JPS5841545B2 - Shading removal circuit - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a shading removal circuit used in optical character reading devices and the like.

An optical character reader (hereinafter abbreviated as OCR) that automatically reads handwritten or printed characters.

) is constructed as shown in Figure 1.

As shown in Fig. 1, when a form 2 with printed or handwritten characters is sent by the paper feed mechanism 1, light from a light source 3 is applied to the form 2, and the text is displayed according to the character pattern on the form 2. Reflected light 4 passes through lens system 5 and enters scanner photoelectric conversion section 6 .

As shown in FIG. 2, this scanner photoelectric conversion unit 6 reads the character pattern 9 on the form by vertical buttons and horizontal scanning indicated by arrows A and B, and the vertical scanning is performed using a photodiode. This is realized by a line sensor 10 in which photoelectric conversion elements and scanning circuits, etc., are arranged one-dimensionally in the vertical direction.

There are two types of OCR in OCR: a page reader that reads characters across multiple lines within a document, and a document reader that reads characters on one or two lines within a document.When performing horizontal scanning, the page reader uses the line sensor 10. Document readers are designed to move the form horizontally, while mechanically moving the form in the direction of the character string.

Through such scanning, the analog video signal read by the photoelectric converter 6 is converted by the quantization circuit I into a digital signal of two gray levels, black and white.

The recognition circuit 8 takes in this binarized digital data and performs various recognition processes to determine which character the original pattern corresponds to.

The biggest problem with such OCR when converting an analog video signal into a digital signal using the quantization circuit I is shading.

This shading is caused by uneven illuminance in the lighting system, and the influence of this shading makes it extremely difficult to read characters, so a shading removal circuit is used to remove the shading.

FIG. 3 shows the configuration of a quantization circuit including a conventional shading removal circuit, which attempts to remove shading in the vertical scanning line direction.

3 in the figure, the maximum value detection circuit 11 detects the input terminal 12.
Compare the input video signal from the attenuator 13 with the output of the attenuator 13;
Output the larger value.

In analog memory 14, this output is delayed by one vertical scanning period, attenuated by a predetermined amount in attenuator 13, and input to maximum value detection circuit 11.

Therefore, since the maximum value detection circuit 11 compares the currently input video signal with the video signal at the corresponding point one vertical scan ago, the attenuation amount of the attenuator 13 is adjusted appropriately using the control signal C8. By selecting , it is possible to track the video signal in the horizontal scanning line direction and find the envelope of the white level of the video signal.

Since this envelope is a shading waveform of the original video signal, the shading is removed by correcting the shading using the gain control amplifier 15 using this envelope.

The video signal from which shading has been removed in this way is sent to a level tracking circuit 16 to track the black level, and a threshold level signal is output based on the tracking result and the set slice level SL.

The comparator 17 compares this threshold level with the video signal from the amplifier 15, and outputs from the output terminal 18 a binarized signal corresponding to the magnitude of both.

Such conventional shading removal circuits can follow the shading waveform well when the read character pattern (black level signal) is short in the horizontal scanning line direction, but when the character pattern becomes long in the horizontal scanning line direction, shading It becomes impossible to follow the waveform, and accurate quantization is not possible.
There was a possibility of misunderstanding.

Also, if the input video signal contains noise, the resulting white level envelope will be affected by this noise.
There was a problem that accurate shading removal could not be performed.

Furthermore, since the above-described shading removal circuit is of an analog type, it requires analog elements such as the maximum value detection circuit 11 degree analog memory 14 and the gain control type amplifier 15.

When such analog elements are used, it is difficult to adjust the circuit system, the operating range of each element becomes extremely narrow, and stable operation cannot be expected, it is expensive, and high integration is not possible. was there.

A main object of the present invention is to provide a shading removal circuit that can remove shading with extremely high accuracy.

Another object of the present invention is to provide a shading removal circuit that can obtain a shading waveform that is not affected by noise.

Still another object of the present invention is to provide a shading removal circuit that is inexpensive and easy to integrate.

