at t)+) )
Similarly, subscripts are changed sequentially from @' to 90°, 180°, and 270°.

At 90°, Y-o=Yo At 180°, YIw = Ycr RY = ' (ao Y+go)α At 270°, Y270°"Yo ・)) Number of color difference signals R-Y and B-Y If it is noted that the expression format changes alternately, the expression format will be the same throughout the entire cycle of the color signal.

Now that the preparations have been completed, FIG. 1, which is a specific embodiment of the present invention, will be explained.

In the figure, numeral 4 denotes delay elements which are input terminals for the sampled composite signal E, and (2n+1) delay elements are connected in series.

6 and 12 are 20 adders, and each sample value a is attached to the output terminal of the delay element. , al, a2・a1・an . !
:bo, b,...b...bo, 6 is (a+ +t
)+) + 12 calculates (aIb+).

7 is a coefficient multiplier which applies a coefficient y to each output of the adder 6. ,y1・
・y1...yo is multiplied. Coefficient y. ,y+
”' ...yn is yo = ko (8
ko” 3) . )/+ ”k+(8k+” 3
＞．． 9,Y+=k+(8k+'3)...

Specifically, yo =0.154, y+ =0.56
0, yz = 0.365, y3 ”0.267
, Y4 = 0.209, ys = 0.17
2...etc.

8 is a set adder that calculates the sum of each output of the coefficient multiplier 7. This set adder actually has two adders.
Since you can only add between values, do the following. When the number of inputs is (n+1), the first adder is a set adder 8 in which approximately (n+1) adders are connected in cascade.

The output of the set adder 8 is the separated luminance signal Y. Expressed using real coefficients, Y = O. l54(ao +b.) +0.560(a
l +b +) 0.365(a2+l]+) +
0.267 (a, +63) − , +yo (ao ten b
. ).

Here, we will look at the frequency characteristic of Y. Sample value (ao + b
Looking at other sample values centered around o), as a Z operating element, Y
=0.154(1+1)+〇. 560 (z+z-')
0.365(z2+z-2)〒0.267(z3+z
−3) + −・− ・+y. (zn+z-')
It can be written as

If we rearrange it as Y20.308 +1.120cos(WT) -0.
730cos(2WT) +0.534cos(3WT
) 10・・+2y,,cos(nWT) Now, n
If you make the calculation sufficiently large, you will get 4 as shown by the dotted line A in Figure 3.
A frequency characteristic having a pole at a period of Wsc=8πfsc is obtained.

In FIG. 3, the horizontal axis represents the frequency, and the vertical axis represents the amplitude of Y when the input vibration is 1.

For example, if the composite signal is in the NTSC format, fsc = 3.
The frequency included in 58Mllz Y is at most 4.5MH
z, so %T< f = 2 fsc = 7 in Figure 3
．． It is sufficient for A to have a flat characteristic up to near 16 Ml{z. Therefore, the luminance signal separated from the composite {co signal} is complete and does not contain any color signal.

Incidentally, in reality, if n is set sufficiently large, the hardware cost will increase, so it is necessary to limit it from the viewpoint of cost performance.

Therefore, an example of using cant and try with n25 is the solid line B
It is long-lasting.

WT) -8cos(4~VT) +4cos(5
WT)).

If you do this, you will get a peak at f=2.4~11Iz. f=
Although the flatness characteristic is lost near the valley in fsc, this degree is not actually a problem.

Further, by restricting in this way, the coefficients of the multiplication unit 7 are also simplified, resulting in a multiplier with an extremely simple configuration, and the cost reduction can be achieved.

Next, returning to FIG. 1, 9 is the coefficient q. , Q++ +
This is a multiplier that multiplies Q'. IO is a set adder that adds each output of the multiplier 9. Set addition f'l: Vessel 8 is Q(a+b1) =qo(ao +bo) +q+(
a+ 10bυ+qz (a2+b2) 10-10 play, release to the east.

■ ks(ks”・7)20.114.

13 and 14 are r. This is a set adder that performs an addition operation on each output of this multiplier 13 and the multiplier I2 which is multiplied by the coefficient of l rII l rn. This set adder 14 is R(at −
b+) =ru(an-bo) r+(at-b
+) +r2(a2-b2) +r3(a.+-b.
) + releases + −.

However, ru= ko'= 0.405, r+=k
+”=0.045, r20.016, r,”k3”
=0.008, r4=k4'=0.005.
rs=ks""0.003, ra"kg'
= 0.002, , , .

l1 and 15 are 1/β=1/0.3483 =2.8
This is a multiplier that multiplies by 71 times.

17 is an adder, which outputs the color difference signal B-Y-■2.871 1
R(at b1) ■2.87I Q (a, +
b+) l is calculated.

18 is also an adder, color difference signal (R - Y) /0.561
5-■2.871 1 R(a+ -b.) l■2.
871 I Q(at +b+) l is calculated. ■ is the calculated position for the color-based pregnancy signal...A No. 8 φ
and Rfa1b+) and Q(a, +b1) mean that + and 1 are determined.

The phase signal φ is added to adders 17 and 18.

l9 is a multiplier that multiplies by 0.5615, and the output of adder l8 (R - Y)/0.5615 X0.5G15 = R-
This is to make it Y.

Ru.

Then, 2.218cos(WT) -1.414cos(2W
T) 10- 10+ +2.871 X2qncos(n
WT). The frequency amplitude characteristic of this field has an imaginary part.
? ), so it is calculated using absolute values. Let n be a sufficiently large field,
The B characteristic shown in Fig. 4 is improved.

