JPS6236977A - Picture reader - Google Patents

Abstract

PURPOSE:To attain a proper binary-coding processing with a few memories and simple hardware by applying operation of frequencies at a prescribed level before and after detection as well as prescribed detected frequencies in forming a histogram. CONSTITUTION:When an optical system is subject to forward scanning and a coordinate on an original platen is detected, a CPU 211 calculates an area to sample a peak value for deciding a slice level for binary-coding from a coordinate data. In starting reversing, after a main scanning line synchronizing signal is counted by a prescribed distance, white/black peak values are detected to form a white peak histogram and a black peak histogram by using white/ black peak histogram areas prepared in a RAM 213. Then density value representing the maximum frequency of the black peak histogram and the white peak histogram are decided as the information part and the back ground part of an original and the median of them in decided as the slice level. A comparator 210 uses the said slice level to binary-code a picture signal from a latch 203 and to output the result.

Description

[Detailed description of the invention] (Field) The present invention relates to an image reading device.

(Prior art) As a method for binarizing image signals, there is a method that statistically processes the image density to determine the slice level. Unable to obtain stable output.

(Objective) An object of the present invention is to provide an image reading device that eliminates the above-mentioned drawbacks and is capable of performing appropriate binarization processing with a small amount of memory and a simple hardware circuit.

The present invention also enables stable binarization processing that is less affected by noise and changes in light amount.

(Structure) The present invention includes a first means for detecting a predetermined value of document image density for each main scanning line, a second means for forming a histogram of the detected values, and a statistical method based on the histogram obtained by the second means. a third means for determining the slice level;
The image reading device is characterized in that when forming a histogram in the second means, not only the frequency of the detected predetermined value but also the frequency of a predetermined level before and after the detected value is calculated. - For example, better effects can be obtained by setting the predetermined values to the white peak and black peak values of density.

(Embodiment) FIG. 1 is a schematic diagram of a document reading device (hereinafter referred to as a reader) to which the present invention can be applied. Original 102 placed on original platen glass 101 while being pressed by original force/<-110K
In order to read the image information, an image sensor 103 such as 00J) is used, and the illumination light from the light R104 is reflected on the surface of the original 102 and is transmitted onto the image sensor +03 by the lens 10B via the mirrors 105 and 106 and 107. is imaged. Light source 104, mirror 105 and ξ sea 106, 10 is 221
It is designed to move at a relative speed of This optical unit is operated by a PLL 1ll by a DC servo motor 109.
Move from left to right at a constant speed while applying J 1111. This moving speed is 45gm/s depending on the magnification on the outward journey.
Variable from ea to 500 points 31/8 e Q, always 800 trends/B@O on the return trip. The main scanning direction (hereinafter referred to as the
The optical unit is moved from the left end to the right end while reading at a resolution of oes/1 nch, and then moved back to the left end to complete one scan.

FIG. 2 shows a block diagram of a circuit that processes image signals from the image sensor 105. The image signal VD read by the image sensor 103 is converted into a 6-bit digital signal by the A/1) converter 201, and is converted to a 6-bit digital signal via the latch 202. 208.

The comparator 204 compares the 6-bit image signal sent from the latch 202 with the 6-bit image signal sent from the latch 203 one clock ago, and if the latch 202
If the new image signal sent from 02 is smaller, a comparison output is output to the AND gate 206.

AND gate 206 sends the comparison output from comparator 204 to latch 205 in synchronization with sampling clock SQL.

The comparator 207 compares the 6-bit image signal sent from the latch 202 with the 6-bit image signal sent from the latch 203 one clock ago.
If the new image signal sent from 2 is larger, a comparison output is output to AND gate 209. AND gate 209 synchronizes the comparison output from comparator 207 with sampling clock 80'll and sends it to latch 208.

When the latches 205 and 208 receive the comparator output, they send the image signal sent from the latch 202 to the 0PU 211. In addition to the comparator output and the sampling clock 80L, the 206 and 209 have an image sensor 1.
Enable signal Et indicating the valid section of the image signal from 03
N is input, and the comparison result of the image signal in a predetermined section for each main scanning line is sent from the latch 205 and 208 to the 0PU 211.

The 0PU 211 stops the image signals from the latches 205 and 208 in synchronization with the main scanning line synchronization signal MS, and distinguishes between the lowest density level (hereinafter referred to as white peak) and the highest density level (hereinafter referred to as black peak) of each main scanning line. peaks) can be detected.

