JPH04367170A - Ghost remover - Google Patents

Abstract

PURPOSE:To remove ghost by removing a reference signal component from an input waveform, by adding the reference signal component after non-linear processing is carried out so as to effectively reduce a noise component, and by using a training waveform that is regarded as a ghost distortion. CONSTITUTION:Video signal that is interrupted by ghost is input in an A/D 11 and timing signal generator 19, the video signal input in the A/D 11 is input in the A terminal of switch 12 and in an input waveform memory 21. Further, in the B terminal of switch 12 the output signal from a training waveform memory 20 is input, the output from switch 12 is input in both a delay device 13 and a TF 14. Furthermore, both the output from delay device 13 and the output from the TF 14 are input in a subtractor 15, whereby the output signal from TF 14 is subtracted from the output signal of delay device 13, and the result is input a subtractor 15, whereby the output signal from TF 14 is subtracted from the output of delay device 13, and the result is input in a subtractor 16. Thus, a noise component is effectively reduced, thereby being able to remove the ghost.

Description

[Detailed description of the invention]

[Object of the invention]

0002

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a ghost removal device, and more particularly to a ghost removal device that reduces the time required for ghost removal.

0003

[Prior Art] In recent years, in television broadcasting, GCR (ghost can) is used as a reference signal for ghost removal.
A television receiver has been developed in which a cell reference (GCR) signal is inserted into the vertical retrace period, and this GCR signal is used to perform ghost removal.

FIG. 5 shows the configuration of a conventional ghost removal device. The reference for this device is “NEW GH
OST REDUCTION TUNER USING
A TRAINING METHOD” (IEEE
Transactions on Consumer
Electronics, Vol 36, No. 3,A
UGUST 1990) is fried.

In FIG. 5, a video signal subjected to ghost interference is inputted to an A/D (analog-to-digital converter) 11 and a timing signal generation circuit 19 via a terminal 10. The video signal converted into a digital signal by the A/D 12 is input to the A terminal of the switch 12 and the input waveform memory 21. Further, the output signal of the training waveform memory 20 is input to the B terminal of the switch 12. The output of this switch 12 is input to a delay device 13 and a TF (transversal filter) 14. The output signal of the delay device 13 and the output signal of the TF 14 are input to a subtracter 15. Here, the output signal of the TF 14 is subtracted from the output signal of the delay device 13 and outputted to the subtracter 16. The output signal of TF17 is also input to the subtracter 16, and the output signal of TF17 is subtracted from the output signal of the subtracter 15.
It is output to F17, the output terminal 18, and the output waveform memory 22.

Input waveform memory 21 and output waveform memory 22
and store the GCR signals inserted into the digital video signal before and after ghost removal, respectively. The MPU (microprocessor) 25 uses these GCRs.
TF14 using the signal, RAM23, and ROM24
, 17 tap coefficients are calculated. Note that the ROM 24 stores the same reference signal as the GCR signal inserted into the input video signal. Further, the timing signal generation circuit 19 generates and outputs clock signals for these systems.

FIG. 6 is a flow chart for explaining the ghost removal operation in the circuit shown in FIG. First, when power is turned on or channel switching is performed in S1, the MPU 25 sets an initial state in S2,
Initialize the tap coefficients of TF14 and TF17.

Then, in S3, the B terminal of the switch 12 is selected, and in S4, the GCR waveform of the input signal is loaded into the input waveform memory 21, and the MPU 25 performs synchronous addition. The reference waveform whose noise has been reduced by this synchronous addition is pulsed and used as a training waveform. A training waveform is set in the training waveform memory 20 in S8.

Next, in S9, the variable i is set to "0", and in S1
0, the training waveform is output from the training waveform memory 20. This training waveform is used for two TF14,
The ghost component is reduced by an equalization circuit including 17, and the output waveform is taken into the output waveform memory 22 (S11). The reference waveform stored in the ROM 24 is subtracted from this output waveform to obtain an error waveform (S12). The tap coefficients of the TFs 14 and 17 are determined from the determined error waveform (S13), and the respective tap coefficients are changed (S14). Thereafter, "1" is added to the variable i (S15) and compared with a predetermined value X. If i<X, the process returns to S10 and repeats changing the tap coefficient.

