JPH03297277A - Signal processing unit - Google Patents

Abstract

PURPOSE:To prevent occurrence of smear by selecting and outputting a digital signal through a 1st digital low pass filter under a prescribed condition when a level difference is a prescribed level or below. CONSTITUTION:Three outputs of an LPE 2, an LPF3 and a delay circuit DL4 are inputted to a comparator circuit 5 and an electronic switch 7 selects an output of the LPF 2 only when the level difference between the LPF2 and DL4 and between the LPF3 and DL4 is both at a prescribed level or below. The discontinuity of a level delta is improved when the output of the LPF2 and DL4 is selected and switched.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a signal processing device that reduces quantization noise and the like generated when converting an analog signal into a digital signal.

(Prior Art) In recent years, digital signal processing has been performed in many video devices.

However, even in video equipment that processes these digital signals, human signals are usually analog signals, so
When digitizing a circuit, an AD converter is required, and quantization noise that occurs in proportion to the accuracy of the AD conversion may pose a problem.

By the way, if the signal to be handled is a video signal, the required A-
The number of quantization bits of the D converter may be about 8 to 10 bits, which is less than that for audio signals (usually 16 bits are required), but the conversion speed is on the contrary, the number of quantization bits for audio signals is several hundred. You'll need twice as much.

Therefore, as shown in FIG. 4, two types of reference voltages Vc c I have conventionally been used for such video signals.
V) A 2" resistor 33 (n is the number of quantization bits) is connected between the input terminals 31 and 32 of B, thereby dividing the divided voltage and the analog signal input from the input terminal 34 into 2+
A so-called flash type A that compares with one comparator 35, encodes the result with an encoder 36, and outputs it as an n-bit digital signal to an output terminal 37.
-D converters are commonly used.

By the way, with this flash type A-D converter,
As is clear from the above explanation, each time the required number of quantization bits n increases by 1, the number of required comparators doubles. need to be kept to a minimum.

On the other hand, as mentioned above, in ordinary video equipment, it is said that a quantization bit number of 8 can provide a reasonably satisfactory image quality, and for this reason, video A-D converters, which are currently popular as general-purpose products, The standard is an 8-bit type.

However, in broadcasting equipment that requires the highest quality video, especially television cameras that require gamma correction, 8 bits often lacks quantization accuracy.

In such cases, an A-D converter with 9 bits or more is usually required, but using these A-D converters, which are not very general-purpose, may cause problems in terms of cost, size, power, etc. Similar cases often occur in fields that handle things other than video signals.

For this reason, Japanese Patent Application Laid-Open No. 127883/1983 discloses a method of reducing quantization noise while achieving high bit precision for low frequency components by using an A-D converter with a small number of bits. A method is being considered.

To briefly explain this method using FIG.
PF) 42, and the other is a low pass filter (LPF) 43.
The small amplitude signal (noise component) of the output of the HPF 42 is clipped by a clip circuit 44, and added to the output of the LPF 43 whose bit precision has been improved by an adder circuit 45 to be output.

Here, to explain the basic operation of a digital low-pass filter, it takes a weighted average of n-bit digital signals sent at regular intervals over a certain period on the time axis. The quantization error (noise) is also averaged, and normally the quantization precision of this output is equivalently n+1
Becomes more than a bit.

For example, - The A-D conversion sampling frequency of digital video equipment that is popular for boat fishing is often around 14 to 18 MHz, but the band of this A-D conversion output is filtered using a digital filter, which corresponds to the sampling frequency of 178. When the frequency is limited to about 2 MHz, the equivalent quantization precision of the output of this filter is improved to approximately n+1 bits, and when limited to about 0.5 MHz, which is 1/32, it is improved to n+2 bits.

In other words, for low frequency signal components, by filtering the A-D conversion result with a digital filter, an n-bit A-D converter is used to convert the A-D conversion result to n+1 bits or more.
It becomes possible to obtain a digital signal with quantization precision equivalent to that obtained using a converter.

Figure 6 shows an analog signal waveform A whose value changes with a gentle slope (does not include high frequency components), a waveform B which is an analog image of a digital signal obtained by converting this signal from A to D, and a waveform B which is an analog image of a digital signal obtained by converting this signal into a digital signal. It shows a waveform C passed through a pass-pass filter.

An example of this LPF is shown in FIG. 7, and its operation will be briefly explained below.