The aptr 21t sends the slice level to the comparator 210, which determines the slice level based on the white peak and black peak detected for each line in this way using an algorithm described later. In the comparator 210, the image signal from the latch 203 and the 0PT
Compare the slice levels from T211 and create a binary signal V4
D] Generate IiO.

FIG. 3 shows a state in which a document is placed on the document table 101 of the reader shown in FIG. In this case, 1 on the document table 101,
Coordinates p of four points on the document from the reference point SF in seven directions
+(:cl, 71), P2(X2.72), I)
5 (X3.

73) *"("97') is detected by preliminary scanning with the optical system.

The document cover 110 in FIG. 1 is mirror-finished so that image data outside the area where the document is placed is always black data.

The circuit diagram of FIG. 4 shows the logic for detecting the coordinates. The main scanning counter 351 is a down counter and represents the scanning position in one main scanning line. This counter is set to the maximum value in the main scanning direction (X direction) by the horizontal synchronizing signal H8YNO, and counts down every time the image data clock OLK is input. The sub-scanning counter 552 is an up counter, which is reset to @lON at the rising edge of vsyNa (image leading edge signal), counts up at the H8YNO signal, and represents the scanning position in the sub-scanning direction. .

is input to the shift register 301 in 8-bit units.

When the 8-bit input is completed, the gate circuit 302 checks whether all of the 8-bit data is a white image, and the yxa
If so, output 1 to the signal line 50M. F/'IP 304 is set when the first 8-bit white appears after scanning the original. This ψ is reset in advance by VSYNO.

After that, it is left set until the next VSYNO comes.

When the F/7304 cassette is inserted, the latch V1P305
The value of the main scanning counter 351 at that time is loaded.

This becomes the X1 coordinate value. The value of the sub-scanning counter 350 at that time is still loaded into the latch 306, and this becomes the y coordinate value. Therefore, Pi(Xs y 7t ) can be found.

Also, each time 1 is output to the signal 603, the main scanning counter 351
Load the value from into latch 307. The value from the main scanning counter when the first 8 bits of white appear is latch 3.
07, the latch 310 (which has VBYN
A comparator 309 compares the data with the data (set to the maximum value in the X direction at time O). If latch 3
If the data in latch 307 is smaller, the data in latch 307 is loaded into latch 310. Also, at this time, the value of the sub-scanning counter is loaded into the latch 311. This operation is processed until the next 8 bits enter shift register 301.

In this way, if the data of the latch 307 and the latch 310 is applied to the entire image area, the minimum value of the document area in the X direction remains in the latch 510, and the coordinate in the y direction at this time is
will remain in the Since the main scanning counter 351 is a down counter, the coordinate corresponding to the minimum value in the X direction represents the coordinate farthest from Sp in the main scanning direction, which is Ps (Xs
*7g).

The F/IP 512 outputs a horizontal synchronizing signal H8YN at ψ, which is set when 8-bit white appears for the first time in each main scanning line.
It is reset with O, the first 8 bits are set to white, and the next H
Hold until 8YNO. When the XPAP 512 is set, the value of the main scanning counter corresponding to the position of the first white signal appearing in one line is set in the latch 313. Then, the latch 315 and the comparator 516 compare the magnitude. Latte 315 is set to the minimum value of 10'' in the X direction at the time of VByMO occurrence.
If the data of 5 is smaller than or equal to the data of latch 313, signal 317 becomes active and latch 313
data is loaded into latch 315. This action is H
This is done between 87M0 and HBfNO.

If the above comparison operation is performed for the entire image area, latch 5
15, the maximum value of the document coordinates in the X direction, that is, the X coordinate of the white signal of the point closest to the scanning start point in the main scanning direction remains K. This is X.

It is. Also, when signal line 317 outputs, the value from the sub-scan is loaded into latch 318. This becomes y, and the Pt ("t e 7t ) coordinate is obtained.

Each time 8-bit white appears in the entire image area, the latches 319 and 320 are loaded with the main scanning counter value and the sub-scanning counter value at that time.

Therefore, when the document pre-scanning is completed, the count value at the time when 8-bit white appears last remains on the counter. This is P4 ("a t 74 ).

The above eight latches (51M, 3rd, 520, 818.
The data line of 15°310.515,519) is connected to the pass line BUEI of CPH in FIG.
will read this data at the end of the previous scan.

Figure 5 shows the flow of the document reading seconds sequence. First, in step 501, the optical system performs forward scanning from the left end to the right end in Figure 1, and detects the coordinates of the original on the document table as described above. .