On the other hand, if i=X, the process advances to S17, and the switch 12 selects the A terminal, thereby ending the training operation (S18). As a result, the video signal containing the ghost component inputted from the terminal 10 is outputted from the terminal 18 with the ghost component removed.

This method is called a training method in the above-mentioned literature, and uses a waveform whose noise has been reduced by performing synchronous addition as a training signal. Thereby, the tap coefficient correction cycle can be made equal to or less than the insertion cycle of the reference waveform in the input signal. Furthermore, since a waveform with a good SN ratio can be used, the tap coefficient correction period can be further shortened.

FIG. 7 shows a GCR signal currently inserted into television broadcasts. This signal was approved as a standard by the Telecommunication Technology Council in March 1989, and is composed of a sinX/X bar waveform and a pedestal waveform. These waveforms are inserted into the 18th H (horizontal period) or 281H of the vertical blanking period. The sinX/X bar waveform and pedestal waveform are inserted in an 8-field sequence, and the 1st, 3rd, 6th, 8th
The field contains sinX/X bar waveforms (S1, S3, S
6, S8) are inserted, and pedestal waveforms (S2, S4, S5, S7) are inserted in the second, fourth, fifth, and seventh fields. The ghost removal device extracts the GCR waveform SGCR shown in FIG. 7(i) by calculating the following equation.

[0013]

[Equation 1] SGCR = (1/4) x {(S1-S5) + (S6-
S2)
+(S3-
S7)+(S8-S4)} (1)

0
Ghost removal and waveform equalization are performed based on this waveform. Further, by performing one sample difference on this waveform, a pulse waveform can be obtained. If this waveform is used every time to remove ghosts, one tap correction will require eight fields plus calculation time. Therefore, it takes time for the tap coefficients to converge. However, when the training method is used, the 8-field sequence operation in the equalization loop is not necessary, so the convergence time of the tap coefficients can be significantly improved.

However, if the SN ratio is poor or if the signal contains a beat component, the time required for synchronous addition to sufficiently remove noise will be from several seconds to several tens of seconds. Therefore, the time required from the time the power is turned on until the ghost removal is completed becomes long, and the shortening of the tap coefficient convergence time, which is a feature of the training method, cannot be achieved.

On the other hand, if the number of synchronous additions is insufficient, noise is not completely removed, so errors in tap coefficients due to malfunctions and grandchild ghosts may be buried in the noise and cannot be removed. In this case, in addition to the remaining ghost,
Problems such as the addition of ghosts and color blurring and blurring occurred due to malfunctions. In some cases, oscillation of the equalization circuit,
A divergence phenomenon was occurring. This kind of problem was common to other division methods as well.

[0017]

[Problem to be solved by the invention] As mentioned above, SN
If the ratio is poor or there is a beat component in the signal, the time required for synchronous addition to sufficiently remove noise will increase, resulting in a longer time required from power-on to complete ghost removal. was. Therefore, it was not possible to reduce the tap coefficient convergence time. Furthermore, if the number of synchronous additions was insufficient, noise would not be completely removed, resulting in malfunctions. Due to this malfunction, tap coefficient errors and grandchild ghosts were sometimes buried in noise and could not be removed. In such cases, in addition to remaining ghosts, problems such as ghosts being added due to malfunctions, color bleeding, and blurring occurred. In some cases, oscillation and divergence phenomena occurred in the equalization circuit. Such problems were also common in the division method.

The present invention has been made in view of these problems, and even when the S/N ratio of the signal is poor or a beat component is present, the ghost removal time from power-on can be shortened and the ghost removal can be effectively performed. An object of the present invention is to provide a ghost removal device capable of removing ghosts.

