JPS61214865A - Picture processing device - Google Patents

Abstract

PURPOSE:To eliminate bad influences of dirt, dust, etc. other than an original picture by detecting the picture density levels of the original picture and invalidating levels higher or lower than a prescribed level out of detected picture density levels. CONSTITUTION:A CPU 211 consisting of a microcomputer takes in picture signals from latches 205 and 208 synchronously with a synchronizing signal MS for every one main scanning line to detect the lowest density level (white peak) and the highest density level (black peak) of each main scanning line. By this constitution, picture density levels of the original are detected to process the picture signal obtained by reading the original on a basis of detected picture density levels, and levels higher or lower than a prescribed level out of detected picture density levels are invalidated to eliminate effectively band influences due to dirt, dust, etc. other than the original picture, and an optimum picture processing is performed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to an image processing device that electrically processes a document image.

[Prior art]

2. Description of the Related Art Conventionally, image processing apparatuses have been proposed that photoelectrically read an image of a document such as a facsimile and electrically process the read image signal. In such image processing apparatuses, digital processing has recently become common, as it has advantages such as ease of signal processing and resistance to the influence of external noise.

Therefore, it is necessary to accurately convert the original image into a digital image signal.

By the way, the conditions (density, size, etc.) of original images to be read vary widely, and it is impossible to handle all these original images with the same processing. Therefore, a configuration is adopted in which an operator determines the state of the document image and adjusts the image processing operation based on this judgment. However, the adjustment operation is troublesome, and there are cases where good image processing cannot be performed due to incorrect adjustment.

Therefore, it is conceivable to provide a function to detect the state of the document image to be read and automatically execute an image processing operation according to the document image according to the detection result. According to this, the operator's adjustment effort as mentioned above can be saved, and
Even an operator unfamiliar with the apparatus can perform image processing to a certain degree.

However, in such automatic processing, for example, when the image density of a document is detected and the document image is digitized based on the detected image density, dirt, dust, etc. other than the document may be mistakenly detected as the document image. This may cause inconveniences such as the digitization operation not being performed accurately.

〔the purpose〕

The present invention has been made in view of the above points, and it is possible to automatically perform image processing suitable for the condition of a document image to be read, and to eliminate inconveniences in this automatic processing, thereby creating an image that can be subjected to optimal image processing. The purpose is to provide processing equipment.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a schematic diagram of a document reading device to which the present invention is applicable. In order to read the image information of the original 102 held by the original cover 110 and placed on the original platen lot, an image sensor 1 such as a CCD line sensor consisting of several thousand light receiving elements is used.
03 is used, and the illumination light from the light source 104 illuminates the original 102.
It is reflected on the surface and is imaged onto the image sensor 103 by the lens 108 via the mirrors 105, 106, and 107. Light source 104. Mirror 105 and mirrors 106 and 107 are configured to move at a relative speed of 2:l. This optical unit consisting of the light source 104 and mirrors 105, 106, and 107 is reciprocated at a constant speed under PLL control by a DC servo motor 109. The moving speed is 90mm/90mm depending on the reading magnification on the outward path from left to right.
sec to 360 mm/sec, and is always 630 mm/sec on the return journey from right to left.

The direction in which this optical unit moves is called the sub-scanning direction.
While reading the main scanning line perpendicular to the sub-scanning direction with a resolution of 16 pefL/mm using the image sensor (main scanning), the optical unit is moved forward from the left end to the right end, and then moved back to the left end to perform one scan. By the zero or more operations that complete the process, the entire surface of the document 102 placed on the document table 101 is sequentially read every l line, and the image sensor 103 reads l lines.
An analog image signal indicating the density of the read image is repeatedly output for each line.

FIG. 2 shows a schematic block diagram of a circuit that processes analog image signals from the image sensor 103.

The image signal VD read by the image sensor 103 is converted by the A/D converter 201 into a 6-bit digital signal that digitally represents the image density of each pixel. A/D converter 2
The digital signal output of 01 is synchronized with the image data clock for reading out the image signal from the image sensor 1O93 via the latch 202, and is output to the latch 203 and the converter lator 204.
7, sent in 6-bit parallel to latches 205 and 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 l 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 SCL.

Similar to the comparator 204, 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, and determines whether the 6-bit image signal sent from the latch 202 If the new image signal is larger, a comparison output is sent to the AND gate 209. The AND gate 209 synchronizes the comparison output from the comparator 207 with the sampling clock SCL and sends it to the latch 208.