Next, in step 508, the area in which the peak value for determining the slice level for binarization should be sampled is calculated from the single bond data detected in step 501.

For example, from the coordinates detected for a document such as the shaded area in Figure 3, the peak value sampling area of this document is 7.
Selecting the rectangular area bounded by g+7 and xI+''4 actually prompts.

This is because the original is usually placed as parallel to the original table as possible, and even if it is placed at a slight angle as shown in Figure 3, there is no risk of picking up information from outside the original. be. Of course, the sampling area may be determined using other methods. As can be seen from FIG. 6, when the document coordinate detection is completed, the optical system is at a point in the sub-scanning direction YmaX and the peak value sampling start point y2 and end point y1 are known, so steps 504, 505, and 506 are executed. I can make a schedule. That is, when the return movement is started in step 506, 0PU21 + is the distance (7maX -
7f) After adding the main scanning line synchronization signal by a corresponding amount,
After starting the detection of the white peak value/black peak value described above, and counting the main scanning line synchronization signal for the distance corresponding to the distance (72+7s) from that point, the detection of the peak value is finished, and furthermore, after counting the main scanning line synchronization signal for the distance corresponding to the distance (72+7s), After counting the main scanning line synchronization signals for a corresponding number of times, the backward motion is stopped.

Further, each time the CPU 2N detects a peak value for each line between steps 504 and 505, it forms a white peak histogram and a black peak histogram by means to be described later.

It goes without saying that at step 504, at the start of peak value detection, the above-mentioned energy pull signal EN is set corresponding to the detection points 4) JXI and X4Vc as shown in FIG.

With the above operations, the white peak and black peak of the image density for each main scanning line in the document located at an arbitrary position 1t on the document table are detected, and each histogram is formed.

Next, the procedure for creating a histogram will be explained.

Using the above-described means, the PU 211 renormalizes the black peak value and white peak value for each main scanning line from within the document area. The black peak on the younger brother 1 main scanning line is BPl, and the white peak is wp.
Since the image data of A/D l is a 6-bit value, each value takes any value from 0 to 63 and BP
It is i\wp1.

Create a histogram using a 2-byte black peak histogram area at 64 and a 2-byte white peak histogram area at 64 prepared in the CPU RAM.
Figure 11 shows the RAM '7. As mentioned above, the backward movement speed of the optical system is about 800nφθC, the maximum peak value detection area is 420 nm in the sub-scanning direction, and the main scanning line synchronization is 552.7μsθC, so the maximum number of lines to be sampled, that is, black, The number of samples for each white peak value is approximately 1,500, so if 2 bytes are prepared to count the frequency of each detected peak value, the number of samples will be approximately 40 times, and even with the histogram creation means according to the present invention, which will be described later. That's enough.

The simplest histogram creation means is a method of counting only the frequencies of areas corresponding to the detected data BPl and WPi, and this is called method (0). For example, BP1=
62,! When vPi = 1, the RAM addresses ADRB + 2, ADRB + 1, 2) < 4) in Fig. 11 are used.
The contents of the 2-byte area of ADRw+125 are each set to +1.

After that, wait for the first main scanning line synchronization signal MS, and
The same operation is performed again by taking the data WPi+I and BPi+1 from the line, and this is continued with the sample H completed from the tape 504 to the tape 505.

In each of the black peak histogram A and the white peak histogram A created in this way, the density values showing the maximum frequency are estimated as the information part and background part densities of the document, and from these densities, for example, their median value is calculated. Determine the slice level.

An example of the black peak histogram A created in this way is shown in FIG. 7-1, and an example of the white peak histogram is shown in FIG. 7-2. In this example, the density of the information area of the document is determined from the black peak histogram by 50. Further, the background density of the original can be estimated to be 10 from the white peak histogram, and the median value of 20 can be used as the streak level for binarization, as shown in FIG. 7-1.7-2, for example.

FIG. 7-3 shows an example of a document on which the histograms shown in FIGS. 47-1 and 7-2 are generated and an example of its output copy.

In FIG. 7-3, the text rAJ, which is the information part of the original image, has a slightly low density, which indicates that it is reproduced black by the above-mentioned binarization and output as a copy.

By the way, consider the manuscripts shown in Figures 8-3 and @9-3. In addition to the original shown in FIG. 7-3, this document has a thin line in the main scanning direction of higher and more uniform density than Kf1v@portion "A".