[Configuration of the invention]

[0020]

[Means for solving the problem] The means according to the present invention are as follows:
an equalizing means including a transversal filter with variable tap coefficients; a first extracting means for extracting a waveform of a reference signal portion for ghost removal inserted into the input signal of the equalizing means; and an output of the equalizing means. a second extraction means for extracting the waveform of the reference signal portion inserted into the signal; and synchronously adding the waveform of the reference signal portion extracted by the first extraction means;
error waveform detection means for detecting an error waveform obtained by subtracting the reference signal; nonlinear processing means for performing nonlinear processing on the error waveform to remove noise components; and a tap coefficient correcting means that performs an operation to correct the tap coefficient using the output signal of the adding means and the output signal of the second extracting means, thereby eliminating ghost components. Noise components are effectively removed.

[0021]

[Operation] Only the ghost component is extracted by subtracting the reference signal component from the waveform of the reference signal portion of the input signal and performing nonlinear processing to remove only the noise component. The ghost component is reliably removed by adding a reference signal to this ghost component and performing an operation to modify the tap coefficients in the equalization means.

[0022]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The structure and operation of an embodiment of the present invention will be explained below with reference to the drawings. Here, FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a flowchart explaining the operation in FIG. 1, FIG. 3 is an explanatory diagram explaining waveforms in FIG. 1, and FIG. FIG. 3 is an explanatory diagram illustrating output amplitude characteristics.

In FIG. 1, a video signal subjected to ghost interference is inputted to an A/D 11 and a timing signal generation circuit 19 via a terminal 10. The video signal converted into a digital signal by the A/D 12 is input to the A terminal of the switch 12 and the input waveform memory 21. Further, the output signal of the training waveform memory 20 is input to the B terminal of the switch 12. The output of this switch 12 is input to a delay device 13 and a TF 14. The output signal of the delay device 13 and the output signal of the TF 14 are input to a subtracter 15. Here, the output signal of the TF 14 is subtracted from the output signal of the delay device 13 and outputted to the subtracter 16. The output signal of the TF 17 is also input to the subtracter 16, and the output signal of the TF 17 is subtracted from the output signal of the subtracter 15, and the result is output to the TF 17, the output terminal 18, and the output waveform memory 22.

Input waveform memory 21 and output waveform memory 22
and store the GCR signals inserted into the digital video signal before and after ghost removal, respectively. The MPU 25 receives these GCR signals and the RAM 23, R
The tap coefficients of TF14 and TF17 are obtained using OM24. Note that the ROM 24 stores the same reference signal as the GCR signal inserted into the input video signal. Further, the timing signal generation circuit 19 generates and outputs clock signals for these systems.

FIG. 2 is a flowchart for explaining the ghost removal operation in the circuit shown in FIG. First, when power is turned on or channel switching is performed (S1
), the MPU 25 sets an initial state and initializes the tap coefficients of the TFs 14 and 17 (S2). and switch 12
(S3), the GCR waveform of the input signal is taken into the input waveform memory 21, and the MPU 25 performs synchronous addition (S4). The reference signal component is subtracted from the waveform of the reference signal portion whose noise has been reduced by this synchronous addition (S5).
. As a result, a waveform to which a ghost component and a noise component are added is extracted.

[0026] Here, most of the noise components have small amplitudes and are distributed over a wide range. Therefore, nonlinear processing is performed on the waveform obtained in S5 so that the small amplitude portion is not output, and the high amplitude waveform is output as is (S6). This removes most of the noise components. A reference signal component is added to this waveform (S7), and this is set as a training waveform in the training waveform memory 20 (S8).

Next, the variable i is set to "0" (S9), and the training waveform is output from the training waveform memory 20 (S10). This training waveform is used for two TF1
The ghost component is reduced by an equalization circuit including 4 and 17, and the output waveform is taken into the output waveform memory 22 (S11). The reference waveform stored in the ROM 24 is subtracted from this output waveform to obtain an error waveform (S12). Find the tap coefficients of TF14 and TF17 from the obtained error waveform (S13)
, change the respective tap coefficients (S14). Thereafter, "1" is added to the variable i (S15) and compared with a predetermined value X. If i<X, the process returns to S10 and repeats changing the tap coefficient.

On the other hand, if i=X, the process advances to S17, and the switch 12 selects the A terminal, thereby ending the training operation (S18). As a result, the video signal containing the ghost component inputted from the terminal 10 is outputted from the terminal 18 with the ghost component removed.

