JPH04317261A - Image area identification system for image processor - Google Patents

Abstract

PURPOSE:To provide an image area identification system which is capable of identifying with high performance the respective area of an image in which color characters and intermediate tones are mixed and which has excellent hardware realization. CONSTITUTION:A blocking means 14 for plural picture elements, characteristic value detection means 16 to 18 detecting plural characteristic values from the picture element within the block, and area identification means 20 to 24 identifying the area of characters/intermediate tones from plural character values are provided. The plural characteristic values are available in three kinds: an average value within the block, and the total values of high level picture elements and intermediate level picture elements within the block, respectively. The area identification means 20 to 24 are provided with quantization means 20 to 22 for plural characteristic values, a LUT23 preliminarily set to a threshold value for the total value of the intermediate level picture elements for the area identification and a comparison means 24, and obtains the threshold value for the total value of the intermediate picture elements from the LUT23 while using a value obtained by quantization and synthesizing the average value and the total value of the intermediate level picture elements within the block as an address, compares the threshold with the actually obtained total value of the intermediate level picture elements and identifies the areas in the block.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image area identification method for an image processing apparatus that identifies areas of color characters/halftone images.

0002

[Prior Art] Color image processing devices such as color copying machines use Y (yellow), M (magenta), C (cyan),
A four-color toner developer consisting of K (black) is used, and these images are superimposed to reproduce a full-color original. Therefore, scanning is repeated four times in accordance with the development process of each toner, and the full-color image data obtained by reading the document is processed each time.

FIG. 8 is a diagram showing an example of the configuration of a conventional color image processing device and an example of the configuration of an image area identification circuit.

Generally, in a color image processing device, a line sensor is used to optically read a document, and
When image data is captured using the color separation signals of green (green) and red (R), the image data is converted to EN as shown in FIG.
D conversion 31, color masking (color correction)
32, it is converted into color toner signals YMC. Then, the UCR 33 generates a black plate (K) and removes the undercolor, and turns on/off the toner signal The data is converted into binary data, and a laser beam is controlled using the binary data to expose a charged photoreceptor to form an image of each color.

[0005] The hue separation type nonlinear filter section includes a UCR
In step 33, image data X of the developed color is selected according to the development process from the YMCK signals generated by performing black plate generation and undercolor removal processing, and this image data X is input and branched into two systems. A smoothing filter 34 is connected to one of them, and a γ conversion circuit 36, an edge detection filter 37, and an edge enhancement LUT 35 are connected to the other, and the former system performs smoothing processing on the halftone image.
In the latter system, edge enhancement is performed on character images. Then, these outputs are combined by an adder 39 and output as a signal after nonlinear filter processing. In this case, in edge processing, the hue of the input image is detected by the hue detection circuit 35, and it is determined whether the developed color at that time is a required color. Here, if the input image is a black area, Y
, M, and C, but only K is controlled to be emphasized according to the edge amount.

[0006] By introducing non-linear filter processing as described above, binary images such as characters and line drawings of different types and original documents such as photographs and halftone printed matter are processed. Nonlinear filter processing is performed to emphasize edges to increase sharpness, and to smooth halftone images such as photographs and halftone printed materials to improve image smoothness and graininess.

However, if it is easy to specify in advance the document or area for binary images such as characters and line drawings, and halftone images such as photographs and halftone printed materials, the By specifying the image type for each area, you can select the optimal parameters for each area.
Although reproducibility can be improved, in the case of mixed images, parameters that can be reproduced to a certain degree as both binary images and halftone images are selected. In this case, since the processing is not optimal for either binary images or halftone images, it is difficult to obtain an image that is satisfactory for both.

[0008] For example, in a binary image, the edge emphasis is weak and blurry, and the characters are not clear, and in the case of black characters,
There is a problem in that the edges and lowercase letters become cloudy. On the other hand, in the case of a halftone image, since the vicinity of the edge detection frequency is emphasized, the smoothness of the halftone image is lost, resulting in a rough image with strange moiré and edges being strangely emphasized. Therefore, it is necessary to identify three types of image areas, black characters, color characters, and halftones, and to switch the parameters of the filter section, and many identification processing methods of this type have already been proposed.