To achieve such an objective, the present invention calculates the difference or ratio between the input video signal of the current scan and the white level envelope signal of the corresponding point of at least the previous scan, and calculates the difference or ratio according to the difference or ratio. The present invention is characterized in that a predetermined function is generated, the output of the function is added or multiplied by the envelope signal, and the result is output as a new white level envelope, that is, a shading waveform.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 4 shows a block configuration of an embodiment of a quantization circuit including a shading removal circuit according to the present invention.
22 is a subtraction circuit that calculates the difference between two input signals; 23 is a function generator that generates a desired function according to the input; 24 is an addition circuit that calculates the sum of the two input signals; 25 is a subtraction circuit that calculates the difference between two input signals; 2 is a delay circuit that delays an input signal by one vertical scanning period; 26 is a black level detection circuit that calculates the ratio of two inputs and outputs a black level signal from which shading has been removed;
T is a black level tracking circuit, 28 is a comparison circuit for obtaining a binary signal, and 29 is an output terminal for the binary signal.

Among these, a subtraction circuit 22, a function generator 23, an addition circuit 24
, delay circuit 25 constitutes a white level envelope tracing circuit.

For such a configuration, when the video signal x(nT) at time nT is inputted from the input terminal 21 to the subtraction circuit 22, the output y( A signal y((n ], )T:] which is obtained by delaying nT) by one vertical scanning period is similarly input to the subtraction circuit 22, and a signal corresponding to the difference is obtained from the subtraction circuit 22.

A function generator 23 generates a tracking function F(x) corresponding to the difference signal X from the subtraction circuit 22.

Figure 5 shows the input X and output F that are pressed to this function generator 23.
This is a characteristic diagram showing the relationship between the current input video signal x(nT) and the envelope output y(I(nT) l vertical scanning period ago).
1) While the difference signal Therefore, the predetermined value X,
While it is larger than Xl, the output F(x) is proportional to the difference signal X but has a small slope.
is 0.

FIG. 6 shows another example of the characteristics of the function generator 23.
Only in the range x2<x<x3 where the difference signal X is close to O, an output F(x) a3 x (a3 <82) with a very small slope can be obtained.

This means that when the difference signal This shows that it follows and absorbs noise.

Further, as can be seen from FIG. 5 and FIG. 6, in the function generator 22, when the difference signal X becomes smaller than the predetermined value X,
That is, if the current input video signal becomes extremely smaller than the envelope signal from one vertical scanning period ago, the envelope tracking is stopped from 1 to 1, and the envelope output from 1 from the previous vertical scanning period is changed as described later. Output as is.

The function F(x) obtained in this way by the function generator 22
and the output yC(n 1)T of the delay circuit 25 mentioned above]
The adder circuit 24 calculates the sum of
After a delay of the vertical scanning period, an output y(n 1)T is obtained.

Therefore, the output y (n T ) of the adder circuit 24 is expressed by the following equation (1).

As can be seen from FIGS. 5 and 6, in the range where the difference signal X is larger than X, its output F(x
) is expressed by the following equation (2), so the envelope output y(n
T) changes following the input video signal x(nT), but when the difference signal X becomes smaller than X, that is, when the scanning character pattern changes from the white level to the black level, the tracking function F (x) becomes zero, so the envelope output y(
nT) is the envelope output yl: (n-1
) T) will be kept as is.

Therefore, even if the character pattern becomes longer in the horizontal scanning line direction, the envelope output retains the signal of the previous vertical scanning line as it is, so the envelope output accurately represents the shading waveform.

Furthermore, as can be seen from the characteristics in Fig. 6, in the region where the difference signal It is possible to absorb a certain amount of noise caused by variations in reflection on the paper surface.

The envelope output y (nT) and the input video signal x (nT) obtained in this way are input to the black level detection circuit 26, and the black level signal X (nT) as shown in the following equation (3) is inputted to the black level detection circuit 26.
T ) is obtained.

As the output X (n T ) of the black level detection circuit 26, a black level video signal from which shading has been removed is obtained, and the output is directly input to the comparison circuit 28, while
It is input to the black level tracking circuit 27.

The black level tracking circuit 27 tracks the black level of the input signal, and uses the higher of the signal level corresponding to the tracking result and the set slice level signal as a threshold signal X.

(nT) is input to the comparator circuit 28.

The comparison circuit 28 compares the video signal X (n T ) from which shading has been removed with the threshold signal Xo (n T), and generates a binary signal depending on whether X (n T) is larger or smaller than Xo (n T). Outputs 1pa or 0°'.

The purpose of the black level tracking circuit is to increase the resolution by changing the slice level for binarization as described above.

In addition, in FIG. 4, the subtraction circuit 22 was used to find the difference between the two inputs x(n'r) and y[(n 1)T], but
Alternatively, a ratio between the two may be calculated.

In that case, the adder circuit 24 becomes a multiplication circuit that multiplies two inputs.

FIG. 1 shows an example of a specific configuration of the circuit shown in FIG. 4, which is realized using a digital circuit.

In the figure, 31 is an analog-digital converter (hereinafter referred to as an AD converter).

), 32 is a read-only memory (hereinafter referred to as ROM) that generates a logarithmic function of an input signal.

), 33 is a latch, 34 is a subtracter, 35 is a ROM that generates a desired function for the input signal, 36 is an adder, 37
is a latch, 38 is a shift register having a capacity for vertical scanning lines, and 39 is a multiplexer (hereinafter referred to as MPX).

) is shown. These circuits 34 to 39 constitute a white level envelope tracking circuit as shown in FIG.

Further, 41 is a subtracter, and 42 is a latch, which corresponds to the black level detection circuit 26 in FIG. 4.

43 is a subtracter; 44 is a ROM that generates a predetermined function for the input signal; 45 is an adder; 46 is a ROM that generates a predetermined function according to the input signal level and the set slice level SL; These latches constitute a black level tracking circuit 27 shown in FIG.

Furthermore, 48 is a latch, 49 is a comparator, and 50 is a latch, which correspond to the comparison circuit 28 in FIG.

It is assumed that the latches 33, 37, 42, 47, and 48 have a capacity that can store the number of quantization bits of each input signal.

Although not shown in the figure, it goes without saying that a clock generation circuit is provided that generates various clocks that control the operations of the above-mentioned circuits.

FIG. 8 shows the signal waveforms of each part in FIG. 7, where a is a waveform diagram of the video signal input from the input terminal 21, and b is the output waveform of the adder 36, which is the output of the white level envelope tracking circuit. In the figure, C is an output waveform diagram of the latch 42 which is the output of the black level detection circuit, and d is the ROM which is the output of the black level tracking circuit.
46 output waveform diagram, e is latch 5 which is the output of the comparison circuit
0 is an output waveform diagram.

FIG. 9 shows an example of the functions generated by the ROM buttons 35 and 44 in FIG.
F1H(x) shown by the solid line represents a function selectively generated in the ROM35, and F2V(X) shown by the solid line represents the function generated selectively in the ROM35.
It represents a function generated in M44.

Further, FIG. 10 shows a function generated in the ROM 46 of FIG. 7, and shows that the characteristics shown by a solid line or a dotted line can be selected according to an external selection signal.

Hereinafter, the operation of the circuit shown in FIG. 7 will be explained in detail with reference to FIGS. 8 to 10.

A video signal x (as shown in FIG. 8a) is input from the input terminal 21.
When the input video signal x(nT) is inputted, the input video signal
32.

The ROM 32 converts the input digital signal into a logarithmic relationship and stores it in the latch 33 as a multivalued digital signal.

Therefore, the output x'(nT) of the latch 33 is expressed by the following equation.

Here, the reason for converting the output of the AD converter 31 into a logarithmic function is that when the input video signal is normalized according to the equation (1) using the shading waveform in the black level detection circuit as described later, This was done in view of the fact that it is extremely difficult to perform division digitally.

At the same time, it has the effect of keeping the average value of noise caused by variations in the reflectance of the paper surface constant.

The white level envelope tracking circuit includes outputs j and n of the latch 33.
Since x(nT) is input, its output is ffny(
nT) and is input to the subtracter 41.

Therefore, the output X'(nT) of the latch 42 is (5)
As shown in the equation, the output tn of the white level envelope tracking circuit
y (rtT) and logarithmically transformed input video signal/
-nx(nT), and if A, which replaces equation (3), is set to 1, the following equation (6) is obtained.