Next, a waveform processing method up to setting a training waveform will be explained with reference to FIGS. 3 and 4. In FIG. 3, by subtracting the reference waveform component shown in (b) from the synchronously added waveform (a) with reduced noise, a waveform with added noise and ghosts as shown in (c) is obtained. By applying nonlinear processing to this waveform as shown in Figure 4,
A waveform (d) with noise removed is obtained. In this waveform (b
By adding the reference waveforms of ), it is possible to obtain a waveform from which noise components have been removed, as shown in (e). By setting this waveform as a training waveform, ghost removal can be performed reliably.

In the nonlinear processing described above, as shown in FIG. 4, when the amplitude of the signal before processing is less than ±α, the nonlinear processing is performed such that the amplitude after processing is smaller than the amplitude of the signal before processing. In the case of α or more, linear processing is performed in which the amplitude after processing changes in proportion to the amplitude before processing. Here, the value of α can be changed according to the noise component.

As described above, the reference waveform component is subtracted from the waveform of the reference signal portion whose noise component has been reduced by synchronous addition, and the noise component is removed by nonlinear processing. By adding the reference waveform component to this waveform, a training waveform free of noise components is generated. By performing ghost removal calculations using this training waveform free of noise components, accurate ghost removal that is not affected by noise components can be performed and the ghost removal time can be shortened. This eliminates the addition of ghosts, color blurring, and blurring caused by mistaking noise for ghosts. In addition, grandchild ghosts and great-grandchild ghosts are effectively removed, and divergence and oscillation of the equalization circuit become less likely to occur. Although the present invention cannot remove small-amplitude ghost components, it can be improved on the screen if we take into account that they are not noticeable to begin with and that the interference caused by noise is greater.

In addition to the method shown in FIG. 4, the nonlinear processing may be performed by, for example, a method in which amplitudes below α are set to "0" after processing. Further, considering that large noise components have high frequencies, noise may be removed using a low-pass filter. Furthermore, if the device has a function of detecting the SN ratio, it is also possible to perform noise removal and ghost removal according to the SN ratio. In addition to this, an equalization circuit for calculating tap coefficients may be provided separately. Further, the calculation equivalent to the equalization circuit may be implemented using a product-sum calculation unit or by software. The same applies to other waveform calculation processes.

[0033]

[Effects of the Invention] As described above, since noise cannot be mistaken for a ghost, troubles due to malfunctions do not occur. Furthermore, since it is not affected by noise, effective ghost removal is possible. Furthermore, since the number of times of synchronous addition can be reduced, the time required to remove ghosts can be shortened.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention.

[
Figure 2: Flowchart explaining the operation of Figure 1

[Fig. 3] Explanatory diagram explaining the waveform in Fig. 1

[Figure 4] Explanatory diagram explaining input/output amplitude characteristics in nonlinear processing

[Figure 5] Block diagram showing the conventional configuration

[Fig. 6] Flowchart explaining the operation of Fig. 5

[Figure 7]
Explanatory diagram explaining the waveform of the GCR signal

[Explanation of symbols]

11...A/D 12...Switch 13...Delay device 14,17...TF 15, 16...subtractor 19...Timing signal generation circuit 20...Training waveform memory 21...Input waveform memory 22...Output waveform memory 23...RAM 24...ROM 25...MPU

Claims (2)

[Claims] 1. Equalization means including a transversal filter with variable tap coefficients; first extraction means for extracting a waveform of a reference signal portion for ghost removal inserted into an input signal of the equalization means; A second extraction means for extracting the waveform of the reference signal portion inserted into the output signal of the equalization means and a waveform of the reference signal portion extracted by the first extraction means are synchronously added to generate the reference signal. error waveform detection means for detecting the subtracted error waveform; nonlinear processing means for performing nonlinear processing on the error waveform to remove noise components; and addition for adding the reference signal to the output signal of the nonlinear processing means. means, the output signal of the adding means and the second
1. A ghost removal device comprising: tap coefficient correction means for performing a calculation for correcting the tap coefficients based on the output signal of the extraction means. 2. The ghost removal apparatus according to claim 1, wherein the nonlinear processing means changes the shape of the nonlinear processing depending on a noise component.