The digital input signal is sequentially delayed by unit time delay elements 60-63. The outputs and digital input signals of each delay element 60-63 are sent to multipliers 64-68, each of which is multiplied by 4 to a coefficient KO~, which is added to an adder 69 to produce an output of 0. 4 is set to a value that becomes 1 when added together, and as a result, the output of each multiplier 64 to 68 has a value below the decimal point.

Therefore, if this is added without being rounded down, the output of this LPF will have a value below the decimal point.

That is, if the input signal is n bits, this output signal is n bits.
This is equivalent to becoming a digital signal of +1 bit or more. Furthermore, as mentioned above, the output of the digital filter is the result of weighted averaging of the input signal on the time axis, and the quantization error (noise) is also averaged, so the data below the decimal point also has meaning. It's full.

However, this method has the following drawbacks.

This drawback will be explained below using an example.

Let us now consider a case where a step-like signal as shown in FIG. 9(A) is input manually. At this time, the outputs of the LPF 43 and HPF 42 become as shown in FIGS. 9(B) and 9(C), and the output of the clip circuit 144 becomes as shown in FIG. 9(D). The output obtained by adding the outputs of the LPF 43 and the clipping circuit 44 has a waveform as shown in (E), which is considerably distorted compared to the original input waveform.

That is, in the conventional method, a part of the high frequency component of the signal is removed along with the noise, resulting in deterioration of resolution, and furthermore, it is not possible to prevent smear (image blurring)-like distortion from occurring before and after the contour of the image.

(Problem to be Solved by the Invention) As described above, quantization noise becomes a problem in video signals, etc. when capturing an image of a subject whose light intensity is changing with a gentle slope. In many cases, this is true for signals containing many low-frequency components, but in conventional technology, in order to reduce quantization noise for low-frequency signal components, quantization of the A-D converter is necessary. Increase the number of bits or
Since it was necessary to increase the quantization accuracy with the configuration shown in Figure 5, it was necessary to increase the cost, reduce the resolution of high-frequency signal components, and reduce the smear-like distortion that occurs before and after the contour. The problem could not be resolved.

The present invention solves these drawbacks, achieves high quantization accuracy especially for low-frequency signal components below a predetermined frequency, and eliminates the occurrence of smear-like distortion before and after the contour at low cost and low power consumption. The purpose is to realize a signal processing device that does not

(Means for Solving the Problem) In order to achieve the above object, the present invention provides an n-bit A-D
The digital signal of the conversion output is smoothed to reduce the bit length to n.
one or more digital low-pass filters for bit expansion that extend the signal to +1 or more for output; one or more digital low-pass filters for determining the content of high-frequency components of the input signal; means for comparing the output of the A-D converter with the output of the digital low-pass filter for determination, and depending on the comparing means, either the output of the A-D conversion or the output of the digital low-pass filter for bit expansion; The configuration includes means for selecting and outputting.

Alternatively, a digital low-pass filter with a predetermined cutoff frequency that limits the frequency band of the A-D converted digital signal and removes high-frequency noise components, and a digital low-pass filter that limits the frequency band of the A-D converted digital signal and filters the digital signal for the same time as the delay time of the digital low-pass filter. a first delay means for delaying; a means for comparing the output of the digital low-pass filter and the first delay means; and a means for comparing the output of the digital low-pass filter with the output of the first delay means; means for determining a certain period and outputting first information indicating it; and time adjustment for receiving this first information and outputting second information indicating a period longer than the visit indicated by this information by a predetermined time m T 1. means, a second delay means for further delaying the force of the digital low-pass filter and the first delay means by T2 (TI>72), and the second delay means based on the second information. The configuration includes means for selecting and outputting either the digital low-pass filter delayed by T2 or the Ryokawa power of the first delay means.

(Function) As a result, the present invention provides a signal that does not pass through the digital low-pass filter when the digital signal contains high-frequency components higher than a predetermined value, and quantizes the signal that passes through the digital low-pass filter at other times. By selecting and outputting a signal with improved resolution, it is possible to prevent deterioration of resolution, and furthermore, it is possible to prevent discontinuity that occurs at the selection/switching point of both signals.

(Example) FIG. 1 shows a first example of the present invention, and the present invention will be described in detail below.