Since the color signal has only a band of at most 3MI-1z centered around f=fsc, the characteristic shown in the dotted line box A is desirable, but since it is not included in the signal, there is no problem with B.

As with luminance signals, hardware cost is an issue, so
An example in which n is limited will be explained.

Look at the wavenumber characteristics.

B Y = 1.164 (ao bo) - 0.129 (
a+ b+)+0.047(a2b2)+
+0.692(ao+l)o) +1. LO9(
a l +b l)-0.707(a2+bz) +
aQ "1) O = l, a+:z-'
, b+=z, a2-Z2, b2-z-2, Z river bank 1
WT)) /4.

C in FIG. 4 shows this characteristic.

The embodiment of the present invention shown in FIG. 1 now has complete Y/C separation, resulting in proper luminance signal Y and color difference signals R-Y and B-Y.

Next, another embodiment of the present invention in which the focus point of the virtual sample value is changed will be described with reference to FIG.

In FIG. 5, the same numbers and symbols as in FIG. 1 have the same means or meanings.

α2 = 8k2' = 0.203, (13
=8k3" =0.090 --(ao + bo)
+(a+ +b+) +...+ (al +bl)
There are things that take advantage of this. 21 is a set adder, Y−Σα
*(at+b1) is calculated.

23 is a multiplier, β, = −2 k. =0.637
, β2 = 2k2 = 0.318 , β3 = 2
k3=0.212 , , , coefficient of (a1b+)
It is something that rides on.

24 is a set adder, S-Σβ. It calculates (a, -b, ). 22 is an adder that calculates (ao Y).

Reference numeral 25 denotes a switch for switching the outputs of the inputs S and (ao Y) according to the phase signal φ.

26 and 27 are multipliers that multiply the output of the switch 25 by 1.14 and 2.03. The outputs of the multipliers 26 and 27 become color difference signals RY and 113-Y.

In the case of FIG. 5, the same frequency amplitude characteristics as those of FIGS. 3 and 4 are obtained, and perfect Y/C separation characteristics are obtained.

Finally, a simple method for calculating color signals will be explained with reference to FIG.

The luminance signal Y is calculated by the method of the present invention shown in FIG. 1 or FIG. 30 is an input terminal for Y, and 31 is an input terminal for sampled value D whose sampling phase has been confirmed.

32 is an adder which calculates D-Y. 33 is preferably a bandpass filter (B) having the frequency characteristic of A in FIG.
PF). This bandpass filter (I3-Y
) emit alternately. 35 is a T time delay element, 36 and 37
is a 2T delay element. 39 is a clock φ with a period of 2T
2 human power terminals. (B-Y) configured in this manner are individually emitted at 2T periods.

Multiplying by 1.14 and 2.03 by multipliers 26 and 27 yields color difference signals Y and BY.

34 is a delay element that matches the sum of the delay of the bandpass filter 33 and the 2T delay of the delay element 36 or 37.

In this case, the data amount of the color difference signal is halved for the luminance ia number Y. Furthermore, the reason why the bandpass filter 33 is provided is that unlike other embodiments, the same point is not calculated.
This is to reduce mixing of luminance signals.

However, although the separation characteristic shown in FIG. 7 is poor, it is characterized by low cost.

〔Effect of the invention〕

As described above, according to the present invention, from a plurality of sample values,
Since infinitely small adjacent virtual sample values are obtained, this method can assume that the luminance signals and color signals for the three {7 values are the same.

According to conventional technology, Y/C was separated under various conditions, but the effect of achieving complete Y/C separation unconditionally is revolutionary.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the digital Y/C separation method according to the present invention. The second I2I is a diagram illustrating the order of sample value sequences and virtual sample values used in the present invention. FIG. 3 is a diagram showing frequency 4PC width characteristics of a luminance signal according to the present invention. FIG. 4 is a diagram showing frequency amplitude characteristics of color signals. FIG. 5 is a block diagram of another embodiment of the digital Y/C separation method according to the present invention. FIG. 6 is a diagram showing the positions of virtual sample values adopted in FIG. 5. FIG. 7 is a long diagram showing an example in which only the luminance signal is simplified according to the present invention, and the color signal is simplified. FIG. 8 is a diagram showing an example of a bookmarking position. FIG. 9 is a diagram showing an example of a dintal Y/C separation method according to the prior art. ” Applicant Seiko Epson Co., Ltd. Agent Patent Attorney Kisaburo Suzuki and others -05! - "Fuken"i Figure 4Sjj 5 Q -480@ -90' 0a 90@ 27o@ 360@ Figure 7

Claims (2)

[Claims] (1) Digital Y, which samples a composite signal consisting of Y (luminance signal) and C (chrominance signal) using a sampling clock that has a predetermined phase relationship with the color reference signal, and separates Y/C in digital quantities.
In the /C separation method, the following sampled value sequence (-
-a_1, a_f_-_1--, a_1, a_0, b_
0, b_1, --, b_f_-_1, b_1, --) in a cascade manner at a predetermined period; and three adjacent infinitesimal sample values obtained by preferably applying a sampling function to the sample value sequence. The first coefficient string related to Y and the second and third coefficient strings related to C calculated from the composite signal constitutive equation are selectively combined with the sample value strings (a_1+b_1) and (a_1-b_1).
) A digital Y/C separation method characterized in that Y/C separation is carried out in digital quantities using arithmetic means that act on the signal. (2) In claim 1, the length of the sample value sequence is set to a minimum of i=2, and the first, second, and third coefficient sequences are modified without significantly changing the frequency characteristics. A digital Y/C separation method characterized by: