JPH03190379A - Ghost elimination device - Google Patents

Abstract

PURPOSE:To reduce the converging time by taking a difference with a ghost elimination reference signal (GCR) fetched at 4 preceding field from a current field for each field when the GCR is fetched and forming one field type for a minimum tap gain correction period. CONSTITUTION:A current fetch waveform is always stored in an area M1 and a waveform of 4 preceding field is stored in an area M5 by sending the content of a memory sequentially every time a GCR is fetched. Thus, the calculation of a GCR(SGCR) has an advantage that it is implemented fixedly by using the areas M1, M5. However, it takes time for a job to transfer the content of the memory sequentially before the GCR is fetched. Then a method of storing a signal in an area in which the oldest GCR of 4 preceding field to the current field is taken every time the GCR is fetched. Thus, the transfer from the memory area to the other memory area is omitted and the time is reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a ghost removal device that is used in a television system and removes ghosts contained in a video signal.

(Prior Art) In recent years, a reference signal for ghost removal (hereinafter referred to as OCR) for removing television ghosts has been standardized. This standard is shown in a document ("Ghost Removal Method Lecture Material J 1989.04.13, Broadcasting Technology Development Council)".

FIG. 7 shows the ghost reference signal shown in the above-mentioned document. This is called 8-field sequence OCR. This OCR includes the first, third, sixth,
WRB inserted in the 8th field (Vlde
Reverse Bar) waveform (stepped waveform that rises suddenly from black to white and falls smoothly from white to black) and the black flat waveform inserted in the 2nd, 4th, 5th, and 7th fields. It consists of Therefore, in order to detect a ghost, the following 8-field sequence calculation is performed in order to remove the influence of the horizontal synchronization signal of the previous line, the current line, and the color burst signal of the current line.

5OCR-1/4((Sl-S5) + (S6-
S2 ) + (Si −8y ) + (Sl!l
S4)l...(1) FIG. 4 shows the configuration of a ghost removal device adapted to 8-field sequence GCR.

This device is disclosed in the literature (Japanese Unexamined Patent Publication No. 59-211315). Further, FIG. 5 shows the operation sequence of this device. When the device is powered on or the channel is switched (step A1 in FIG. 5), the tap coefficients C-1 to C-1 held in the tap gain memory 44 are
Initial state settings such as setting C1 to 0 are performed (step A2).

In FIG. 4, a video signal is supplied to the input terminal IO, and an analog-to-digital converter 20 and a timing circuit 3
0 is introduced. The digitized video signal is supplied to a tapped delay line 41 of a transversal filter 40 . The output of each tap delayed by the time T in the tapped delay line 41 is multiplied by tap coefficients C-1 to C in a coefficient multiplier 42, and is supplied to an adder 43. The coefficients to be applied to the coefficient multiplier 42 are stored in a tap gain memory 44. The output of the adder 43 is led out to an output terminal 50 and is also supplied to an output waveform memory 51.

The timing circuit 30 determines the period required for this device (e.g. T - about 70 ns = 174 (f sc ) f sc
-3,579545MHz-color subcarrier) Nori O-/R CK is generated, and the first GC shown in FIG.
Upon detecting the arrival of R(Sl), the microprocessor 54
and the output waveform memory 51, the first OCR (S
1) to the 8th 0CR (Ss). Each GCR (S, -Ss) is taken into the output waveform memory 41 field by field as shown in the memory map of FIG. 6, and then stored at a predetermined address in the work RAM 52 (step A3). Microprocessor 54 then outputs output GCR (S
, ~Ss) and executes the 8-field sequence operation shown by equation (1) to obtain the final 0CR (SOCR) (step A4). The final 0CR (SOCR) is
It is composed of one word (1 word - 8 bits) and is expressed as a sample value as shown in the following formula.

5GCR= (SGCR, -1(k-0~1023)・
...(2) Next, the difference waveform (y,) defined by the following formula is calculated,
It is stored in the RAM 52 (step A5).

yh -8acR, *++ 5ccR, *
-(3) Next, the microprocessor 54 detects the maximum peak of the output difference waveform (yk) (step A6
). The peak position will be denoted as p.