The analog input signal is converted into a digital signal by an A-D converter 1, and the output is converted to a digital low-pass filter (LPF) 2 with a predetermined cutoff frequency. Judgment L
P F 3. and is sent to a delay circuit (D L ) with a delay amount equal to the delay time of L P F 2°3.

Note that when the input signal is an image signal, the image signal is a signal that has correlation in the horizontal, vertical, and time axis directions, so the above delay circuit uses a 1-horizontal line delay element or 1-frame delay element. 2D, 3D by combining
It is also possible to construct a dimensional LPF.

Now, among the LPF2.3 configured as above and the DL4 having a delay amount equal to the delay time of the filter, the LPF
The outputs of DL2 and DL4 are respectively applied to switching terminals of an electronic switch 7. Further, the outputs of LPF3 and DL4 are sent to a comparison circuit 5, and the results are sent to a determination circuit 6. To explain the operation of the judgment circuit 6 and the electronic switch 7, the judgment circuit 6 calculates the difference between the output of the DL4 and the output of the LPF3, and
It is determined whether this difference is below or above a predetermined level. Here, the difference between the output of LPF3 and the output of DL4 is nothing but the amount of high frequency components included in the human input signal.

The present invention thus determines the amount of high frequency components contained in the input signal, and when this amount is greater than a predetermined value, the output of DL4, which does not apply band limitation to the human input signal, is changed to When it is less than the value, the electronic switch 7 selects and outputs the output of the LPF 2 whose bit length has been expanded.

By the way, the presence or absence of a high frequency component contained in an input signal can be determined from the difference between LPF2 and DL4 without using the determination LPF3, but the presence or absence of a high frequency component contained in the input signal can be determined from the difference between the two and the electronic switch 7 is activated. If you switch, the following problems will occur.

This will be explained below using an example.

Now, consider a case where a step-like signal is manually generated. At this time, the DL4 output and the output of the LPF2 are represented by a waveform 8-1 shown by a solid line and a waveform 8-2 shown by a dotted line in FIG. 8(A). (It is originally a digital signal, but for the sake of explanation, it is depicted as an analog signal.)

Here, when the level difference between these two signals is more than the predetermined value 8, the solid line DL4 output is used, and when it is less than δ, the dotted line LPF
If 2 outputs are selected, the second selected output is shown in Figure 8 (B
) is the waveform 8-4 shown in FIG. That is, at the point where the above two signals are selected and switched, a level δ discontinuity occurs in the output signal. These discontinuous points cause ghost-like false contours when the signal to be handled is an image signal, and also causes various problems when using other signals.

In order to solve this problem, in the present invention, an LPF3 having a lower cutoff frequency or a higher order than LPF2 is used for selection switching between LPF2 and DL4 output, and this LPF3
A method is used in which the presence or absence of a high frequency component is determined based on the level difference between the output of LPF2 and the pressure of DL4, and selection switching between the LPF2 output and the DL4 output is performed.

That is, when the level difference between the outputs of LPF3 and DL4 is greater than the predetermined value 8, the output of DL4 is used, and when this difference is less than δ, the output of DL4 is
The output of PF2 is selected by electronic switch 7.

In this way, if we consider the above-mentioned step-like signal input as an example, the output of LPF3 will be the waveform 8-3 shown by the dotted chain line in Fig. 8(A), so the level of this output and the output of DL4 will be The result of selecting and switching the outputs of LPF2 and DL4 based on the difference is waveform 8-5 shown in Figure 8(C).
become that way. That is, the discontinuity of level δ, which has been a problem in the past, is significantly improved.

Note that depending on the input signal, there may be a case where the level difference between LPF3 and DL4 is larger than the level difference between LPF2 and DL4. In this case, as shown by the dotted line in FIG. DL4 no 3
2 outputs manually, LPF2 and DL4 and LPF3 and D
Only when the level differences of L4 are both below a predetermined value,
A method may be adopted in which the output of the LPF 2 is selected using the electronic switch 7.

Next, FIG. 2 shows a second embodiment of the present invention.

In this example, the pressure of the A-D converter 1 is changed to the first one with a different cutoff frequency or order. Second. Third digital low pass filter (LPF) 8, 9.10 and delay circuit (DL)
4, and compare four outputs including this LPFB output or three outputs not including this in a comparison diagram 1111,
This result is sent to the determination circuit 12 and the LPFB and LPF9
．． This selects and applies pressure to one of the outputs of DL4.

