JPH01181380A - Binarization processing circuit for picture information - Google Patents

Abstract

PURPOSE:To reproduce an edge part clearly even for a character with thick line width by separating accurately input multi-value picture information into a contrast picture area and a binary level area and applying optimum simple slice binarization and pseudo intermediate tone binarization respectively. CONSTITUTION:An image area discrimination circuit 14 discriminates input multi-value picture information as to whether it belongs to a binary level picture area or a contrast picture area in the unit of picture elements and allows a selector circuit 13 to select the simple binary picture information or pseudo intermediate binary picture information in response to the result of discrimination. Then the selector circuit 13 outputs the binary picture information in the mixture of simple binary picture information and pseudo intermediate tone binary picture information. The front and tail ridges are represented as black level in its waveform and the middle part is represented as the contrast level in response to the original. Thus, even a character with a thick line width is reproduced as a picture easy to see with clear contour.

Description

[Detailed Description of the Invention] [Summary] Image information that determines each image area and performs optimal binarization processing on input multivalued image information such as a mixture of grayscale images and monochrome two-level images. Regarding the binarization processing circuit.

The object of the present invention is to realize a means that can accurately determine whether an input image is a grayscale image or a two-level image with a simple configuration.

A simple slice processing circuit that converts input multilevel image information into simple binary image information, a pseudo halftone processing circuit that converts input multilevel image information into pseudo halftone binary image information, and a simple slice processing circuit that converts input multilevel image information into pseudo halftone binary image information; A selector circuit inputs halftone binary image information and selects one of them pixel by pixel according to an instruction, and a grayscale image for each pixel of input multilevel image information to determine whether it belongs to a two-level image area or not. It determines whether it belongs to the area, in particular detects the short edges and edges, determines it to be a two-level image area, and sends simple binary image information or pseudo-intermediate iJ! to the selector circuit. and an image area determination circuit that instructs to select binary image information.

[Industrial application field]

The present invention relates to an image information binarization processing circuit that determines each image area and performs optimal binarization processing on input multivalued image information such as a mixture of grayscale images and monochrome two-level images. .

It distinguishes between images that require shading, such as photographs, and images that emphasize resolution, such as characters and figures, and selects the optimal two to obtain high-quality binarized image information for both.
Provides technology that performs value processing.

[Conventional technology]

In a general image processing system, when inputting an image using an image scanner, etc., the original image is scanned using COD and multivalued (
(multi-gradation) image information, and then performs simple slicing processing using a threshold value.

Binarized image information is obtained.

However, manuscript images include photographic images expressed in gray levels (halftones), images of characters and line drawings expressed in two levels such as black and white, etc.
There may be a mixture of various things, and if you perform binarization processing by simple slicing using the same threshold value for all of them.

Depending on the identification of the original image, the quality of the original may be significantly degraded.

Therefore, there are two types of images: images with grayscale level expression (hereinafter referred to as grayscale images) and images with two-level expression such as black and white (hereinafter referred to as grayscale images).

A method has been developed that performs binarization processing by discriminating between two-level images (called two-level images) and applying an optimal threshold to each.

In this case, when a document image is read, it is necessary to determine for each pixel whether it is in a grayscale image area or a two-level image area. Conventionally, from multiple pixels surrounding an arbitrary pixel of interest on two images.

a, tHow to check the 11 degree histogram.

b. Method for determining the difference between maximum and minimum concentration.

A method to find the average value and variance of C0 concentration.

d. A method of calculating the run length and determining whether the pixel of interest belongs to a gray or two-level image area based on the magnitude of the run length value.

etc. were taken.

In this last method d, with X as the reading position, D
Binarization processing is performed using a conversion function g (D(X)) with (x) as the density level. This conversion function g (D (x
)) by the threshold α.

g(D(x))=1: D(x)≧αg (D
(x) ) =O: D (x) < α.

The run length around the pixel of interest is calculated based on the pixel information binarized using this conversion function. When the calculated run length is larger than a certain threshold β, the pixel of interest is It is determined that it belongs to the grayscale image area.

In this case, the run length of a two-level image such as a document mainly composed of characters is generally extremely short, but the determination is made based on the fact that the run length of a photographic image is long.

Next, explanation will be given using a specific example.

Figure 9 shows C on the original image in the image scanner.
This shows the scanning direction of OD, ■, ■.