That is, y is the peak of the impulse of the main signal.

Next, the microprocessor 54 generates an output difference waveform (y
, ), the reference signal waveform (r, l) stored in advance in the ROM 53 is subtracted from the peak position to obtain the error waveform (
e, ) and stores it in the work RAM 54 (step A7). This calculation formula is as shown in equation (4).

e * −yh r k
(4) Next, the microprocessor 54 corrects the tap gain based on the proportional method with leakage shown in equation (5) (step A8).

C+, sew -C+, ova (EX e
k - β New and old indicate after modification and before modification, respectively. Also, α indicates a positive minute correction amount,
β is a leakage amount that stabilizes the equalization operation, and is a positive minute amount smaller than α.

Ghosts are removed by repeatedly executing the equalization loop consisting of the above operation sequence (steps A3 to A8).

(Problems to be Solved by the Invention) In the ghost removal apparatus described above, when OCR calculation is performed in an 8-field sequence, it takes 8 fields of OCR acquisition time for one tap gain correction. As a result, there is a problem that the tap gain convergence time becomes long.

Therefore, an object of the present invention is to provide a ghost removal device that can shorten the convergence time.

[Structure of the Invention] (Means for Solving the Problems) This invention improves the efficiency of each field in which OCR is captured by
The difference from the OCR captured before the field is taken, and the minimum tap gain correction cycle is set to one field type.

(Operation) By the above means, it is possible to stably remove ghosts with a short convergence time.

(Example) Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing an operating procedure in an embodiment of the present invention. The overall block configuration of this invention is the same as the configuration shown in FIG. 4, but since its signal processing function is different, this part will be mainly explained.

When the power is turned on or the channel is changed,
Tap coefficients C-1~ held in tap gain memory
Initial state settings such as setting Co to 0 are performed (steps A1 and A2).

In the next step B1, the OCR of each field is captured. The captured OCR is written to a memory (RAM) having a memory map shown in FIG. M1~M5
is the memory area.

Before each OCR is captured, the contents of each memory area are sequentially transferred from M1 to M.2, from M2 to MS, from MS to M4, and from M4 to M5, and the latest captured waveform is stored in M1. By doing this, the latest captured waveform is always in Ml, and the waveform 4 fields before it is in Ml.
It is stored in 5.

Now, after turning on the power or changing the channel, the initial settings are completed, and the time for 5 fields has elapsed.
The GCR that has arrived at the field number will be stored in area M1. From this point on, the calculation of SGCH starts.

Let us now assume that the OCR captured in the i-th field is Yl. 0CR (Y5) captured in the 5th field is
It is stored in area M1. GCR (Yl) captured four fields before this is stored in area M5. Therefore, when calculating (Y5-Yl), as shown in Fig. 7, the horizontal synchronization signals of the previous line and current line,
WR that removes the influence of the color burst signal of the current line
B (step waveform) can be obtained.

Here, the i-th captured OCR (Yl) is WRB
Then, 0CR (Yl-4) four fields before is a black flat waveform. Conversely, the i-th imported OCR (
When YI) is a black flat waveform, the GCR (Yl-4) four fields before it is WRB. This is the seventh
This can be understood from the principle of the 8-field sequence shown in the figure.

In this embodiment, the result of subtracting the black flat waveform from WRB is always used in every field. For this purpose, in step B2, the latest OCR that has been imported is
It is determined whether it is a B or black flat waveform.

When the latest OCR (Yi) is WRB, area M
Ml-MS-OCR (SGCR) is obtained by subtracting the black flat waveform data of area M5 from the WRB data of 1.
Step B3), when the latest 0CR (Yf) is a black flat waveform, from the WRB data of area M5,
Subtract the black flat waveform data of 1 and get M5-Ml-O
Obtain CR (SOCR) (step B4).

The OCR (SOCR) obtained in this way is processed in step A5 as follows.

That is, the (k+1)th and kth S GCRs and the difference waveform yk are obtained by yk-8acR, h++5GCR. This is the same as equation (3) explained earlier. below,
Up to steps A6 and A7, the same processing as described in FIG. 5 is performed.