Note that the cutoff frequency of each LPF is LPFIO.

LPF9. The order increases in the order of LPF8, or the order of the filter increases in the order of LPF8. LPF9. It is assumed that the values increase in the order of LPFIO.

As mentioned above, the digital low-pass filter takes a weighted average of the human input signal on the time axis, so generally, the lower the cutoff frequency or the higher the order, the more the output quantization error will be averaged and reduced.

In this embodiment, the comparison circuit 11 uses DL4 and the second and third LPFs 9.10, or in addition to these, DL4 and the first LPF.
FB and sends it to the determination circuit 12, and when the level difference between the output of DL4 and the output of LPF9, or the level difference between the output of DL4 and the outputs of LPF8 and LPF9, are both less than a predetermined value δ. is the output of LPF8, or the level difference between LPF10 and DL4 at other times, or? 1DL4 output and LPF 10 and LPF
When the level difference with the output of DL 9 is less than a predetermined value δ, the output of LPF 9 is selected by the electronic switch 13, and in other cases, the output of DL 4 is selected and output. In this way, it is possible to realize a device that outputs a more precise A-D conversion result in accordance with the amount of high frequency components contained in the input signal in more detail than in the first embodiment.

In this explanation, the condition for selecting the output of LPF 9 is when the level difference between the output of DL 4 and the outputs of LPF 10 and LPF 9 are both below a predetermined value δ. In some cases, it may be better to set the condition that the level difference between DL4 and DL4 both satisfy a condition that the predetermined value δ or less is satisfied.

In addition, as shown in Fig. 3, L
A determination filter LPF 14 having a lower cutoff frequency or higher order than the LPF 9 may be used instead of the PF 9.

It is also possible to increase the number of digital low-pass filters in the same way, and even further increase the number of digital low-pass filters. Second
The effects of the present invention can also be obtained by using a digital low-pass filter whose cutoff frequency is set to the upper limit of the frequency band of the required signal instead of DL4 in the embodiment.

Another embodiment of the present invention is shown in FIG. 10 and will be described in detail below.

The digital human input signal is passed through a digital low-pass filter (L
P F )2 and a delay circuit (D L )4 having the same delay amount as the LPF2. L
The outputs of PF2 and DL4 are sent to a comparison circuit δ, and the results are sent to a determination circuit 6. The determination circuit 6 takes the level difference between the output of the DL4 and the output of the LPF2, and determines whether this difference is below or above a predetermined level. Here, the level difference between the output of LPF 2 and the output of DL 4 is nothing but the amount of high frequency components contained in the input signal. When the amount of high-frequency components is above the specified value, the DL4 output is output without band-limiting the human input signal, and conversely, when the amount of high-frequency components is below the predetermined value, the LP is output with band-limiting and noise components suppressed.
This selects and outputs the F2 output.

By the way, if the presence of high frequency components is determined from the difference between the output of DL4 and the output of LPF2, and the electronic switch 7 is directly switched based on this result to selectively output the output of DL4 and the output of LPF2, the above-mentioned problem will occur. . below,
This will be explained using an example in which a step-like input signal is added as in the previous example.

Now, the DL4 output and the LPF2 output are shown as a solid line and a dotted line, respectively, in FIG. If the LPF2 output is selected, this selected output has a waveform as shown in FIG. 11(B). That is, at the point where the above two signals are selected and switched, a level δ discontinuity occurs in the output signal.

Therefore, in this embodiment, the comparison circuit δ compares the level difference between the LPF2 and the DL4, and sends this to the determination circuit 6. Output the information 28 and send it to the time adjustment port 1ia2
9, and obtains second information 20 indicating a period longer than the period indicated by the first information 28 by a predetermined time TI, and this second information 20 is sent to
According to the information 2o, the LPF2 output and the DL4 output are respectively set to the second output for a predetermined time T2. A method is used in which one of the outputs delayed by the third delay means 24 and 25 is selectively output using the electronic switch 7 to prevent the occurrence of the discontinuity described above.

Using FIG. 11, this operation will be explained in the case where a step-like signal is input as the input signal as in the previous example.The period indicated by the first information 28 output from the determination circuit 6 is indicated by a pulse. 11(C), and on the other hand, the second information 2 output from the time adjustment circuit 29
The period indicated by o is as shown in FIG. 11(D). On the other hand, the dotted line indicates the filter 2 output that has passed through the second delay means 24.25.
The output of L4 is shown as a solid line as shown in FIG. 11(E). (Note that the case where T2=TI/2 is shown here).