Each arrow of ■... represents a main scan (line scan) electrically performed within the CCD, and the sequential main scans ■, ■
, ■... The vertical scanning is shown by the array.

This represents sub-scanning performed by mechanically moving the COD downward or mechanically moving the document upward.

When counting the run length, in the main scanning direction, it can be easily calculated from pixel information continuously obtained during main scanning of the CCD, but i! In one scanning direction, it is not possible to continuously obtain information for each pixel from the COD, so a line memory or large-capacity buffer corresponding to the maximum value required for the run length is prepared, and each pixel information of consecutive multiple lines is Pixel information is temporarily stored in them, and the information is obtained by reading out the same pixels on a line in the vertical direction.

Figure 1Q is an example of 2-image information, and represents the density signal waveform when a document image containing a mixture of characters, line drawings, photographs, etc. is read at 16 gradations in the main scanning direction (X). In the example, the center part is 2 levels such as text and line drawings, image area, left,
The right part is a grayscale image area such as a photograph.

Figure 11 shows the result of simple slice binarization of the 16-gradation representation image information in Figure 10 using a threshold of density level 4.
Value represents image information.

In this case, the aforementioned conversion function g (D(x)) is.

It is given as follows (α=4).

g(D(x)) = 1: D(x) ≧4g
(D (x) ) = g (D (x)) = 0: D (x) < 4 Here, the pixel with g (D (x)) = 1 is a black pixel, and g (D (x)) = A pixel of 0 is defined as a white pixel.

Therefore, ■ to ■ in FIG. 11 represent a succession of black pixels, that is, a black run length. The number of pixels of each black run length (■) to (■) is 14.4°3.20.

So, in the case of the example in Figure 11, ? ! As a threshold value β that separates the grayscale image area and the two-level image area.

By setting "5", the black run lengths ■ and ■ in FIG. 6 become the grayscale image area and the black run length ■.

2 can be determined to belong to the 2-level image area.

FIG. 12 shows an example of a conventional binarization processing circuit having a function of discriminating between a gray scale image area and a two-level image area using such a method, where 1 is a binary converter.

2 is a demultiplexer, 3-1 to 3-n are n input buffers (n is the maximum value of the desired run length), 4 is a multiplexer, 5 is an n-stage shift register,
6 is a reference pattern generator, 7 is a comparator, 8 is a two-stage shift register, and 9 is an AND circuit.

The binary converter 1 is a CO converter that operates according to the scanning method shown in FIG.
Multivalued image information as shown in FIG. 10 outputted from D is converted into binary image information as shown in FIG. 11 using a threshold value α.

The binary image information output from the binary converter 1 is as follows.

The demultiplexer 2 cyclically writes each line to each of the input buffers 3-1 to 3-n. In other words, each of the input buffers 3-1 to 3-n has a pixel capacity for one line, and the demultiplexer 2 selects the next empty input buffer after writing line data to one input buffer. and then
Image information for six lines is written continuously.

On the other hand, the multiplexer 4 selects, in a fixed manner, the input buffer for which writing has been completed, and transfers its contents to the shift register 5.

When determining the run length in the main scanning direction, the multiplexer 4 selects one of the input buffers 3-1 to 3-n in which the pixel of interest exists, and transfers information on n pixels in the horizontal direction to the shift register 5. However, when determining the run length in the sub-scanning direction, the multiplexer 4 reads out the data corresponding to the vertical direction of each input buffer 3-1 to 3-n for each target pixel position (x) on the line. The information of n pixels is scanned and continuously read out to the shift register 5.

The reference pattern generator 6 is composed of a ROM, and generates reference patterns of various runs for distinguishing between a grayscale image and a two-level image. That is, a reference pattern of a run is generated using all possible combinations of n pixel values.

The comparator 7 compares the contents of the image information with each reference pattern each time the image information is written into the shift register 5, and inputs the comparison result with the reference pattern with a run length larger than the threshold value β to the shift register 8. Write in columns. Here, the comparison result when the run length of the 9 image information is larger than the reference pattern is assumed to be "l".

This run length detection operation is performed continuously in the main scanning direction and the sub-scanning direction for each pixel of interest.

The shift register 8 has a two-stage configuration, and when a comparison result is input to the input stage, it performs a shift operation and holds the comparison result in the main scanning direction and the comparison result in the sub-scanning direction in each stage.

When the contents of each stage of the shift register 8 are both 11'', the AND circuit 9 outputs a determination result 11'' indicating that the pixel of interest belongs to the grayscale image area.