That is, the microprocessor outputs the differential waveform (y,)
The maximum peak of is detected (step A6). If the peak position is denoted by p, then y becomes the peak of the impulse of the main signal.

Next, the microprocessor generates the output difference waveform (yk)
, the reference signal waveform (rk) stored in the ROM in advance according to the peak position is subtracted to obtain the error waveform (e, l
Find this and store it in the working RAM (step A7)
. This calculation formula is as shown in formula (4).

The error waveform fe obtained in this way is used to correct the tap coefficients stored in the tap gain memory, but the amount of correction is is the latest OCR (Yl)
differs depending on whether it is a WRB or a black flat waveform. The reason for making the correction amounts different in this way will be described later.

Therefore, first in step B5, the latest GCR (Y
i'') is a WRB or a black flat waveform is determined.

And if OCR(Yi') is WRB,
In step B6, C1,sew -C1,o Id + (!
If OCR (Yi) is a black flat waveform, in step B7, CL aev -C1,old+α/8xet-βx
Perform the operation c,, old.

In other words, if OCR (Yi) is WRB, the correction amount will be the conventional 174, and if it is a black flat waveform,
It becomes 118. After the processing in step S6 or S7, the process returns to step B1 and the same processing as described above is repeated, and the tap gain for ghost removal is corrected. The reason why the correction amount was selected in this way will be explained below.

In the tap gain correction means shown in FIG.
) is synchronously added as shown in the formula. This means that the WRB waveform is added four times, so if WRB is 81 and the noise in the signal at that time is 01 to n4, then 1(S+nl)+(S+n2)+(S+n3)+(S+
n4)l/4-s+(1/4)(nl+n2+n
3+n4) - (6), and when the included noise is n, n is expressed by the following formula.

n = (1/4) (nl+n2+n3+n4) The variance (square mean) of this noise is (1, j-1, 2, 3, 4) (7) where - indicates the average value. Since nl is uncorrelated noise, n1nj=0, n12-σ2...(
8). Therefore, σfi'-σ2/4...(9
). In other words, the S/N ratio becomes 1/4, which means that the S/N ratio is improved by 6 dB by the synchronous addition. Noise variation σ of tap gain. is based on the literature (JEERTran
s, CP-28No, 3 P629 Aug, 1980
, ANovel Automatic Ghost
Canceller). According to this, atcm J-CrTD-#a. (a correction jl)
−(10). In the case of this embodiment, the standard deviation σ of the noise is smaller than that of the conventional one. is doubled (-2σfi), so if the amount of correction at this time is α', then (7α - (ff' /2) " 2 (7a - (1
1). Here, if we set equation (10) - equation (11),
α′ −α/4. Therefore, when the latest 0CR (Yi) is WRB, the correction amount is α/4 (Step B
6). Also, when the latest 0CR (Yl) is a black flat waveform, it means that the ghost detected by WRB 4 fields before is used, and the latest ghost is not used, so the amount of correction is For example, α
18 (step B7). If this is determined,
During 8 fields, the correction amount is α in the conventional case, but in this embodiment, it is (α/4)X4+(α/8)X 4-1.5α(12), which is 1.5 The amount of correction will be doubled.

As a result, the stability remains the same as before, and the convergence time can be shortened.

FIG. 3 shows another embodiment of the processing in steps B6 and 87, and the other parts are the same as those in FIG. 1, so they are omitted. In this embodiment, tap coefficient correction is performed using incremental control. The correction gain at this time is C1, new −C+, o +, + ahaX sgn on the step B6 side.
(e k)-β +a + α/8xsgn(
e k )−βX sgn(C+, ova ). At this time, the standard deviation of the variation due to noise in the tap gain is σa. is based on the above-mentioned document (IEEE Tran
s, CE-26No, 3 P629Aug, 1980
), σTG-0.79J7τ-... (13)(
Δ: correction amount). According to this embodiment, the standard deviation of the noise is doubled, so σTG-0,7fl/'τ7]co- (14). When formula (13)-(14) is used, Δ'-Δ/2 is obtained. Therefore, when the latest 0CR (Yi) is WRB, Δ
/2, and when 0CR (Yl) is a black flat waveform, it is half of that, Δ/4. Then, whereas in the conventional case, the amount of correction is Δ, in this embodiment, it becomes (Δ/2)x4+(Δ/4)X4-3Δ...(15), which is three times the amount of correction as before. . Therefore, the stability is the same as before, but the convergence time can be significantly shortened.