For example, the configuration shown in FIG. 8(b) (see Japanese Patent Application No. 2-2940) introduces an identification method based on run length. The hue detection unit 41 detects one of eight colors (Y, M, C, K, W, B, G, R) for each pixel, and the comparator 42 detects one of eight colors (Y, M, C, K, W, B, G, R), and the comparator 42 detects one of the hues of eight colors (Y, M, C, K, W, B, G, R). This is to detect whether or not the value is greater than or equal to a threshold value thmax. Since the hue detection unit 41 also detects low-density noise, etc., in this case, a problem arises in that the run length, which will be described later, becomes short and it is mistakenly recognized as a character area. Therefore, in order to select pixels having at least a certain level of density as character candidates, a threshold value thmax is set.
The comparator 42 performs a comparative judgment with the noise and removes noise.

The blocking unit 43 blocks several pixels, for example 4×4.
Each pixel is divided into blocks of ~8×8, and the block determination unit 44 performs seven-color determination. Seven-color determination is to determine the maximum frequency color within a block for seven colors (Y, M, C, K, B, G, R) excluding W, and use that as the block color. For example, count the frequency of each color with 4 x 4 pixels as one block, K = 6, M = 2
, R=1, C=1, and W=6, then the color K, which is the largest among the colors excluding W and black pixels, is adopted as the block color.

The main scanning direction color run count section 46 includes:
The comparator 47 receives one bit indicating whether it is W (0) or other black (1) from the block color determination unit 44 and counts the run length of the color/black block in the main scanning direction. It is determined whether the count value (run length) of each block is shorter than the threshold value thrun. Here, if the run length of a color/black block is shorter than the threshold thrun, it becomes a character candidate.

When viewed in terms of run length, the run length of the halftone dot area is short in pixel units, just like the character area, and it is difficult to distinguish it from the characters, but in the block unit, the run length of the halftone dot area is long, In the case of a character area, the block run length is short because the distance between characters is longer than the pitch of halftone dots. In this way, by looking at the block run length, it becomes possible to separate the characters and the halftone dot area.

The sub-scanning direction edge detection section 45 detects whether or not there is an edge within a range of several blocks in the sub-scanning direction, and if an edge is present, it is determined as a character candidate. This is because the horizontal lines used in the character area also have a long run length, so the above run length in the main scanning direction does not serve as a character candidate, so this is to detect it as a character candidate in such cases as well. .

The OR gate 48 and the AND gate 49 are used when a character candidate exists in either the main scanning direction or the sub-scanning direction from the above detection result, and the max flag of the block is 1.
If so, this block is determined to be a character area, and the block color signal determined by the block color determining section 44 is output. Note that the max flag is set to Y by the comparator 42.
This is a signal indicating that there is one or more pixels in the block whose maximum MC is determined to be equal to or greater than the threshold value thmax.

The above example focuses on the fact that the run length around the character portion is short. In other words, the text area consists of a block of characters and a block of background, and because it is in the background, the density changes sharply, whereas the image area is in the background of the image, and the density changes gradually. . Therefore, when observing the run length, the run length is short in the character area and long in the halftone area. Moreover, it was difficult to distinguish between halftone dot areas and text areas on a pixel basis, but in the case of halftone dots, the run length becomes longer by dividing them into blocks, so it is now possible to recognize them in halftone areas rather than in text areas. Can be done. In other words, due to blocking, halftone patch areas with high frequencies also tend to occur as color blocks with long run lengths.
It can be captured as a halftone area. Furthermore, if only the main scanning direction color run count section 46 and the comparator 47 detect character candidates, lines with long run lengths in the main scanning direction will be mistakenly recognized. In some cases, recognition accuracy is improved by using the characters as candidates.