Latches 205, 208 and gates 206, 20 respectively
When the comparator output is received from 9, the 6-bit image signal sent from the latch 202 is latched, and the CPU 211
By sending 0 or more to latches 205 and 207, the minimum and maximum values of the image signals inputted up to that point are latched, respectively.

In addition to the comparator output and the sampling clock SCL, the AND gates 206 and 209 also include the image sensor 10.
An input signal EN indicating the valid section of each line of the image signal from No. 3 is input. Therefore, the comparison results for the image signals in a predetermined section for each main scanning line are sent from the latches 205 and 208 to the CPU 211.

A CPU 211 consisting of a microcomputer operates a latch 205.2 in synchronization with a synchronization signal MS for each main scanning line.
By taking in the image signal from 08, 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 can be detected.

Based on the white peak and black peak detected for each line, the CPU 211 determines a slice level using an algorithm described later according to a program stored in advance in the ROM 212, and sends the slice level to the comparator 210. The comparator 210 receives the 6-bit image signal from the latch 203 and the CPU 211.
A z-valued signal (image data VIDEO) indicating white/black of each pixel is generated by comparing the 6-bit slice levels from .

Note that the coordinate detection unit 215 detects the coordinates (position) on the document table lot of the document placed on the document table 101 of the document reading device.

Further, 214 is an operation display section provided on the top surface of the document reading device shown in the first diagram, and 213 is a memory RAM that temporarily stores calculation data of the CPU 211.

FIG. 3 shows a state in which a document is placed on the document table 101 of the document reading device (FIG. 1).

In this case, the coordinates of four points (Xr, Y
t), (X2.Y2), (X3.Y3), (X4, Y
4) is detected by scanning the optical system forward. The document cover 110 (FIG. 1) is mirror-finished so that image data outside the area where the document is placed is always black data. The pre-scanning includes main scanning and sub-scanning to cover the entire glass surface.

In FIG. 4, which shows the detailed circuit configuration of the coordinate detection unit 215, the main scanning counter 351 is a down counter, and represents the scanning position in one main scanning line. It is set to the maximum value in the direction (X direction) and counts down every time the image data clock CLK is input. The sub-scanning counter 352 is an up counter, which is reset to "0" at the rising edge of VSYNC (image leading edge signal), counts up at the H5YNC signal, and represents the scanning position in the sub-scanning direction.

During pre-scanning, the image data VIDEO binarized by the comparator 210 is input to the shift register 301 in 8-bit units. In addition, during pre-scanning, CPU2
11 supplies a predetermined fixed slice level to comparator 210. When the 8-bit input is completed, the gate circuit 302 shifts ◆Checks whether all the 8-bit data in the register 310 is a self-image (θ level), and if all is a white image, the signal line 303
Outputs 1 to .

F when the first 8-bit white appears after starting pre-scanning of the original.
/FC flip-flop) 304 performs -t='/). This F/F 304 is reset in advance by VSYNC (image leading edge signal output at the start of forward movement). After that, it is left set until the next VSYNC comes. The main scanning counter 351 uses a clock synchronized with each pixel output of image data from the comparator 210.
It is a down counter, and when the F/F 304 is set, the latch 305 & main scanning counter 351 at that time
The value of is loaded. This becomes the x1 coordinate value. Further, the sub-scanning counter 352 counts up the signal synchronized with the scanning of every l line, and the latch 306 is connected to the F/F 30.
The value (number of lines) of the sub-scanning counter 352 when 4 is set is loaded.

This becomes the Y1 coordinate value. Therefore Pl(Xt.

Yl) is found.

Also, each time 1 is output to the signal 303, the main scanning counter 35
Load the value from 1 into latch 307. When the value from the main scan counter 351 when the first 8-bit white appears is loaded into the latch 307, the data in the latch 310 (which is set to the maximum value in the X direction at the time of vSYNC) and the comparator 309 are loaded. are compared in size. 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 352 is loaded into the latch 311. This operation is processed until the next 8 bits enter the shift φ register 301. If data comparison between the latch 307 and the latch 310 is performed for the entire image area in this way, the minimum value of the original area in the X direction remains in the latch 310, and the coordinate in the Y direction at this time remains in the latch 311. That is, since the main scanning counter 351 is a down counter, the coordinate corresponding to the minimum value in the X direction represents the coordinate closest to SP in the main scanning direction, which is P2 (X2).

