JPH1146301A - Image distortion correction device and image-processing device and method using the correction device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To rapidly correct shading with high accuracy despite the variance of peak values of a reading signal by multiplying the input data by the peak value of data on a reference image and also multiplying the value of a signal peak register, i.e., the peak value covering up to the preceding line of the input data by the peak value of data on the reference image to perform the shading correction, based on those two multiplication results. SOLUTION: A 1st mode is a shading storage mode. In a signal reading mode, i.e., a 2nd mode, a shading waveform is read out synchronously with the reading of an image. Then an input signal VI is multiplied by a reference peal signal Vwp from a reference peak detection part 101 with a multiplier 106. Meanwhile, the shading waveform Vw stored in a shading storage part 102 is multiplied by the peak signal from a signal peak detection part 104 with a multiplier 107. The signal Vic multiplied by the multiplier 106 and the signal Vwc acquired through the multiplier 107 are inputted to a shading correction circuit 105 which performs the shading correction.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for processing an image read by a scanner or the like, and particularly to an apparatus for correcting image distortion in an image processing apparatus requiring high-speed and high-precision correction. And methods.

[0002]

2. Description of the Related Art An image reading apparatus generally reads image information with a scanner, and a shading correction method is known as a method for correcting unevenness of a light source and sensor sensitivity which are characteristics of an optical system of the scanner. In the shading correction method, first, a shading waveform serving as a reference (a read waveform of a white reference plate) is read in advance, and when an image is read next, an image signal read using the previously read shading waveform is normalized to reduce shading distortion. This is a method for correction. However, it is often assumed that a signal higher than the shading waveform is not input. In this case, when a signal higher than the shading waveform enters, the signal is clipped to white. On the other hand, in the case of reading a facsimile or the like, since the shading waveform is read and updated for each communication or each page, there is no time difference between the storage of the shading waveform and the time of image reading, and a signal higher than the shading waveform enters and is clipped to white. However, that part does not matter because it is not really different from white.

[0003] However, some reading devices are constantly operating and can only update the shading waveform weekly or monthly.

[0004] In this case, there is a time difference between the time of reading the shading waveform and the time of reading the image. Therefore, there is a possibility that the shading waveform may be changed due to a change in environmental temperature, and good correction cannot be performed. Although the accuracy is somewhat deteriorated, if the input signal is lower than the shading waveform, a peak value after the shading waveform is detected and normalized by the peak value, thereby preventing the deterioration of the image quality such that the image is completely blackened. However, when a signal higher than the shading waveform enters, a necessary signal is clipped in white at the time of shading correction, so that image quality deterioration cannot be prevented even in the subsequent processing. In this case, the image becomes white as a whole, and, for example, there arises a problem that the image of the black thin line rubs white.

As a method for solving such a problem, for example, as described in Japanese Patent Application Laid-Open No. 60-259063, an attenuator is provided in front of an amplifier for amplifying a sensor output, and an attenuator is detected by detecting overflow of shading correction. , And adjust the signal within the shading waveform.

There is also known a method in which a shading waveform value stored in advance is multiplied by a certain ratio to make it higher than an actual shading waveform, and a margin is provided so that a signal does not exceed the shading waveform.

[0007]

SUMMARY OF THE INVENTION However, Japanese Patent Application Laid-Open
According to the method described in Japanese Patent Application No. 60-259063, the processing when an overflow occurs is not closed by the digital section, and extends to the analog section.

Further, all the methods of multiplying a previously stored shading waveform value by a certain ratio can be realized by digital processing, but there is a problem that high-precision correction cannot be performed.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide an image distortion correction apparatus which can realize high-precision shading correction at a low cost even if the peak value of a read signal fluctuates. It is the purpose.

[0010]

In order to achieve the above object, the present invention provides a means for storing shading data of a reference image, a means for detecting a peak value of shading data, and a method for detecting a peak value of read input data. Means for multiplying the input data by the peak value of the data of the stored reference image, and the value of the signal peak register, which is the peak value up to the previous line of the input data, and the data of the stored reference image. Means for multiplying by a peak value, 2
And a shading correction unit for inputting two multiplication results and performing shading correction.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows an image processing apparatus to which the present invention is applied. The image processing apparatus includes a light source, an image sensor 1 for converting read light into an electric signal, an amplifier 2, an AD converter 3, an image reading apparatus 10, a recognition processing unit 6, a printer 7 as an output apparatus, and control of the entire apparatus. The CPU 8 controls the motor (not shown) for transporting the read medium, and executes LINE END after reading one line.
It comprises a mechanical control section 9 for outputting a signal.

The operation of this image processing apparatus is as follows. Light is applied to a reading medium from a light source, the reflected light is photoelectrically converted by an image sensor 1, an analog image signal is amplified by an amplifier 2, and A
The data is converted into 10-bit digital data by the D converter 3,
The image distortion correction unit 100 removes image distortion and performs a correction process. The distortion-removed signal Vc is subjected to a filtering process in the image preprocessing unit 4 to enhance or smooth the signal. The binarization processing unit 5 has a function of converting the input multi-level signal into a binary value and a function of outputting the input multi-level signal as it is, based on the binarized signal or the multi-level signal. First, the recognition processing unit 6 performs image recognition of characters and the like. Further, the printer 7 outputs a binary image or a multivalued image.

Next, the image distortion correcting section 100 of the present invention will be described.

The image distortion correction unit 100 includes a register 10
8, reference peak detection unit 101, shading storage unit 1
02, signal peak detection unit 104, multipliers 106 and 10
7. It is composed of a shading correction circuit 105.

The image distortion correction unit 10 thus configured
0 is executed in two modes based on a command from the CPU 8. The first mode is a shading storage mode, in which known image data such as white and black is read as reference data and stored in the shading storage unit 102. At this time, the white image data is read by, for example, reading a blank sheet, and the black image data is read with the light source turned off. At this time, the peak values of the white and black reference data are stored.

The second mode is a signal reading mode. In this mode, the shading waveform is read in synchronization with the reading of the image, and the input signal Vi and the reference peak signal Vwp from the reference peak detection unit 101 are multiplied by the multiplier 106.
And the multiplier 107 multiplies the shading waveform Vw stored in the shading storage unit 102 by the peak signal from the signal peak detection unit 104. Next, the signal Vic multiplied by the multiplier 106 and the signal Vwc obtained by the multiplier 107 are input to the shading correction circuit 105, and the shading correction circuit 105 performs correction according to the following equation.

[0018]

Vc = Vic / Vwc * 1024 = (Vi * Vwp) / (Vw * Vip) * 1024 (Equation 1) As a result, the input signal can be suppressed between the white shading waveform and the black shading waveform. Therefore, white clipping can be prevented, the entire image can be prevented from rising to the white side, and advanced shading correction can be performed.

Here, the fact that good shading correction can be performed by the image distortion correction unit 100 will be described in detail.

FIG. 2 shows a conceptual diagram of shading correction. In FIG. 2A, the horizontal axis indicates the pixel arrangement direction, and the vertical axis indicates the luminance. Here, the image distortion correction unit 100
Input is a 10-bit input, so 0 to 10
23 indicates a luminance value. Black shading is a value obtained by reading a black reference level, and white shading is a value obtained by reading a white reference level. The reason why the brightness at the center is high and the brightness at the periphery is low is due to the way of illumination and the characteristics of the lens. The portion that looks like noise is due to the effect of the level difference that occurs because the characteristics of the amplifier that amplifies the output from the sensor are different between the odd and even bits of the sensor. In such a state, the input signal normally exists between white shading and black shading as shown in FIG. 2B. However, the read signal may exceed the shading waveform as shown in FIG. 2C due to the fluctuation of the light amount due to the temperature or the ground density of the read document.

In this case, if the input signal is Vi, the shading waveform is Vw, and the shading correction result is Vc, and the correction shown in Expression 2 is performed, if the input signal Vi is larger than the shading Vw, Vc> 1. I will.

[0022]

Vc = Vi / Vw (Equation 2) In this case, when the shading correction is performed, the signal above the white shading waveform is clipped in white, and the entire image also rises to the white side. It becomes a thin image.

Therefore, even when the peak value of the input signal fluctuates as shown in FIG. 2C, the input signal is suppressed between the white shading waveform and the black shading as shown in FIG. 2D. Thus, high-precision shading correction that prevents white clipping and prevents the entire image from rising to the white side can be realized.

Therefore, the brightness is 0 to 1023, that is, 10
Since this value is normalized using the peak value Vwp of shading and the peak value Vip of the signal,

[0025]

Vc = (Vi / Vw) * (Vwp /) * 1024 (Equation 3) When (Equation 3) is transformed, (Equation 1) is obtained.

As described above, the image distortion correcting section 10 of the present invention
With 0, good shading correction can be performed. In addition, since the image distortion correction unit 100 is configured to calculate the peak value Vip of the signal and the peak value Vwp of the shading symmetrically and immediately divide the same, the normalization process using the peak value after the shading correction is performed. compared to,
A well-balanced and highly accurate calculation can be realized.

Next, each unit constituting the image distortion correction unit 100 will be described in detail.

FIG. 7 shows the configuration of the reference peak detector 101 for detecting the peak value of the shading waveform. The reference peak detection unit 101 includes a comparator 1011 and a register 101
2, a reference peak register 1013. When storing the shading waveform, the maximum value of the data of the shading waveform input by the comparator 1011 and the register 1012 is detected, and the peak value between one line is detected by a LINEN signal (a signal indicating completion of reading of one line). Is stored in the reference peak register.

FIG. 8 shows the configuration of the signal peak detection unit 104. The CPU 8 sets the initial mode in the register 1048 in advance and switches the selector 1045 to write the data (peak initial value) from the CPU 8 into the register 1046. Next, at the time of reading, the reading mode is set in the register 1048 from the CPU 8, and the output of the comparator 1044 is selected as the selector 1045. The read signal (Vi) is compared with the data of the register 1046 by the comparator 1043, and the comparator 1044 compares the value of the higher signal compared by the comparator 1043 with the reference peak register value (Vwp). Compare the value of the higher signal to the register 1046.
Set to. Then, the peak value is detected by comparing the next read signal (Vi) with the data in the register 1046. When the input of the read signal (Vi) for one line is completed, the LINEEND signal is input and the peak value for one line is set in the signal peak register 1047. Now, when this LINEEND signal is input, the selector 1
042 is switched from the register 1046 to the register 10
The value obtained by subtracting the value of 40 is input to the register 1046. Since the LINEND signal is one clock, a value obtained by subtracting only once is set in the register 1046.

Next, details of the shading correction circuit 105 will be described. Before describing the configuration of the shading correction circuit 105, the concept of shading correction will be described.

FIG. 3A is a conceptual diagram of the shading division process. The GI axis in which G is 0 in the figure indicates the level of the read signal, and the AG axis in which G is 0 indicates the level of the shading waveform. The height direction indicates the output level after correction. In other words, it is a conversion table for obtaining the result after shading correction by inputting the shading waveform level and the read signal level of the target pixel to be corrected. FIG. 3B shows this two-dimensionally. If the shading waveform has a level of, for example, 1023, it has a linear relationship with the input signal. If the shading waveform has a level of 512, the slope is doubled, and the output reaches 1023 when the input signal is at the 511 level. As described above, the values of the correction table are obtained by tabulating the calculation results to be corrected.

However, a table having two series of 1024 level (10 bit) inputs requires a memory of 1M words, so that an increase in cost cannot be avoided. Therefore, folding is performed to reduce the number of tables. First, since this table stores the result of division of two inputs, the result does not change even when each input is divided by the same constant. Therefore, by making each input １ ／ (9 bits), the correction space AGIC shown in FIG.
Map to HE. Although the bit precision drops,
As a result, the memory can be reduced to 1/4.

This operation will be described with reference to FIG. FIG. 4 is a diagram conceptually showing a relationship between a shading waveform and an input signal. The area a in FIG. 4A has a small white and black shading range and a small signal. For example, if the value of white shading is less than 511 (9 bits), it corresponds to the subspace DGHE in FIG. 3B. The area b in FIG. 4 has a large shading range and a large signal fluctuation (0 to 10).
23) This is the case. This is a portion corresponding to the space AGIC in FIG. In order to dynamically switch both the area a and the area b in FIG. 4 to a signal of 9 bits (0 to 511), as shown in FIG.
If the most significant bit of the higher value of the two inputs is 1, then
If the least significant bit of each of the two inputs is deleted and the most significant bit is 0, the most significant bit is deleted to realize 9 bits (0 to 511). If the least significant bit of the signal is simply reduced to 1/2, the original 9
If there is only a bit signal, it will be 8 bits,
By selecting the bit to be deleted by the most significant bit of the shading waveform (the higher value of the two inputs), it is possible to maintain the calculation accuracy at 9 bits whether the signal is small or large.

Returning to FIG. 3B again, further reduction of the subspace DGHE will be considered. Since the shading waveform W is always higher than the input signal S, if the signal S is higher than the shading waveform W,
The value is 1 (upper limiter). Therefore, the subspace DG
Among the HEs, meaningful division data is written in the triangle DGE, and the result of the triangle EGH is 511 (upper limiter) without referring to the table. Therefore, the table can be compressed by monitoring the address of the table. In other words, the triangle EJL describing the division data is converted to the area LG where the result is obvious.
Consider folding into N. This is possible with simple address translation. This allows the table to be reduced to half.

The shading correction circuit 105 constructed based on the concept described above will be described.

FIG. 6 shows the configuration of the shading correction circuit 105. First, the two inputs are compared by the range determination unit 1050, and it is determined whether the highest value is “1” or “0”. If the value is “1”, the range adjustment unit 1051 , 1052, the address is shifted. In addition, the return detection unit 1057 determines whether or not access to the triangle DGE is performed.
If this is the case, the address change units 1053 and 1054 perform the address return processing. Using these inputs, the division R
The result is obtained by accessing the OM 1055. In the mask processing 1056, if the signal exceeds the shading waveform, the signal is treated as an overflow and the data indexed in the table is masked and rejected.

FIG. 9 shows the configuration of the range adjustment unit 1051. The input signal (for example, 10 bits) is divided into two paths, and is simply shifted by し て to remove the lower one bit (9 bits), and the other path is shifted twice to remove the upper bit, Later, the input signal is halved and passes through the path of the signal in which the upper bit is simply deleted (9 bits).
Then, the selector 10513 selects one of the signals by the switching signal and outputs the selected signal.

FIG. 10 shows the structure of the overflow detecting unit 1058. Overflow detection unit 105
8 compares the multiplication value of the shading waveform output from the shading storage unit 102 with the signal peak value of the signal peak detection unit 104, and the multiplication value of the input signal Vi and the output of the reference peak detection unit 101 which is the shading waveform peak value.
When the multiplication value of the input signal Vi and the output of the reference peak detection unit 101, which is the peak value of the shading waveform, is high, the overflow detection unit 1058 outputs the overflow signal “1” and passes it to the mask processing unit 1056.

FIG. 11 shows the structure of the mask processing unit 1056. When the overflow signal is “1”, the mask processing unit 1056 outputs a constant value of “511” (9 bits: decimal value) instead of the output of the division ROM 1055.
61 is selected by the selector 10562.

FIG. 12 shows the configuration of the turn-back area detection unit 1057. Return area detection unit 1057
Is the product Vic of the shading waveform and the signal peak value.
And the most significant bit of the multiplication value Vwc of the input signal Vi and the peak value of the shading waveform by the AND circuit 105
71.

FIG. 13 shows the structure of the address changing unit 1054. The range-adjusted signal passes through the path as it is and the path through which the range-adjusted signal from the constant "511" (decimal value) 10541 is subtracted by the subtractor 10542, and one of them is switched by the selector 10543 by the signal of the return area detection. When the signal of the return area detection is "1", since the area is the LMEj area in FIG. 3B, it is a return area, and therefore a constant "511" (decimal value) 1054
A signal to be subtracted from the signal whose range has been adjusted from 1 is selected as a subtractor 10542. Otherwise, the range-adjusted signal is output as it is.