Furthermore, the configuration shown in FIG. 8(C) (see Japanese Patent Application No. 2-136071) detects the maximum value and minimum value from a plurality of pixels in each block, and detects the difference between them when the difference is larger than a predetermined value. In addition to identifying the character area based on whether or not the character area is the same, error correction is also performed based on the area determination results of surrounding blocks.The area identification circuit is broadly divided into primary identification and secondary identification. The primary identification is an 8×8 blocking circuit 61
The character area is identified by forming blocks of 8 x 8 pixels into blocks, and for this purpose, the maximum value and minimum value are determined from the 8 x 8 pixels of the block by a maximum value detection circuit 62 and a minimum value detection circuit 63, respectively. The arithmetic circuit 64 calculates the difference between them. Then, a comparator 65 compares the output of the arithmetic circuit 64, that is, the difference between the maximum value and the minimum value, with a threshold th, and if the difference is larger than the threshold th, outputs a character area identification signal, for example, logic "1". . In the secondary identification, errors in the identification area of the block based on the primary identification are further corrected using 3×3 blocks.

For example, when the block of interest is identified as a character area in the primary identification, if most of the surrounding blocks are not character areas, the identification content of the block of interest is corrected to be a non-character area. Similarly, when the block of interest is identified as a non-text area in the primary identification, and most of the surrounding blocks are text areas, the identification content of the block of interest is corrected to be a text area.

In addition, there are also methods that use the average value and standard deviation within a predetermined pixel block (for example, Japanese Patent Laid-Open No. 63-205
No. 783) and those that use binary output converted by multiple dither conversions with different phases (for example, Japanese Patent Application Laid-Open No. 63-1937).
No. 70), the maximum and minimum values within the block are determined for each of the R, G, and B color separation signals, and at least 1
When the difference between two color components is greater than a predetermined value, it is determined to be a character area (for example,
Various methods have been proposed so far, such as (see Publication No. 631).

[0019]

[Problem to be Solved by the Invention] However, although the various conventional area identification methods described above certainly improve identification performance by performing block judgment, in reality, it is difficult to increase the speed of the device. In order to obtain high-quality color copies, the document must be scanned at high speed, and the scanned image data must also be processed at high speed. There is a problem in that noise causes variations in area recognition during the development process, resulting in defects such as interruptions in black characters.

That is, in color copying, as mentioned above, a full color image is reproduced by overlapping Y, M, C, and K toner images in four scans, so the difference between the maximum and minimum values, the run Even if the area is judged block by block using each characteristic value such as length, average value, standard deviation, etc., different judgments may occur during the four development processes of Y, M, C, and K. It is difficult to reliably obtain the same judgment each time.

For example, if the M and C development processes determine a black character area, but by chance the Y and K development processes do not determine a black character area, the black characters may be interrupted in that block. It will be output as Y. That is, the output of M and C is suppressed because the judgment is for a black text area, and the output of K is also suppressed because the judgment is for a non-black text area, and only Y is output. Moreover, since the determination is made in blocks and output, it becomes more noticeable that some black characters become colored characters compared to output in pixel units. Since such defects similarly appear in color characters and halftone areas, there is a problem of significant deterioration of image quality.

[0022]Furthermore, the conventional system is unsatisfactory because the hardware configuration is complicated, and conversely, if the hardware configuration is simple, the identification performance is poor. In particular, it has been difficult to match the hardware for black character/color character recognition in order to handle color images, and the system has had to be configured in a separate system from area recognition.

The present invention solves the above-mentioned problems, and is capable of identifying each region of a color text/halftone mixed image with high performance, and is excellent in hardware implementation and is capable of black/color discrimination in text areas. It is an object of the present invention to provide an image area identification method for an image processing apparatus that has good hardware compatibility with processing.

[0024]

[Means and effects for solving the problem] To achieve this, the image area identification method of the image processing device of the present invention includes a blocking means that blocks a plurality of pixels, and a characteristic that detects a plurality of characteristic values from pixels in the block. The method includes a value detection means and an area identification means for identifying a character/halftone area from a plurality of characteristic values. The feature is that the character/halftone area of the original image is identified, and the plurality of characteristic values within a block are determined by quantizing the average value within the block and the image signal into three levels using two thresholds. , and the total sum value of high-level pixels and intermediate-level pixels in each block, and character areas are identified in these spaces. This allows determination to be made in three-dimensional space, thereby increasing the accuracy of region identification.