Two examples of histograms created for such originals are shown in Figures 8-1.8-2, 9-1, and 9-2.

Both FIGS. 8-2 and 9-2 are white peak histograms, which are exactly the same as the white peak histogram shown in FIG. 7-2 for the sake of explanation, and the background density can be estimated to be 10. Figures 8-1 and 9-1 are black peak histograms.

These two histograms both have the same large peak having a maximum density of 3°, and this peak is formed by the black peak value sampled from the original information part molecule AJ.

However, there are differences in the histograms formed by the black peak values sampled from the thin line area, and both have a density of 6.
0 is taken as the maximum value, but in Figure 8-1, some dispersion can be seen on both sides, and the frequency relative to the maximum value of 60 is the maximum value of 3.
smaller than the frequency for o.

On the other hand, in FIG. 9-1, a steep peak is generated only for density 60, and its frequency is greater than that for density 30. For example, as described above, in the case where the density value indicating the maximum frequency in the histogram is estimated as the density of the document information part and the background part, in the case of Figure 8-1, the information part density is estimated to be 30, and the density value of the information part is estimated to be 30, and In the case of the figure, it is estimated to be 60.

Furthermore, if the slice level is the median value with the estimated background density, then in FIG. 8, 20. In FIG. 9, it is 35. Therefore, in the output copy example of the result of binarization at each slice level, both the thin line and the "mu" are reproduced in black in Figure 8-3, and in Figure 9-3.
In the figure, rAJ is not copied, but only thin line C is copied.

As mentioned above, the reason why the histograms formed for the same document are different is that the amount of light is unified for each reading operation, and the mechanical scanning of the optical system and the mechanical reading are originally performed asynchronously. It is conceivable that the relative positional relationship between the original image and the main scanning line may change. For example, if there are many samplings on thin lines, the result will be as shown in FIG. 9, and if there are many samplings between thin lines, the result will be K as shown in FIG. 8.

A histogram creation method (1) for eliminating the instability of binarization where the output results differ for each reading operation for the same document will be described below.

For example, when the detected black peak is BPi:50, the address ADRB+60 of the histogram area and ADRB
The contents of the 2-byte area +61 are counted up by 1, and the address ADRB+58 of the histogram area before and after that, that is, corresponding to BPi=29.

Address ADRB-)-62 of the 2-byte area of ADRB+59 and the histogram area corresponding to BPi:31
.

The contents of the 2-byte area of ADRB+63 are also counted up by 1 each.

Generally speaking, for the detected black peak BPi: J, a 2-byte histogram area ADuB+2(l-n) corresponding to each of 2n thousand and one consecutive densities from BPi = l-n to BPi = l +n, nR:a+2
(J-n)+1 to ADRB+2(A'+n), AD
Count up by 1 for each 2n+1 2-byte area up to RB+2(A'+n)+1, and count up the continuous values from WPi=R-m to BPl=R-m for the detected white peak WP,j:H. 2m+1 2-byte histogram area corresponding to each density ADRW+2 (
R-m), ADRW+2 (R-m)+1 to ADRB+
2(R+m), ADRB+2(R+m)+1, each of 2m+1 2-byte areas is counted up by 1.

1.2 with n and m predetermined constants
．． 3rd place would be appropriate. Further, n and m may be equal.

In other words, the 10th example shows an example in which three consecutive areas are counted up by 1 for one sampling.
As shown in the figure.

The black peak histograms shown in Figures 8-1 and 9-1 were re-formed using the above method (INK), and part A in Figure 10 corresponds to part A' in Figure 9-1. However, part B corresponds to part B' in FIG. 8-1.

As can be seen from Fig. 10, even if an unstable histogram is formed as shown in Figs. 8-1 and 9-1 according to the method (2), according to the above method (1) K, The density of the information part can be estimated to be 30, and the output as shown in FIG. 8-3 can be stably obtained.

In the above explanation, we have given an example where a thin line with uniform high density affects the black peak histogram, but in addition to that, dirt on the platen glass or noise-like dots on the original image can affect the black peak histogram. Method (1) is also effective when affecting the white peak histogram.

In addition, if black/white peak histograms are created using method (11), when estimating the information density and background density of the original image, it is possible to simply detect the density level that indicates the maximum frequency in each histogram without calculating the integration or variance. Dewell CPH
The calculation time can be reduced.

Furthermore, when counting up each histogram area in method (1), instead of uniformly counting up each area determined by ff1n and m by 1, a method (IT) may be considered in which the value to be counted up is weighted. .