Y2) coordinates.

F/F 312 is an F/F that is set when 8-bit white appears for the first time in each main scanning line, and the horizontal synchronizing signal H5Y is set.
It is reset by NC and the first 8 bits are set to white and held until the next H3YNC. At the time when the F/F 312 is set, the value of the main scanning counter 351 corresponding to the position of the first white signal appearing in one line is set in the latch 313. and the value of latch 315 and comparator 316
are compared in size. The latch 315 is preset to the minimum value in the X direction, ie, O at the time VSYNC occurs. If the data in latch 315 is less than or equal to the data in latch 313, signal 317 becomes active and the data in latch 313 is loaded into latch 315.

This operation is performed by latch 31 when the comparison operation of 0 or more between H3YNC and H3YNC is performed for the entire image area.
5 remains the maximum value of the document coordinates in the X direction, that is, the X coordinate of the white signal from the point farthest from the scanning start point in the main scanning direction. This is x3. Also, the signal line 31
7 outputs, the value from sub-scan is loaded into latch 318.

This is Y 3 Ninari, and P3 (X3, Y3) is obtained.

Latches 319 and 320 are loaded with the current main scanning counter value and sub-scanning counter value each time 8-bit white appears in the entire image area. 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 (X4.
Y4).

The above 8 latches (306,311°320.318
, 305, 310.31'5゜319) are connected to the line BUS of the CPU 211.
U211 reads this data at the end of the forward movement in the previous scan.

FIG. 5 shows a flowchart of the document reading resequence, and steps 501 to 507 relate to pre-scanning. First, in step 501, the optical unit performs forward scanning from the left end to the right end in FIG.
The illustrated coordinate detection unit 215 detects the coordinates of the document on the document table 101, and captures each detected coordinate data.

Next, in step 502, the area in which the peak value for determining the slice level for binarization should be sampled is calculated from the coordinate data detected in step 501. From the coordinates detected for the original, the peak value sampling error 1 of this original is calculated.
As Noah, IQ1* at Y3, y2 and Xt, X4
Select the rectangular area that will be displayed. This is because the original is usually placed as parallel as possible to the original table, and even if it is placed at a slight angle as shown in Figure 3, unnecessary information from outside the original will not be mistaken as data from the original. This is because there is no danger of it being stolen. Of course, the sampling area may be determined using any other method.

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.

At this point, the peak value sampling start point Y2 and end point Y3 are known, so the 51st! i! Step 5 of I
04, 505, and 506 can be scheduled to be executed.

That is, when the return movement is started in step 503, C
After counting the main scanning line synchronization signals for the distance (Ymax-Y2), the PU211 starts detecting the self-peak value/black peak value described above in FIG. 2, and then calculates the distance (Y2-Y3) from that point. After counting the main scanning line synchronization signals for a corresponding number of times, peak value detection is finished.

Further, after counting the main scanning line synchronization signals for the distance 111Y3, the backward movement is stopped.

Furthermore, when starting peak value detection in step 504,
Needless to say, the above-mentioned energy pull signal EN is set to be output in correspondence with the detection coordinates Xi and X4 as shown in FIG.

With the above operation, it is possible to reliably detect the self-peak and black peak of the image density for each main scanning line in the document placed at any position on the document table 101, and the detection operation can be performed using something other than the document, such as the document cover. to prevent it from being influenced.

Next, an algorithm for determining slice levels for binarization will be explained.

According to the procedure described above, the CPU 211 takes in the black peak value and the self-peak value for each main scanning line from within the document area.
Now, if the black peak on the i-th main scanning line is BPi and the own peak is WPi, since the image data is a 6-bit value, each takes a value from "" (HEX) to 3F (HEX). , and BPI≧WP1.

The CPU 211 uses a 64 x 2 byte black peak histogram area prepared in the RAM 213 and a 2 byte area H within the 64 x 2 byte self peak histogram area.
B (J), data B detected from the contents of HW (j)
The areas corresponding to Pl and WPi are counted up one by one. Then, after waiting for the primary main scanning line synchronization signal MS, the data BPi41 and WPi from the f+1th line are
+1 is taken in, the corresponding area of the histogram is counted up again, and the process continues until the end of the sample in step 505.