[0025] Furthermore, the area identification means stores the identification results in an LUT.
The area identification means stores the plurality of characteristic values in the LUT and reads out the identification result in the LUT by using a value obtained by combining the plurality of characteristic values as an address, or comprises means for quantizing the plurality of characteristic values. It is configured to perform non-linear quantization, compress the word lengths of each word, and then synthesize it to create address data for the LUT. It is equipped with a set LUT, a comparison means for comparing the threshold value set in the LUT and the sum value of intermediate level pixels, and uses the quantized and synthesized value of the intra-block average value and the sum value of high level pixels as an address. By subtracting , a threshold value for the sum of intermediate level pixels is obtained, and the area within the block is identified by comparing this threshold with the actually obtained sum of intermediate level pixels, which reduces the identification performance. An adder/subtractor is provided after the LUT in which the threshold value of the sum of the intermediate level pixels is set, and the threshold value of the LUT and the adjustment value are added/subtracted to reduce the hardware scale without causing any problems. Since the determination threshold value is configured to be adjustable, the determination area for character identification can be adjusted.

The present invention is characterized in that a quantization means for quantizing pixel data is provided before the blocking means, and the word length of the pixel data is compressed when calculating the intra-block average value. The size of the blocking means can be reduced, and the blocking means is characterized in that it can expand the observation width of the block in the sub-scanning direction by thinning out according to the line synchronization signal, so that the identification accuracy during expansion can be improved. can prevent a decline in

[0027]Furthermore, there is a hue identification means for identifying hue in pixel units, a blocking means for dividing a plurality of pixels into blocks, and a block hue determining means for counting the hue signals of each block and determining the hue of the entire block by majority decision. After having a discrimination means and dividing the hue signal of each pixel into blocks,
The method is characterized in that the count value of each hue signal within the block is determined, and the hue of the entire block is determined by a majority decision.

[0028] The present invention is characterized in that it includes an area identification means for identifying three types of image areas of a document image, a black character area, a color character area, and a halftone area, from the character/halftone area identification signal and the hue identification signal. , comprising a quantization means for quantizing the area identification signal and the hue identification signal, and the word length of the signal obtained by combining the area identification signal and the hue identification signal is one of the block forming means.
The present invention is characterized in that the area identification signal and the hue identification signal are restricted to be within a finite word length of a pixel and can be simultaneously blocked by the same blocking means.

[0029] A means is provided for comparing pixels in a block with a threshold value for background determination, and an identification result is obtained when the block is determined to be a character area on the condition that there is at least one pixel in the block that is equal to or greater than the threshold value. It is characterized by the adoption of

[0030]

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

FIG. 1 is a diagram for explaining one embodiment of the image area identification method of the image processing apparatus according to the present invention, FIG. 2 is a diagram for explaining the distribution of color characters/halftone images, and FIG. 3 is a diagram for explaining the distribution of color characters/halftone images. FIG. 4 is a diagram for explaining an example of the distribution state of character areas, and FIG. 4 is a diagram for explaining parameter optimization conditions.

[0032] In a document containing a mixture of binary images such as line drawings and characters, and halftone images such as photographs and halftone dots, in order to collect the feature quantities of the text and halftone regions in block areas, as described above, Although various methods have been used in the past, these methods use only one feature, such as run length or the difference between the maximum value and the minimum value. In contrast, in the present invention, a predetermined number of pixels are divided into blocks, the intra-block average value Pa, the intra-block high-level pixel number Ph, and the intra-block intermediate-level pixel number Pm are collected, and their three-dimensional distribution state is An example of the basic configuration is shown in FIG. 1.