The relationship expressed by equation (6) is shown in FIG. 11 in a graph.

Therefore, originally, the output of this latch 42 should be X'(n
T) must be converted into the normalized signal x(nT) shown in FIG. 11, and this output XI'(nT) is shown in FIG.
As shown in , the threshold level T of the original normalized signal X (n T ), which corresponds to the black level signal, is equal to the black level signal
Since the signal X'(nT) corresponds to the threshold level T' at '(nT), in the embodiment shown in FIG. 7, the signal X'(nT) is used as it is and is binarized using the threshold level T'.

On the other hand, as described above, the output cuff nxn of the latch 33 is input to the subtracter 34, and as described later,
An output obtained by delaying the envelope output tny (nT) by one bit within the vertical scanning period or an output delayed by one vertical scanning period is input, and the output X of the subtracter 34 is a combination of the output of the latch 33 and the delayed output. The difference will be output.

For example, if the output obtained by delaying the envelope output tny (nT) by one vertical scanning period is selected by the MPX 39, the subtracter 34
The output X is expressed by the following equation (7).

By the way, when tracking the white level envelope, the tracking direction is the vertical scanning line direction, which tracks data between adjacent pips within the same scanning period of the line sensor, and the corresponding bit direction over multiple vertical scans. In view of the fact that there is a horizontal scanning line direction in which data is traced in between, in the embodiment of FIG. 7, the tracking function generated in the ROM 35 is changed to FIV( ) and Fl
(X), the latch 37 that produces an output delayed by 1 bit within the vertical scanning period and the shift register 38 that produces the output delayed by 1 vertical scanning period are connected to the MPX 39.
It can be switched with .

First, the switching operation will be explained in detail.

When used as a document reader, as mentioned above,
The line sensor is fixed and readings are taken as the form moves and passes under the line sensor.

Therefore, since the leading portion of the form is blank, tracking is performed in the vertical scanning line direction in order to capture a rough white level envelope during this blank period.

Therefore, by using the switching signal SW, the vertical function "tv(x)" of the ROM 35 is selected, and the output of the latch 37 is
Select with PX39.

This latch output is used to perform vertical tracking as will be described later, and the results are sequentially stored in the shift register 38, so that a rough white level envelope is obtained in the shift register 38 during the vertical scanning period.

Thereafter, the tracking direction is switched, and the horizontal function FIH(X) of the ROM 35 and the output of the shift register 38 are selected to perform tracking in the horizontal scanning line direction.

In this case, if the shift register 3.8 used can be initialized, there is no need to perform the switching described above, and the shift register 38 is initialized when the blank part of the form is reached, and the horizontal scanning is performed. If line direction tracking is performed, a white level envelope can be obtained in several vertical scanning periods thereafter.

On the other hand, when used for page league, when the reference mark at the right edge of the form is detected, the feeding of the form is stopped and the line sensor is moved in the direction of the character string to perform reading. It is necessary to track the line direction.

This is because when tracking is performed using the horizontal tracking function FIH(X) as described later, if the difference between the current input video signal and the previous video signal becomes large, the previous video signal may be held as the envelope output. Therefore, if the black reference mark is at the right end of the form, a correct white level envelope cannot be obtained.

Therefore, in this case, tracking is performed in the vertical scanning line direction when detecting the reference mark, and tracking is performed in the horizontal scanning line direction when reading characters, that is, when the line sensor moves at a constant speed in the reading direction.

Next, the white level envelope tracking function obtained in the ROM 35 will be explained in detail.

In the ROM 35, Fl y (X
) l "IH(X).

That is, the vertical tracking function Fty(x) is defined in the following three areas.

In the region where the input X is positive and larger than X3, the function F, ■(X
) is defined by x 2 < x (in the region of x 3 , the function Ft V (X ) is defined by the following equation (9).

However, 0 (a ()).

Moreover, in the area of x (x 2), it is defined as Flv(x)=C.

Further, the horizontal tracking function FIH(X) is −X.

In the region up to (X, <X<x2), the vertical tracking function F
Same as t v (x), but in the domain of x(x 1, Fly(x) is defined as zero.