For example, when the detected black peak EP1 = l, BP,
j=A! For +n, count up by 1 and BPi2l
2 for -n+1, 3 for A-n+2, etc.
. . , n+1 for l±0. About A'+1 n
For l . . . , /+n, a method of counting 1 can be considered. FIG. 12 shows the side temperature diagram of the black peak histogram obtained by method (summer) regenerated by method (II).

As can be seen from FIG. 12, method (1) also allows noise components to be removed and stable binarization to be achieved, as with method (1).

Furthermore, in method (■), the count-up value with the highest frequency is given to the detected peak value, and the count-up value decreases as the distance from the detected peak value increases. It can be seen that tall mountains with the same width are formed.

In other words, it is possible to form a histogram that respects the detected data and is more faithful to the data while excluding unstable housing. An example is shown in FIG.

If method (0) is used, a histogram as shown in Figure NX14-1 is given. If a histogram is generated using method (1), where n = m = I, it will be as shown in Figure 14-2, and the concentration will show the peak of frequency. It can be seen that the value has changed from the detected data.

This is because there was no weighting, so the influence of count-up values around the detected data was too large. So n=
Figure 14-3 shows a histogram that has been regenerated using method (I) where m = I.
It can be seen that the jagged histogram in Figure 4-1 has been smoothly shaped.

Figure 13-+, 2.3 shows the flow of each histogram creation method. Figure 4151 shows the simplest method described above, when the sample is started at the tape 504 in Figure 5 (θp
tsoo), write 0 to the line number No. and all histogram areas to initialize 5p1301) First, detect the peak value of the first line (spl, !102) ap150
3) , 1 count up only the contents of the histogram area corresponding to the peak value, and 5p below (sp + 3o4)
Steps 1302 to sp1+04 are performed for each line until the end of the sample (sp+5os).

FIG. 13-2 shows the method (2) described above, and will explain the ap1310 and subsequent steps that are different from those in FIG. 13-1. When the peak value is detected (8p1309), according to the constant m mentioned above, when wpi=H, ADRW+2(R-m), ADRW+2(R-m
)+l to ADRW+ 2 (R+m), ADRW+
2. Count up the contents of each 2-byte area by 1 up to (R+m)+1 ((ap1311) Set -m in area a on the RAM as a counter. 5i
p1510) 5pL511 until a becomes -m to m
Continuing, (ap1513\Furthermore, set -n in area a as a counter in the same way for BPi2A'.ap1514) ADRB+2(7+a), MuDRB
1 count up (gp1315) of the contents of the 2-byte area of +2(10a)+1 is repeated 2n+1 times until a becomes n from -n (sp1317), and further repeats ap130B to ap1517 until the sample ends.

Figure 13-3 shows the flow of method (II).
Well then (b2m+ 1- l a l (s+p+32
4) or b=n+ 1-1 a l (ap132
9) is counted up.

lal is the absolute value of a.

(Effects) By creating a histogram using the method described above, the influence of noise components on the original image can be removed and stable binarization can be performed.

[Brief explanation of drawings]

Fig. 1 is a cross-sectional schematic diagram of a reader according to the present invention, Fig. 2 is an image signal processing block circuit diagram, Fig. 3 is a diagram showing a document placed on a document table, and Fig. 4 is a document position detection block. Circuit diagram, Figure 5 is an image reading sequence flow diagram, Figure 6 is a schematic image reading sequence diagram, Figures 7-1, 7-2, 8-1, 8-2, 9- Figures 1 and 9-2 are histogram diagrams, Figures 7-3, 8-3, and 9-3 are original output copy diagrams, Figure 10 is a histogram diagram created according to the present invention, and Figure 11. is histogram creation diagram RA
M-up diagram, Fig. 12 is a histogram diagram created according to the present invention, Figs. In FIG. 11, 102 is an original document and 103 is a reading section. O N Guto News-V4-1 Four Swords Yako AO- -1+1 Shuken Xing NU

Claims (2)

[Claims] (1) A first means for detecting a predetermined value of document image density for each main scanning line, a second means for forming a histogram of the detected values, and a slice level statistically calculated from the histogram obtained by the second means. , and when the second means forms a histogram, calculates not only the frequency of the detected predetermined value but also the frequency of predetermined levels before and after the detected value. Characteristic image reading device. (2) An image reading device according to claim 1, wherein a predetermined weighting is performed in calculating the frequency of the detected value and the frequency of a predetermined level before and after the detected value.