In the embodiment shown in FIG. 1A, first, an 8-bit intra-block average value Pa, a 6-bit intra-block high-level pixel number Ph, and a 6-bit intra-block intermediate-level pixel number Pm are converted into a quantizer. After quantizing to 3 bits, 4 bits, and 4 bits in 1 to 3, respectively, the threshold TH is read out using the 7-bit LUT 4 from the intra-block average value Pa and the intra-block high-level pixel number Ph, and the threshold value is read out using the comparator 5. A character/halftone area determination signal is obtained by comparing with the number Pm of intermediate level pixels in the block.

On the other hand, the embodiment shown in (b) uses the quantized intra-block average value Pa, the intra-block high-level pixel number Ph, and the intra-block intermediate-level pixel number Pm to calculate 1.
A 1-bit LUT6 is used to obtain a character/halftone area determination signal, and the embodiment shown in (c) uses the intra-block average value Pa and the number of high-level pixels within the block as they are without quantization. A 20-bit LUT 7 uses Ph and the number of intermediate level pixels in the block Pm to obtain a character/halftone area determination signal.

[0035] Considering a manuscript containing a mixture of characters and halftones, as shown in FIG.
If a low-level threshold th2 is set, a character, such as A, rises steeply from a low-level background, so
There are many pixels with a high level exceeding the threshold th1 and pixels with a low level that does not reach the threshold th2, and there are few pixels with an intermediate level between the threshold th1 and the threshold th2. In other words, relatively speaking, there are more low levels in the background than high levels in the text area. On the other hand, intermediate tones are evenly distributed at each level, for example B, but when viewed in terms of areas, they are distributed between low and intermediate levels, when they are distributed at intermediate levels, when they are distributed between intermediate and high levels. If it is distributed, it will be distributed at a high level. Therefore, when looking at the average value (intra-block average value Pa) and frequency, as shown in (b), characters are distributed on the lower side of the average value, and halftones are distributed on the higher side of the average value. .

That is, looking at each feature amount, it is as follows. The intra-block average value Pa is low in the case of a text area because there is more background than the high level of the text area. Also, the number of high-level pixels in the block P
h represents the total number of pixels in the block that were at a high level when the pixel signal was quantized into multiple levels (3 levels of high, medium and low), so in the case of a character area, at least one pixel exists,
Furthermore, there is also at least one low level pixel. If this condition is not met, for example, if all pixels in the block are high-level pixels, the area will be a halftone area. Then, in the number of intermediate level pixels Pm within the block, pixel signals are converted to multiple levels (
It represents the total number of pixels in a block that were at an intermediate level when quantized into three levels (high, medium, and low), and is small in character areas because the density distribution changes sharply.

[0037] Therefore, when the above three feature quantities are viewed three-dimensionally, the distribution state of the character area is as shown in FIG. That is, it can be seen that the range of characters can be considerably limited by combining three types of characteristic values.

Now, assuming that the total number of all pixels in a block is Pt, the character area must satisfy at least the following conditions.

[0039]■ Ph+Pm≦Pt-1■ 1≦P
h≦Pt-1 That is, when viewed on a plane consisting of Ph and Pm, the character area is distributed in the black background area shown in FIG. 4, and the halftone area is distributed in the white background area. Therefore, it becomes a character area only when it is ■ and satisfies ■.

Furthermore, the actual identification parameters are determined by statistically analyzing each image data and taking into account the above conditions. The configuration for realizing this is shown in FIG. 1 described above, in which (c) stores the identification results in a three-dimensional LUT in advance, The configuration is such that 20 bits obtained by combining the bit results are input as an address of the LUT. On the other hand, in (b), in order to reduce the capacity of the LUT, a quantizer is inserted into Pa, Ph, and Pm, and the data word length of each is compressed.
This is a UT with reduced capacity. In other words, in order to implement an LSI, it is necessary to make the internal memory as small as possible, and in this case, the reduction to 3, 4, 4 bits to 11 has been achieved. Of course, this quantization cannot be applied simply, but by making the distribution state and quantization of Pa, Ph, and Pm nonlinear, it can be confirmed that even with this degree of compression, there will be no problem with the identification performance. Ta. And (a)
The hardware scale of (b) is further reduced by focusing on the characteristics of Pm, and has a configuration using a two-dimensional LUT and one comparator. This is because when looking at Figures 3 and 4, it is noted that the distribution state of Pm in the character area always falls within the range where Pm = 0 is the minimum value and Pm≦Pt-Ph-1 is the maximum value. ing. As a result, the value of Pm becomes Pa
, Ph can be compared with the threshold value of the maximum value of Pm predetermined by Ph, and it is possible to realize a determination method in which the character region is determined if the value is smaller, and the halftone region is determined if the value is larger. This results in
Hardware feasibility can be improved without reducing identification performance.