Therefore, the output of the electronic switch 7 controlled by the second information 20 becomes as shown in FIG. 11(F), that is, the conventional problem of discontinuity in level δ is greatly improved.

Note that when the input signal is an image signal, since the image signal is a signal that has correlation in the horizontal, vertical, and time axis directions, it is possible to use a horizontal line delay element or a frame delay element as the delay circuit. Putting things together in the morning creates 2D and 3D
It is also possible to construct a dimensional LPF.

(Effects of the Invention) As described above, by using the present invention, it is possible to reduce noise including quantization noise that occurs during A-D conversion, and also to reduce smear-like distortion that occurs before and after the contour of an image. can be prevented, and the effect is great.

4.l! Brief explanation of the first aspect FIGS. 1, 2, 3, and 10 illustrate the first aspect of the present invention. Second. Third. A block diagram showing the fourth embodiment, FIG. 4 is a diagram explaining the configuration of a general A-D converter, FIG. 5 is a block diagram showing a conventional example, and FIG. 6 is an analog signal that does not have high frequency components. Figure 7 is a block diagram showing an example of a digital low-pass filter. Figure 8 is a diagram illustrating how the bit precision improves in a signal obtained by A-D converting a signal. Figure 8 is a diagram illustrating the operation of the first embodiment of the present invention. FIG. 9 is a waveform diagram illustrating the prior art, and FIG. 11 is a waveform diagram illustrating the operation of the fourth embodiment of the present invention.

1: A-D converter, 2, 3, 8, 9, 10, 14: Digital low-pass filter (LPF), 4, 24° 25:
Delay circuit (DL), δ, 11, 15: Comparison circuit, 6
, 12, 16: Judgment circuit, 7, 13: Electronic switch, 2
9: Time adjustment circuit.

No. 7 figure No. 9 figure

Claims (1)

[Claims] 1. A first digital low-pass filter with a predetermined cutoff frequency that limits and smoothes the frequency band of a digital signal; A second high-order digital low-pass filter detects a level difference between the digital signal that does not pass through any of the digital low-pass filters and the digital signal that passes through the second digital low-pass filter. means, selects the digital signal that does not pass through any of the digital low-pass filters when the level difference is above a predetermined value, and selects the first digital signal according to the predetermined condition only when the level difference is below the predetermined value.
A signal processing device comprising means for selecting and outputting the digital signal passed through the digital low-pass filter. 2. A plurality of digital low-pass filters arranged in parallel each having a different cutoff frequency to limit and smooth the frequency band of the digital signal, and the digital signal and the plurality of digital signals that do not pass through these digital low-pass filters. means for detecting a level difference between each of the digital signals passed through the low-pass filter, and detecting any one of the digital signals passed through the plurality of digital low-pass filters. , the level difference between the digital signal passed through a digital low-pass filter having a lower cutoff frequency or higher order than the digital low-pass filter and the digital signal not passed through any of the digital low-pass filters is a predetermined level difference. Means for selecting and outputting the digital signal on the condition that it is less than or equal to the above value, or selecting and outputting the digital signal that does not pass through any of the digital low-pass filters when the respective level differences do not meet the above conditions. A signal processing device comprising: 3. The signal processing device according to claim 1 or 2, wherein the digital signal is an image signal, and a two-dimensional or three-dimensional filter is used as the digital low-pass filter. 4. a digital low-pass filter with a predetermined cutoff frequency that limits and smoothes the frequency band of the digital signal;
A first delay means for delaying the digital signal that does not pass through the digital low-pass filter, and a level difference between the outputs of the digital low-pass filter and the first delay means is detected, and this difference is determined to be a predetermined value. means for outputting first information indicating a period in which the value is greater than or equal to a value; means for receiving the first information and outputting second information indicating a period longer than the period indicated by this information by a predetermined time T1; a second delay means for further delaying both the outputs of the low-pass filter and the first delay means by a predetermined time T2 (T1>T2);
A signal processing device comprising means for selecting and outputting either the output of the digital low-pass filter delayed by 2 and the output of the first delay means. 5. The signal processing device according to claim 4, wherein the digital signal is an image signal, and a two-dimensional or three-dimensional filter is used as the digital low-pass filter.