Therefore, in accordance with the output of the subtracter 34, that is, in accordance with the difference X between the current video signal and the white level envelope output obtained at the previous timing, the above-mentioned function output is obtained, and this is added to the adder 36. It is added to the white level envelope output at the previous timing to determine a new white level envelope M skew force as shown in FIG.

Therefore, in the area of x) This immediately modifies the envelope output to a value close to the input level of the current input video signal.

In the next region close to zero, x2 (x<x3, two functions are realized.

One of them is the function of absorbing the noise of the white level of the video signal, and for this purpose, the slope of the function is selected to be small.

This is because the closer the gradient of the function is to 1, the faster the tracking will be.
The closer the slope is to zero, the slower the tracking will be, so noise that appears as small changes in white level will be absorbed.

By the way, in general, the higher the white level of the video signal, the greater the noise, but as mentioned above, the input video signal is logarithmically converted, so in the converted output, the magnitude of the noise component is equal to the white level. is almost the same regardless of the height.

Therefore, the area x 2 (x (x 3 ) can be set to a constant length regardless of the size of the white level of the video signal.

Another function is to press the button in the steady state to give a constant bias α to the white level envelope output signal, and for that purpose, shift the function in the X direction and when X becomes -α,
The function output is set to zero.

The reason for providing this function is that the black level detection circuit described below is composed of a subtracter 41, so that the difference between its output, that is, the white level envelope output at the previous timing and the current input video signal, is affected by noise. The goal is to prevent it from becoming negative.

In the region where X 2 >
When X is less than or equal to, the tracking function Fxv(x)> and “x
There will be a difference in i-i(x).

In other words, in the function FIV(X), even if X<X, the constant value -
C, but the function F I H (x) is zero.

This is because, in general, changes in shading in the vertical scanning line direction are relatively large, so even when the difference becomes large in the negative direction, accurate tracking may not be possible unless tracking is performed. However, changes in shading in the horizontal scanning line direction It's relatively small, so it's not much of a problem.

When the difference becomes large in this way, by preventing the envelope output from following the input video signal, even if the black level video signal continues for a long time, the shading waveform will follow the black level and the output will change. You can't go wrong.

Next, the subtracter 41 subtracts the input video signal from the white level envelope output, as described above.
A black level signal as shown in Figure C is obtained.

At this time, the black level signal has a bias α to prevent the value of the black level signal from becoming negative due to noise in the input video signal.

In this way, a black level signal with corrected shading is obtained, but when this signal is sliced at a constant value and binarized, as shown in D in Figure 8C, when the black levels are close to each other, That separation is not possible.

Therefore, in the black level tracking circuit, the black level is set to a certain constant value β.
When this occurs, the threshold level is varied in accordance with the black level of the input video signal, thereby making it possible to separate adjacent patterns.

Therefore, the configuration of the black level tracking circuit is almost the same as the configuration of the white level envelope tracking circuit, and the shift register 38, M
PX39 simply does not have a switching signal SW.

Then, the ROM 44 generates a tracking function F2V(X) shown by the solid line in FIG.

That is, when the input black level signal X increases, the output amount eF2v (, F2v(x)−ax(0(a(1
), and the output gradually decreases.

This is a so-called digital filter function.

This function output is added to the previous black level signal by one bit in the vertical scanning line direction by an adder 45.

The addition output is input to the ROM 46 and also to the latch 47.

As mentioned above, the output of the launch 47 is 1 in the vertical scanning line direction.
The outputs of 1-1 corresponding to the previous black level signal are input to an adder 45 and a subtracter 43, respectively.

The subtracter 43 calculates the difference between the current black level signal from the latch 42 and the previous bit black level signal from the latch 47, and the difference output is input to the ROM 44.

In the ROM 46, the signal from the adder 45 is input X,
A function ps(x
) occurs.

At this time, by changing the slice level SL, a function as shown by the dotted line can be generated.

Therefore, the output of the ROM 46 has a waveform as shown in FIG. 8d, which is input to the comparator 49 as a threshold level.