FIG. 5 is a diagram showing a configuration example of a processing circuit to which the embodiment of FIG. 1(A) is applied, and FIG. 6 is a diagram showing a configuration example of a processing circuit to which the embodiment of FIG. 1(B) is applied. 7 is a diagram showing a configuration example of a processing circuit to which the embodiment of FIG. 1(c) is applied.

In FIG. 5, input signals L*, a*,
b* is an 8-bit signal expressed as a system value, and the L* axis represents brightness, and a* is orthogonal to this.
Saturation and hue are expressed on a two-dimensional plane of the axis and b* axis. The nonlinear quantizer 11 receives an 8-bit signal L of each pixel.
* to 4 bits, and the quantizer 12 converts the 8-bit signal L* of each pixel into thresholds th1 and th2.
Classified into high level / intermediate level / low level 2
The hue discriminator 13 outputs a 2-bit hue signal that identifies color, black, or white from the 8-bit signals L*, a*, b* of each pixel. It is.

The blocking FIFO 14 holds data of 4 lines in the sub-scanning direction and 8 pixels in the main scanning direction, and
In addition, by thinning out one line at a time from the eight input lines and retaining four lines, blocks corresponding to the size of 8×8 are performed. In other words, when attempting to increase the block size to improve accuracy, thinning blocks is employed. One pixel data held in the blocking FIFO 14 is a compressed 4-bit signal L as shown in FIG.
* It is an 8-bit signal consisting of a 2-bit level division signal and a 2-bit hue signal, each of which is a 4-bit signal L*.
is distributed to an average value calculation circuit 16 and a comparator 15, a 2-bit level classification signal is distributed to counters 17 and 18, and a 2-bit hue signal is distributed to a block hue discriminator 19.

The average value calculation circuit 16 calculates the intra-block average value Pa by adding the signals L* of the pixels within the block. The counter 17 counts the number of high level pixels Ph from the level division signal, and the counter 18
is for counting the number of intermediate level pixels Pm. In the quantizer 12, the low level threshold is set to th1, the high level threshold is set to th2, and the signal L* is set to the thresholds th1 and th.
2, the lower bit of the 2 bits is set to "1", and when it exceeds the threshold th2, only the upper bit is set to "1", the counter 17 counts the upper bits, and the counter 18 will count the lower bits.

After each data is compressed by nonlinear quantizers 20 to 22 similar to those shown in FIG. 1(a), the LUT
23, the comparator 24 outputs a character/halftone determination signal. LUT23 is Pa-P in the three-dimensional space shown in Figure 3.
This is to read the threshold value of Pm in the block indicated by the solid line in the figure, which is a character area in the h plane (left plane in the figure), and if this threshold value is larger in the comparator 24, the solid line in the three-dimensional space shown in FIG. 3 is read out. enters the block, and a character area determination signal is output. The adder inserted and connected between the LUT 23 and the comparator 24 is used to adjust the bias of the threshold value read from the LUT 23 using the bias THbias, and is used to adjust the character area.

The comparator 15 and flag detector 25 are for detecting whether it is a background block or not, and compare the 4-bit signal L* compressed by the comparator 15 with a threshold value THav.
A flag detector detects whether there are pixels in the block that exceed the threshold value THav. Therefore, here, if all the pixels in a block are smaller than the threshold THav, it is determined that the block is the background. In other words, the threshold TH
If there is not at least one pixel that exceeds av, it is treated as a background rather than a character or halftone. The error correction circuit 26 inputs a character/halftone determination signal and a background detection signal, and performs correction in larger block units by comparing it with, for example, the determination signals of surrounding blocks.
Outputs a halftone judgment signal.