FIG. 14 shows the configuration of the range determination unit 1050. A shading waveform output from the shading storage unit 102 as an input signal and a signal peak detection unit
A comparator 10501 compares the multiplied value of the signal peak value of 104 with the multiplied value of the input signal Vi and the output of the reference peak detecting unit 101, which is the shading waveform peak value, and selects the larger one of the most significant bits. At 10502, a selection is made and output.

FIG. 15 shows the configuration of the shading storage unit 102. The address counter 1021 outputs E at the timing of latching data from the sensor.
An address of the memory 1023 is generated by inputting N and LINEEND which is a line synchronization signal. The address counter 1021 can generate an address of the memory 1023 by inputting EN to a clock and setting LINEN to reset. At the time of image correction (normal reading), the selector 1024 is switched from the CPU 8 for reading only and reading is fixed at "L" (Low level). When storing the shading waveform, writing is performed at the timing of the EN signal. At the time of writing, the input data Vi is stored in the memory 1023 as it is.

FIG. 16 shows the structure of the image preprocessing section 4. This configuration is a general smoothing filter. The output Vc of the shading correction circuit is stored in the memory for two lines, and a convolution operation is performed in a 3 × 3 pixel area. Here, 1/9 is added to the entire 3 × 3 pixel area.
And the output E is (a
+ B + c + d + e + f + g + h + i) / 9.

FIG. 17 shows the configuration of the binarization processing section 5. The output of the image preprocessing unit 4 enters a comparator 52,
The value is compared with a threshold value set by the CPU 8 and binarized.
The binarized result is input to the shift register and byte-separated, and either the byte-sequenced signal or the 8-bit data of the output of the image pre-processing unit 4 with the lower part removed as is from the CPU 8 to the selector 54. Is selectable.

FIG. 18 shows LINEND, EN, and the timing of data from the A / D converter 3. L
Both INEND and EN are created from the mechanical control unit 9.

The LINEND signal is a line synchronizing signal for synchronizing the mechanical transport of the image sensor and the reading medium. When the image data is "H" (High level), no data is output from this LINEND signal. EN is a latch pulse that captures valid data from the sensor. VA (V1 to V1) which is data from the A / D converter 3
n) is latched at the rising edge of EN. The register 108 of FIG. 1 can reliably capture data by latching the input signal VA with EN.

As described above, the image distortion correction unit 1 incorporating the present invention
It is also conceivable to put 00 in a one-chip LSI.

FIG. 19 shows the configuration of the LSI of the image distortion correction unit 100.

Since the memory used in the shading storage unit 102 is not large in capacity, an internal memory is used. Further, since the division ROM used in the shading correction circuit 105 has a large memory capacity, the external memory 12
Is used. The memory controller 11 in the figure includes a range determining unit 1050, a range adjusting unit 1051, 1 shown in FIG.
052, a return area detection unit 1057, an overflow detection unit 1058, and address change units 1053 and 1054. The input signal includes a basic clock, LINEND output for each line, and EN / A / D output data for latching A / D output data. Also,
Input signals from the CPU include address and data control signals for chip select, write enable, and output enable. Signals for the external memory include control signals for chip select, write enable, output enable, address, and data (input / output). The output data is output and passed to the subsequent binarization processing unit.

It should be noted that the image distortion correction unit 10 incorporating the present invention
In the image reading apparatus 10 using 0, the output from the A / D converter 3 to the binarization processing section 5 can be put into a one-chip LSI.

[0052]

According to the present invention, it is possible to realize a low-cost circuit for performing high-precision shading correction at high speed even if the peak value of the read signal varies.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of an entire image processing apparatus of the present invention.

FIG. 2 is a diagram illustrating a concept of shading correction in consideration of a peak value.

FIG. 3 is a diagram showing a concept of a shading correction table.

FIG. 4 is a conceptual diagram illustrating a relationship between a signal and shading.

FIG. 5 is a diagram for explaining range adjustment.

FIG. 6 is a diagram illustrating a configuration of a shading correction unit.