Since the current black level signal is inputted to the comparator 49 from the latch 48, these inputs are compared and the latch 5
A binary code as shown in FIG. 8e is obtained from the 0 output terminal 29.

In the example shown in FIG. 7 described above, all the circuits are realized by digital circuits, so that not only very stable operation can be performed, but also high integration and low cost can be achieved.

Although the circuit shown in FIG. 4 can be implemented using an analog circuit, the configuration of the function generator and the like becomes complicated, so it is more preferable to implement the circuit using a digital circuit as described above.

In addition, in the configuration shown in FIG. 4, instead of X(t), X'(t
) is used, but if you want to obtain a multivalued video signal (shading corrected signal) instead of a binary one, you can configure the function in ROM and connect it to the output of the black level detection circuit 26. .

Furthermore, the black level tracking circuit 27 shown in FIG. 4 may be omitted and a constant threshold level may be input to the comparison circuit 28.

Further, in the above-described embodiment, a case was explained in which a black and white gray pattern signal is quantized, but the invention is not limited to this. It goes without saying that this method can be applied to quantizing pattern information that causes shading.

Further, the present invention is not limited to OCR, but can be similarly applied to general pattern recognition, facsimile, etc.

As described above, according to the present invention, when the difference or ratio between the current video signal and the envelope output of the previous video signal becomes less than or equal to a predetermined value, the envelope output of the previous video signal is held. With this configuration, an extremely accurate shading waveform can be obtained, and by using a function generator that can generate a desired function, the influence of noise contained in the input video signal can be reduced.

Furthermore, since the structure can be easily realized using a digital circuit, it has the characteristics of high integration and low cost.

[Brief explanation of drawings]

Figure 1 is a schematic configuration diagram of OCR, Figure 2 is an explanatory diagram of the photoelectric conversion section of OCR, Figure 3 is a configuration diagram of a quantization circuit including a conventional shading removal circuit, and Figure 4 is shading removal according to the present invention. 5 and 6 are characteristic diagrams of the function generator of FIG. 4, and FIG. 7 is a configuration diagram of an example of the specific configuration of FIG. 4, Figure 8 is a waveform diagram of each part of Figure 7, the 9th rotation and Figure 10 are the 7th rotation.
FIG. 2 is an input-output characteristic diagram of the ROM shown in the figure, and a characteristic diagram illustrating the first operation. 21...Video signal input terminal, 21 is shown in FIG. 72...Subtraction circuit, 23...Function generator, 24...Addition circuit, 25 ...Delay circuit, 26...
Black V bell detection circuit, 28... Comparison circuit.

Claims (1)

[Claims] 1. Input means for inputting a video signal that changes between two different signal levels and shading occurs in at least one of the signal levels, and a current scanning video signal and a previous scanning video signal from the input means. detecting means for obtaining a difference or ratio signal between the signal level of one of the scanning video signals and the envelope output; and when the signal from the detecting means is above a certain value, outputting a value corresponding to the signal; function generating means that outputs a predetermined value when the signal from the means is below a certain value; adding or multiplying the output of the function generating means and the envelope output of the video signal of the previous scan, and applying the result to the current scan; calculation means for outputting an envelope of the signal level of one of the video signals;
delay means for delaying the output of the arithmetic means to obtain an envelope output of the video signal of the previous scan; and shading removal according to the envelope output from the arithmetic means and the video signal of the current scan from the input means. and a comparison means for obtaining a video signal obtained by the shading removal circuit. 2. The function generated by the function generating means has a gentle slope within a range where the envelope of the signal level of one of the video signals of the current scan and the video signal of the previous scan is similar. A shading removal circuit according to claim 1. 3. The shading removal circuit according to claim 1, wherein said input means comprises conversion means for converting a video signal into a logarithm. 4. Claim 4, characterized in that the input means comprises an analog-to-digital conversion means for converting a video signal into a multivalued digital signal, and a logarithmic conversion means for converting the signal from the conversion means into a logarithm. The shading removal circuit according to item 1. 5. The shading removal circuit according to claim 4, wherein said function generating means comprises a do-only memory. 6. The shading removal circuit according to claim 4, wherein said logarithmic conversion means comprises a read-only memory.