However, at this time, the detected BPi and WPI are not necessarily used as histogram data. For example,
If there is a uniform density band in the main scanning line direction, whether it is pure white, pure black, or other density, the sample value BPI from that band and W'DI + - LrZ u 217 + 1
．． %+11. MLfThLl m1nx It is not suitable as information for distinguishing and binarizing the information part, and even in the background part, the density is not necessarily uniform, so the black peak of the variation in the background level is Appropriate binarization cannot be performed even if it is treated as an information level. Therefore, C
When PU211 determines that BPi-WPi≦α, that is, the difference between the black peak value and the self-peak value is smaller than a predetermined value and there is little change in density, the BPi and WPi are not adopted as data for creating histograms. , this α is a constant set empirically.

On the other hand, we can also consider cases where a value that is whiter than the actual background level due to dust on the platen glass is sampled as the self-peak, or a case where a value that is darker than the actual information level due to dirt, etc. is sampled as the black peak. It will be done. In these cases.

is not suitable as data for creating a histogram. Therefore, when BPi≧β or WPi≦γ, the CPU 211
That is, the black peak value or the self-peak value deviated from the respective reference values. If the values are unique, the BPi and WPi are not used as data for creating a histogram.
and γ are also constants set empirically.

Also, before starting the sample in step 504, R
It is a matter of course to clear all 64x2x2 bytes of the histogram area of AM213 to \.

FIG. 7 shows a detailed flowchart of peak value sampling in steps 5504, 505 and 506 in FIG. 5, and will be described below.

The CPU 211 first uses the area HW for creating a histogram.
(j) and HB (j) are all cleared to \ (Si2-
1), this HW (j) and HB (J) are shown in Figure 2 RAM
In the area above 213, HW (J), HB for each j
(j) are both composed of 2 bytes, and j is the possible value of the detected peak value from 0 (00uEx) to (J),
j=0, +++++, 63, totaling 2×64=128 bytes, and HB(j) also has a total of 128 bytes.

HW (j) indicates the frequency of the own peak value WPi (=j),
HB (j) indicates the frequency of black beak BPi (=j),
If the histogram creation area is 2 bytes each, 6553
It can count up to 5, and the resolution of 16peJL is sufficient since the total number of scanning lines for A3 is 6720 lines.

Next, the CPU 211 initializes the main scanning line number I to 5.
14-2), when the end of the peak value detection of the first line is detected by the main scanning line synchronization signal MS (S l 4-3
), the self-peak value WPI of the first line from the aforementioned latch
, take in the black peak value BPI (S14-4), if.

When BPI-WPI≦α, that is, when the difference between each peak is smaller than the predetermined value α, the BPI is determined as described above.

Do not use WPl as histogram data (S14-
5).

On the other hand, if BP1-WPt>α, then it is determined whether BPI≧β, that is, the black peak value is larger than the predetermined value β,
When it is large, it is not used as histogram creation data (S14-6).

Also, if BPI<β, then WPI≦γ.

That is, it is determined whether the self-peak value is smaller than the predetermined value γ, and if it is smaller, it is not used as the histogram creation data (314-7).

If wpt>γ, WPI and BPI are used as histogram creation data, and the frequency area HW (WPI) on the RAM corresponding to the white peak value WP1 is counted up by 1 (514-8), and the black peak value BPI is 1 count up for the similar HB (BPl) corresponding to (5
14-9).

After performing peak value sampling and histogram generation or data rejection for the first line as described above, the value of i is increased to 2゜3, 4, -------1514-10), ．． Third. 4th -- 1 --- The same process is performed for each line until the backward movement is stopped, and as a result, when the sample is completed in step 505 (FIG. 5), for example, the process shown in FIG. 8 (1) is performed.

A histogram as shown in (2) is constructed for each of the black peak BP/white peak WP.

After completing the sample, the optical system returns to the starting point and completes the return motion in step 506.

Next, in step 507, a slice level is set.

The density level showing the peak of the set frequency of the slice level of each histogram is considered to be each representative value.

According to the example of FIG. 8, the density of the document information part is 36H, the density of the background part of the document is OAH, and, for example, the median value of 20H is the slice level. Note that there are other ways to determine the slice level, and the operation for determining the slice level will be explained in detail later.

Finally, in step 508, the document reading scan is performed 4:,b-
The CPU 211 uses the comparator 2 to determine the slice level determined in step 507.
Of course, it should be supplied to 10.

A case where a value blacker than the actual information level is sampled as the black peak BP will be explained using FIG. 9.