[0047] The block hue discriminator 19
The hue signal of each pixel divided into blocks by the IFO 14 is used to identify whether the hue is color or black and white in each block by majority decision, and the delay circuit 28 corrects errors by delaying the color/black and white block hue signal. This is to synchronize the character/halftone determination signal of the circuit 26. The final determination circuit 27 finally outputs a color character/black character/halftone T/I (text image) region determination signal from the character/halftone determination signal and the color/monochrome block hue signal. be.

In the example shown in FIG. 6, the nonlinear quantizer of the signal L* is omitted, only the upper 4 bits are input to the blocking FIFO, and the intra-block average value calculated by the average value calculator 16 and counters 17 and 18 is calculated. After compressing Pa, the number of mid-level pixels in the block Pm, and the number of high-level pixels in the block Ph using nonlinear quantizers (LUTs) 20 to 22, these are input to the LUT 23 for direct determination, and are input with 11 bits. It is configured to obtain a 2-bit output.

Further, in the example shown in FIG. 7, the toner signal ymc
is used as an input signal, the maximum value detection circuit 31 is used to collect the same signal as in the previous embodiment, and the intra-block average value P
LU for direct determination without compressing a, the number of intermediate level pixels in the block Pm, and the number of high level pixels in the block Ph.
It is configured to obtain a 2-bit output with a 20-bit input to T32.

Note that the present invention is not limited to the above-described embodiments, and various modifications are possible. For example, in the above-mentioned embodiment, the blocking circuit blocks the blocks to a predetermined size and sequentially outputs area identification signals for each block, but the blocks to be determined may be wrapped. By doing so, it is possible to improve the stability of determination and the accuracy of determination. Also,
The basic explanation has been based on a configuration in which signals L*, a*, b* consisting of lightness, saturation, and hue information are input and processed as area detection signals, but as shown in the example shown in FIG.
It goes without saying that, in addition to the input, the light color separation signal BGR may be input, and the maximum value or minimum value may be used, or the luminance signal may be used. Furthermore, the block size may be changed depending on the reduction/enlargement ratio. For example, when enlarging, the number of lines in the sub-scanning direction decreases, resulting in a problem of reduced identification accuracy, but this problem can be solved by changing the block size.

[0051]

As is clear from the above description, according to the present invention, a plurality of pixels are divided into blocks, and three values are calculated: the average value Pa within the block, the number Ph of high-level pixels within the block, and the number Pm of intermediate-level pixels within the block. Collect two features and
Since character/halftone area determination is performed in a three-dimensional space consisting of these, the degree of freedom in setting area determination is increased, and the accuracy of area determination can be improved. In addition, nonlinear quantizers and LU
By using a compressor such as T, it is possible to easily adjust the determination value and to reduce and simplify the hardware scale. Furthermore, by quantizing the input signal and creating blocks, line memory (FIFO) for blocking is used.
) can be reduced.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining one embodiment of an image area identification method of an image processing apparatus according to the present invention.

FIG. 2 is a diagram for explaining the distribution of color characters/halftone images.

FIG. 3 is a diagram for explaining an example of the distribution state of character areas.

FIG. 4 is a diagram for explaining parameter optimization conditions.

FIG. 5 is a diagram showing a configuration example of a processing circuit to which the embodiment of FIG. 1(A) is applied.

FIG. 6 is a diagram illustrating a configuration example of a processing circuit to which the embodiment of FIG. 1(b) is applied.

7 is a diagram showing an example of the configuration of a processing circuit to which the embodiment of FIG. 1(c) is applied; FIG.

FIG. 8 is a diagram showing a configuration example of a conventional color image processing device and a configuration example of an image area identification circuit.