This invention is not limited to the above embodiments,
The leakage amount β may be changed. That is, although the correction amount α was changed in the above embodiment, the convergence time can be shortened even if the leakage amount β is changed. In this case as well, it is natural that the gain correction amount for the black flat waveform is set to be smaller than the gain correction amount for the WRB. That is,
If the OCR (Yi) captured in the i-th field is WRB, the leakage amount is doubled, for example, and 0CR (Yi) is obtained.
This can be achieved by increasing the amount of leakage when l) is a black flat waveform, for example, by four times.

In the memory control method of the embodiment described above, the contents of the memory are sequentially sent each time a GCR is captured, so that the latest captured waveform is always stored in M1, and the waveform 4 fields before it is stored in M5. Therefore, there is an advantage that the calculation of 5oCR can always be performed fixedly for the regions M1 and M5. However, as described above, it takes time to sequentially transfer the contents of the memory before importing the OCR.

However, if, for example, every time a GCR is imported, it is stored in the area where the oldest OCR, which is 4 fields before the current one, is stored, the transfer from memory area to memory area described above can be omitted, which saves time. ' The time can be shortened.

At this time, the area to be captured changes sequentially, so it is necessary to calculate the difference between the current OCR and the OCR 4 fields ago, but this can be achieved with simple flag processing or simple calculations such as count processing. can.

Furthermore, if it takes more than one field to correct one tap gain, it is sufficient to capture the waveform using an interrupt or the like every time an OCR waveform is received during the equalization loop shown in Figure 1. .

In addition, in this example, the number of memory areas for captured waveforms is 5.
However, it is not limited to this. The number of memories can be reduced to four by performing an operation with the OCR four fields before while taking in the OCR.

Furthermore, when multiple fields are required for one tap gain correction, the number of memory areas required can be reduced by performing processing such as not importing OCR of fields that are not used for correction.

It is clear that various types of memory control may be performed in the present invention, such as those described above.

Furthermore, it is clear that even if the correction amount is the same when the latest captured waveform is WRB and when it is the black flat waveform, the convergence speed will be faster, although accuracy will be sacrificed.

[Effects of the Invention] As explained above, according to the present invention, the convergence time can be shortened without changing the stability from the conventional method.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of an operation procedure showing one embodiment of the present invention, FIG. 2 is an explanatory diagram of an example of a memory area used in this invention, and FIG. 3 is a partial diagram of another embodiment of this invention. Explanatory diagram,
Fig. 4 is an explanatory diagram of the configuration of the ghost removal device, and Fig. 5 is an explanatory diagram of the configuration of the ghost removal device.
FIG. 6 is an explanatory diagram of the storage area of the memory of FIG. 5, and FIG. 7 is an explanatory diagram of the 8-field sequence. 20... Analog-to-digital converter, 30... Timing circuit, 40... Transversal filter, 41
...delay line, 42...coefficient unit, 43...adder,
44...Tap gain memory, 51...Output waveform memory, 52...RAM, 53...ROM, 54...
microprocessor.

Claims (1)

[Claims] By detecting a ghost component using a reference signal for ghost removal sent in an 8-field sequence and correcting the tap gain in the transversal filter, A ghost removal device for removing ghost components, comprising means for obtaining a difference signal between a reference signal for ghost removal captured for each field and a signal for ghost removal captured four fields before the standard signal, and using the difference signal. When obtaining the correction signal for the tap gain, the correction amount when the most recently acquired reference signal for ghost removal has a flat waveform is greater than the amount of correction when the most recently acquired reference signal for ghost removal has a step waveform. A ghost removal device comprising means for obtaining a correction signal of the tap gain by reducing a correction amount.