When a peak value is sampled from an image OG■ in which there is dirt GD on the platen glass, etc., the dirty portion GD is extremely black, resulting in a black peak value BP, and the own peak value WP as in ■,
Black peak BP is sampled. If this is used as histogram generation data as it is, the maximum value of the black peak histogram will be higher in the dirty part than in the information part, as shown in @, so the slice level will be determined to be higher than the density of the information part, and appropriate binarization will not be possible. Not executed.

In this case, if we adopt the value β mentioned above and discard the black peak BP in the dirty area as the histogram generation data, the black peak histogram in that case will have the maximum value at the density value of the information area, as shown in ■, and will be appropriate. The desired slice level can be determined.

Similarly, a case where a value whiter than the actual background level is sampled as the own peak WP will be explained using FIG. 9.

When reading a document (2) with a dark background and dust DS on the document platen glass, the dusty portion DS as shown in (2) is extremely white and is sampled as the self-peak value WP. If this data is used as is to generate a histogram, the slice level will be determined to be whiter than the actual background, as shown in ■, and appropriate binarization will not be performed.

In such a case, if the above-mentioned value γ is not adopted and the white peak value WP of the dusty area is not adopted as the histogram generation data, a mountain of the self-peak histogram will be generated in the actual background area as shown in ■. An appropriate slicing level is determined.

Therefore, by combining the correction operation of the black peak value BP and the own peak value WP using the values of β and γ as described above, it is possible to determine a good slice level that is not affected by noise or the like. Alternatively, a parameter may be set for either the self peak value or the black peak value, and if either peak value is greater than or less than the parameter, that value may be invalidated as data for creating histograms.

FIG. 10 shows an operation display section 214 provided on the top surface of the document reading device for density selection.

Reference numeral 802 denotes a density display section made up of seven LEDs that can indicate seven levels, and shows a state in which density 4 is selected. Each time the key 800 is pressed, the display stage 802 moves one by one to the left, and each time the key 801 is pressed, the display stage 802 moves one to the right.

Reference numeral 803 is a key for selecting the contrast of the original, and each time this key is pressed, the display of three LEDs 804 changes stepwise from normal to high contrast to low contrast to normal.

The information is taken into the CPU 211, and display units such as I and ED perform display operations according to commands from the CPU 211.

Concentration display stage of LED802! (= 1, ---
．． 7), the density level of the background area of the document estimated from the white peak histogram is WpP, and the density level of the document information area estimated from the black peak histogram is BPP.
The slice level sL for binarization can also be determined as follows. However, [] is a Gaussian symbol. For example, in the case of FIG. 8 mentioned above, wpp=φAH9B
Since PP=36H, the relationship between density display and slice level is as shown in the following table.

FIG. 11(A) is a graph showing the relationship between the histogram shown in this table and the slide and slope levels.

In this way, by changing the internal division ratio of the slice level, which is created by internally dividing the area between the background level and the information level estimated by peak value detection, using the density display f, it is possible to completely remove the background while also making it easier for the operator to You can make adjustments such as making the intermediate density part darker or lighter depending on your preference.

For example, in a case where the black peak histogram has a second maximum value, such as point BP2, as shown in FIG. Therefore, the operator selects 15 or more using keys 800 and 801 and selects BF.
It is possible to select whether to set f2 as the background level or select f4 or lower as the information level.

Furthermore, the above method also has the advantage of increasing the resolution of the concentration display f.

For example, as shown in FIG. 11(C), if the slice level corresponding to the density display f is fixed regardless of the black peak or white peak, then the black, For an image showing a self-answering peak histogram, the information read by fl and fl becomes pure white or pure black and cannot be said to be effective.

Furthermore, f6 in this case is located in the mountain of the black peak, and part of the information part is cut off, and furthermore, f2 is located in the mountain of its own peak, and the background is covered, so it is actually from f3. Only three stages of f5 can be expressed.

However, as shown in FIG. 11(A), it corresponds to the density display f between the estimated background level and black level.

If the slice level is determined by the following, expression in seven stages is possible and is effective.

The above is a case of one slice level, that is, binary image output, but in the case of two slice levels, that is, ternary image output, two types of slice levels peak as shown in Figure 11 (B). It is also possible to set the density display f between the background and information level estimated by detection.

FIG. 12 shows a flowchart for calculating the slice level, and will be described below.

After completing the peak value sampling as described above, HW (W p p ) = m a x HW (j
), that is, the density level WPP showing the highest frequency in the white peak histogram, and HB (B p p) = maxHB (J) l1l
pp, that is, the density level Bpp showing the highest frequency in the black peak histogram (315-1, 3
15-2).