[Explanation of symbols]

1 to 3, 11, 12, 20 to 22...quantizer, 4, 6,
7, 23... LUT, 5, 15, 24... Comparator, 13... Hue discriminator, 14... Blocking FIFO, 16... Average value calculator, 17, 18... Counter, 19... Block hue discriminator, 25... Flag Detector

Claims (12)

[Claims] 1. Blocking means for forming a plurality of pixels into blocks; characteristic value detection means for detecting a plurality of characteristic values from pixels within the block; and area identification for identifying character/halftone areas from the plurality of characteristic values. The method is characterized in that it is configured to block a plurality of pixels, detect a plurality of characteristic values in the block, and identify characters/halftone regions of the original image from the plurality of characteristic values. An image area identification method for an image processing device. 2. The plurality of characteristic values within a block are the average value within the block and the sum total value within each block of high level pixels and intermediate level pixels when the image signal is quantized into three levels using two threshold values. 2. An image area identification method for an image processing apparatus according to claim 1, characterized in that there are three types. 3. The area identification means stores the identification result in an LUT, and uses a value obtained by combining a plurality of characteristic values as an address in the LUT.
3. The image area identification method for an image processing apparatus according to claim 1, wherein the method is configured to read out identification results in the UT. 4. A plurality of characteristic value quantization means are provided, and the area identification means is configured to quantize the plurality of characteristic values, compress the respective word lengths, and then synthesize the plurality of characteristic values to obtain address data of the LUT. 4. An image area identification method for an image processing apparatus according to claim 3. 5. The region identification means includes quantization means for a plurality of characteristic values, an LUT in which a threshold value of the sum of intermediate level pixels for region identification is set in advance, and a threshold value set in the LUT and the intermediate level pixels. and a comparison means for comparing the total sum value of the intermediate level pixels, and calculates the threshold value of the total sum value of the intermediate level pixels by pulling the LUT using the quantized and synthesized value of the intra-block average value and the total sum value of the high level pixels as an address, 3. The image area identification method for an image processing apparatus according to claim 2, wherein the area within the block is identified by comparing the threshold value with a total value of intermediate level pixels actually determined. 6. An adder/subtractor is provided at a subsequent stage of the LUT in which a threshold value for the sum of the intermediate level pixels is set, and the LUT
6. The image area identification method for an image processing apparatus according to claim 5, wherein the determination threshold in the comparison means can be adjusted by adding or subtracting the threshold and the adjustment value. 7. Image region identification of an image processing apparatus according to claim 1 or 2, characterized in that quantization means for quantizing pixel data for obtaining a plurality of specific values is provided before the blocking means. method. 8. Image area identification of the image processing apparatus according to claim 1, wherein the blocking means is configured to expand the observation width of the block in the sub-scanning direction by thinning out the blocks in accordance with the line synchronization signal. method. 9. Hue identification means for identifying hue on a pixel-by-pixel basis; blocking means for forming blocks from a plurality of pixels;
It is equipped with a block hue identification means that counts the hue signals of each block and determines the hue of the entire block by majority decision. An image area identification method for an image processing device, characterized in that the hue of the entire block is determined by majority decision. 10. The present invention is characterized by comprising an area identification means for identifying three types of image areas of a document image, a black character area, a color character area, and a halftone area, from a character/halftone area identification signal and a hue identification signal. An image area identification method for an image processing apparatus according to claim 1 or 9. 11. Quantization means for quantizing the area identification signal and the hue identification signal, wherein the word length of the signal obtained by combining the area identification signal and the hue identification signal is a finite word of one pixel of the blocking means. Claim 1 characterized in that the area identification signal and the hue identification signal are configured so that they can be simultaneously blocked by the same blocking means by limiting the signal to within a length.
Alternatively, the image area identification method of the image processing apparatus according to 10. 12. Means for comparing pixels in a block with a threshold value for background determination, wherein the block is determined to be a character area on the condition that there is at least one pixel equal to or greater than the threshold value in the block. The image area identification method for an image processing apparatus according to claim 1 or 10, characterized in that the identification result is adopted.