FIG. 7 is a diagram illustrating a configuration of a reference peak detection unit.

FIG. 8 is a diagram illustrating a configuration of a signal peak detection unit.

FIG. 9 is a diagram illustrating a configuration of a range adjustment unit.

FIG. 10 is a diagram illustrating a configuration of an overflow detection unit.

FIG. 11 is a diagram illustrating a configuration of a mask processing unit.

FIG. 12 is a diagram illustrating a configuration of a return area detection unit.

FIG. 13 is a diagram showing a configuration of an address changing unit.

FIG. 14 is a diagram illustrating a configuration of a range determination unit.

FIG. 15 is a diagram illustrating a configuration of a shading storage unit.

FIG. 16 is a diagram illustrating a configuration of an image preprocessing unit.

FIG. 17 is a diagram illustrating a configuration of a binarization processing unit.

FIG. 18 is a diagram for explaining the timing of image data.

FIG. 19 is a diagram showing a configuration of an LSI on which the present invention is mounted.

[Explanation of symbols]

4 image preprocessing unit, 5 binarization processing unit, 6 recognition processing unit, 7 printer, 8 CPU, 9 mechanical control unit, 10
... image reading device, 100 ... image distortion correction unit, 101
... a reference peak detection unit, 102 ... a shading storage unit,
104: signal peak detection unit, 105: shading correction circuit, 106, 107: multiplier, 108: register.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Yoshiharu Konishi 1st Ikegami, Haruoka-cho, Owariasahi-city, Aichi Prefecture Inside the Information Equipment Division of Hitachi, Ltd. Information Technology Division of Hitachi, Ltd.

Claims (11)

[Claims] 1. A means for storing shading data of a reference image, a means for detecting a peak value of the shading data, a means for detecting a peak value of read input data, the input data and the stored reference image Means for multiplying the peak value of the input data by a signal peak register value which is the peak value of the input data up to the previous line and the peak value of the stored reference image data; An image distortion correction device, comprising: a shading correction unit that inputs two multiplication results and performs shading correction. 2. The means for detecting a peak value of the input data according to claim 1, wherein the comparing means compares the read input data with a peak value of a previous line; and a result of the comparing means and a peak value of shading data. And a signal peak register controlled by the comparing means. 3. An image distortion correcting apparatus according to claim 2, wherein a specific value is subtracted from the value of the signal peak register for each line. 4. An image distortion correcting apparatus according to claim 2, wherein the value of said signal peak register is equal to or larger than the peak value of the shading data. 5. An image distortion correcting apparatus according to claim 2, wherein the updating of the signal peak register is performed by the end of the read input data of one line and before the start of the next line. 6. An image distortion correcting apparatus according to claim 1, wherein said shading correcting means has a table look-up type divider. 7. A method according to claim 1, wherein said shading correction means has means for determining a range based on the state of upper bits of a divisor and a dividend, and means for performing range adjustment based on a result of said determination means. Image distortion correction device. 8. A sensor for inputting light applied to a reading medium and converting the light into an image signal, a converter for converting a signal from the sensor into a digital signal, and inputting the image signal converted to the digital signal An image distortion correction unit configured to correct the distortion of the image signal; and a recognition unit configured to perform a recognition process based on the image signal corrected by the image distortion correction unit. Means for storing shading data of an image of a reference plane, means for detecting a peak value of the shading data, means for storing a peak value of read input data, and a peak value of the input data and the peak value of the shading data. Means for multiplying the peak value of the input data by the shading data; and inputting the two multiplication results. The image processing apparatus characterized by having a shading correction means for. 9. The means for storing the peak value of the read input data according to claim 8, wherein the means for comparing the read input data with shading data at a pixel position corresponding to the input data is controlled by the comparing means. An image processing apparatus comprising: 10. An image processing apparatus according to claim 9, wherein said shading correction means has a table index type divider. 11. A signal peak register which detects a peak value of read input data, multiplies the input data by a peak value of data of a previously input reference image, and obtains a peak value up to a previous line of the input data. And a peak value of the data of the reference image input in advance, inputting the two multiplication results, and performing shading correction.