Thereafter, the slice level is determined in a slice level determining step (515-10).

m= 7-17d s4w-p=----1fcI K
There are 4 types of algorithms A, B, C, D as l^1.
One of these is executed. A desired slice level can be obtained by selecting one of these four algorithms using a mode key (not shown) provided on the operation display section 214. Note that it is not necessary to provide all four algorithms, but it is sufficient to provide at least one.

Algorithm A is an example in which the average value of WPP and Bpp mentioned above is used as the slice level sL for a binary image (S
5-3), Algorithm B calculates WPP and Bpp according to the concentration display step value f described above using FIG. 11(A).
This is an example of determining the slice level sL by changing the internal division ratio (315-4.5).

Algorithm C is an example of determining two slice levels SLA and SLB for ternary image reading.
B internally divides WpP and l1lpp at a ratio of 1:1:1 (
S 15-6), the example in Figure 8 above (Wpp=OA)l
,Bpp=,36)1) then SL A=27H,
SL 5=19H, algorithm D changes the internal division ratio of Bpp and WPP according to the concentration display step value f, and calculates 3.
This is an example of determining two slice levels SLA and SL8 for a value image, and the calculation formula is as follows: When U≧4] When U≧4: When U≧4: When U≧4: According to the formula, the relationship between slice level SL and f is as shown in the table below.

A graph of this is shown in FIG. 11(B).

Value α that determines whether or not each peak value is adopted as data for creating a histogram when the concentration change described above is small.
The contrast key 803 and display 80 shown in FIG.
It is also possible to perform a more faithful binary value by changing the value of 4 as shown in FIG. This will be explained below using FIGS. 13 and 14.

For example, in an original with relatively low density (such as pencil writing) or a completely overlapping original (such as a blueprint), the difference between the white peak and the black peak over the entire surface of the original is as follows: ■ or ■<di or d2
If the above-mentioned value α is small, for example, only α21 in Figure 10 (α2>dx, d2), all of the detected peak values will not be adopted as the data for creating the histogram, and the histogram will not be created at all as shown in ■. There is a possibility that it will not be formed. Therefore, in such a case, by selecting the low-cut trust in the display 804 with the key 803 and selecting the smallest value α1 as α, a histogram like @ is generated and an appropriate slice level can be determined.

On the other hand, when there is unevenness in a part of the background part of the document as shown in ■, and the area occupied by the information section on the document is small, if the above value α is, for example, α2 in Fig. 10, it is sliced as shown in ■. In such a case, when the level is low and the background is covered, you can use the key 803 to select high contrast in the display 804 and select the largest value α3 as α.
As a result, a good slice level is selected as shown in (2), and it becomes possible to remove the fog.

Further, the above-mentioned β and γ can also be changed depending on the concentration display stage f. For example, as shown in FIG.
When γ is set, the closer to f7 the histogram peak moves toward black (to the right in the example in Figure 7), and the slice level becomes higher; (in the example in Figure 7, to the left), and the slice level becomes lower.

Therefore, by setting f according to the condition of the document by the operator, it is possible to eliminate noise and to eliminate the inconvenience such as the image disappearing in the heat.

FIG. 16 shows the flow of peak value sampling when the constants α, β, and γ described above are linked to the operator's operation using the operating section of FIG. 10, and will be described below.

Note that since the operations (S16-1) to (51B-11) are similar to the operations (S14-1) to (S14-11) explained in FIG. 7, detailed explanations will be omitted.

In the case where the constant values of α, β, and γ are linked to the operation unit, the contrast mode selected by the operator is imported before the start of scanning (S t S-12), and according to that mode, α is adjusted as shown in Fig. 13. α1. α2. α3
(516-13).

Next, the concentration display value f selected by the operator is imported (
516-14), and according to that f, β as shown in FIG.
, γ are set (S16-15).

As described above, in Fig. 7, α, which was a fixed value,
By making β and γ variable through operator operations, the operator can perform desired binarization.

As described above, in this embodiment, a histogram of the black peak and the self-peak of the original to be read is created, the density levels of the background and information of the original are estimated based on this, and these levels are used to determine the quantum density of the image. This determines the slice level for conversion. Therefore, a slice level suitable for the document to be read can be automatically set, and good image quantization can be achieved.

Furthermore, when creating a histogram for setting the slice level, peak values that are undesirable due to dust, dirt, etc. are invalidated, and their adverse effects on the setting of the slice level are eliminated. Furthermore, since white peaks and black peaks from areas where density level changes are small are also invalidated, peak values from areas where both background and information exist are used to form a histogram, allowing a good slice level to be set.

Furthermore, since the operator can manually set various parameters that serve as criteria for invalidating peak values, in addition to the automatic slice level setting described above, it is also possible to set any slice level, which is suitable for the document image condition. The slice level can be set arbitrarily. Note that this manual setting function makes it possible to read and satisfactorily quantize documents with image conditions that would be inconvenient with automatic slice level setting.

In this embodiment, the original image is line-scanned, and the black peak and white peak are detected every 1 scans. However, for example, each peak value may be detected every few lines to create a histogram. If the current size or position is known in advance, the current position detection may be omitted and the document image peak detection operation may be performed immediately.Furthermore, if the same document image is read multiple times in succession, The slice level determined based on the peak value obtained in the previous scan is used for all the multiple reading operations, thereby eliminating the need to determine the slice level for each reading.

In addition, in this embodiment, the original placed on a fixed original table was read by moving an optical unit consisting of a mirror, etc., but conversely, a reading device with a structure in which the optical unit is fixed and the original or the original table is moved can be used. can also be applied.

In addition, an operation mode is provided that cancels the automatic slice level determination operation based on peak value detection and enables image reading using a slice level manually set from the operation panel or predetermined. It is also possible to use specific data other than the peak value as data for setting the slice level. Furthermore, in addition to simply determining the slice level for binary or ternary conversion, detection data from pre-scanning is used to select or form a threshold matrix suitable for halftone reproduction, for example, using a dither method, depending on the document image condition. You may also use the original image
A sensor separate from the sensor for reading the document may be provided to detect the state of the image, but in this case, problems such as high cost and complicated configuration arise. Furthermore, in cases where the original background density level and the image information density level are known in advance, the detection of both the white and black peaks may not be performed, and only the peak value of one of them may be detected. In this embodiment, an example is used in which an image is read using reflected light from a sheet or book-shaped document, but the present invention is similarly applicable to a configuration in which an image is read using transmitted light from, for example, a microfilm.

〔effect〕

As described above, according to the present invention, the image density level of a document is detected, and based on the detected image density level, an image signal obtained by reading the document is processed, and a predetermined value of the detected image density level is processed. Since those above or below the level are invalidated, it is possible to effectively remove the adverse effects of dirt, dust, etc. other than the original image, and to perform optimal image processing.

[Brief explanation of the drawing]

FIG. 1 is a schematic diagram of a document reading device to which the present invention is applied;
3 is a block diagram of the image signal processing circuit, FIG. 3 is a diagram showing the coordinates of the document on the document table, FIG. 4 is a block diagram of the coordinate detection circuit, FIG. 5 is a flowchart showing the document reading resequence, and FIG. FIG. 6 is a diagram showing the correspondence between the document placement position and the document reading resequence, FIG. 7 is a flow chart diagram showing the peak value sampling procedure, and FIG. 8 (1). (2) is a diagram showing an example of a black peak histogram and a white peak histogram, respectively; FIGS. 9 and 14 are diagrams showing slice level determination operations for various manuscripts; and FIG.
External view of the operation display section, Fig. 11 (A), (B), (C)
12 is a diagram showing the relationship between the peak value and the slice level, and FIG. 12 is a flowchart diagram showing the procedure for calculating the slice level.
FIG. 13 is a diagram showing the setting operation of the parameter α, FIG. 15 is a diagram showing the setting operation of the parameters α and γ, and FIG. 16 is a diagram showing another procedure of peak value sampling.
103 is an image sensor, 202, 203, 205
．． 207 is a latch, 204 and 207 are comparators, 2
11 is a CPU, 215 is a coordinate detection circuit, and 214 is an operation display section.

Claims (3)

[Claims] (1) comprising means for reading an original image, means for detecting an image density level of the original image, and means for processing an image signal from the reading means based on the image density level detected by the detecting means. An image processing apparatus characterized in that among the image density levels detected by the detection means, those above or below a predetermined level are invalidated. (2) The image processing apparatus according to claim (1), wherein the predetermined level, which is a reference for invalidating the image density level, is variable. (3) The image processing device according to claim (1), wherein the processing means quantizes the image signal from the reading means.