In other cases, a determination result of "0" indicating that the image belongs to the 2-level image area is output.

[Problem to be solved by the invention]

Conventionally, various methods have been used to determine which image a pixel to be processed belongs to when performing binarization processing in real time on halftone image information in which a grayscale image and a two-level image are mixed. But none of them.

It may not be possible to obtain accurate judgment results, or it may not be suitable for real-time processing because it requires complex calculations.

It had drawbacks such as requiring a large amount of memory.

Especially in circuits using run length methods.

In order to determine the run length in the sub-scanning direction, a large number of input buffer sofas (or line memories) are required.

Furthermore, even if the image is originally a two-level image of nine characters and line drawings, if the line width is thicker than a certain level, the run length will be large, resulting in an inconvenience that the image will be misidentified as a grayscale image.

SUMMARY OF THE INVENTION An object of the present invention is to realize a means that can accurately determine whether an input image is a grayscale image or a two-level image with a simple configuration.

[Means to solve the problem]

In order to accurately determine the gray scale image area and the 2-level image area in an input image in the binarization processing circuit in pixel units, the present invention is directed to The run length is obtained by slicing the multivalued image information using an appropriate threshold value, and the small run length value corresponds to the line width of characters, line drawings, etc., and the pixels included in the run are divided into two-level image areas. Also, instead of finding the run length in the sub-scanning direction, edge detection is performed in the main scanning direction, and if there is an edge part where the density level changes rapidly, it is considered as part of a 0 character/line image. If the run length value is thick (if the pixel is in the edge part, a function is provided to determine that the pixel in the edge part also belongs to the 2-level image area, and a single threshold value is applied to the pixel determined to belong to the 2-level image area. This method performs binarization processing using slices.

FIG. 1 is a diagram showing the basic configuration of the present invention.

10 is a delay circuit, which inputs multivalued image information,
It has a timing adjustment function that delays the time required for image area determination.

Reference numeral 11 denotes a simple slice processing circuit, which has a threshold value {circle around (2)} that determines a slice level suitable for binarizing a two-level image, and converts delayed input multi-value image information into simple binary image information.

Reference numeral 12 denotes a pseudo halftone processing circuit, which dithers the delayed input halftone image information using a dither threshold pattern and converts it into pseudo halftone binary image information.

A selector circuit 13 selects either simple binary image information or pseudo halftone binary image information for each pixel.

Reference numeral 14 denotes an image area determination circuit, which determines whether each pixel of the input multilevel image information belongs to a two-level image area or a grayscale image area, and, depending on each determination result, sends a signal to the selector circuit 13, The user is prompted to select simple binary image information or pseudo halftone binary image information.

15 is a simple slice processing circuit similar to the simple slice processing circuit 11;
is, for example, the threshold value when binarizing a two-level image.
Although the value is set to be slightly smaller than that, it is not limited to this value.

16 is a small run length detection circuit which slices the input multilevel image information at an appropriate level and calculates the result as follows.
When the run length of a black run is determined, and if the pixel of interest is included in a run with a run length less than or equal to a predetermined threshold value r, the pixel is considered to be part of a line segment or point, and the pixel is added to the two-level image area. It is judged that it belongs to.

Reference numeral 17 denotes an edge detection circuit, which checks density level changes around the spout pixel in the delayed input multi-valued image information;
If there is a density level change of a certain amount or more, it is considered to be an edge portion, and the pixel of interest is determined to belong to a two-level image area.

[For production]

FIG. 2 is a signal waveform diagram for explaining the operation of the binarization processing circuit of the present invention shown in FIG. 1.

In the figure, fat is a representation of the obtained input multilevel image information in the form of a density level waveform obtained by reading original pixels in the main scanning direction, and in the figure (b) is output from the simple slice processing circuit 15 in Figure 1. Binary image information is expressed as a binary waveform.

In FIG. 2 (al), waveforms I and ■ are character image portions with two levels of black and white, and waveform ■ is a grayscale image inspection.

In addition, the threshold value ■ is determined by the simple slice processing circuit 11 in FIG.
is the threshold value used for simple slicing.

Threshold {circle around (2)} is a threshold used by the simple slice processing circuit 15 for simple slices.

The simple slice processing circuit 11 in FIG. 1 performs simple slices on the waveforms I, II, and Dl of FIG. 2 (al) using the threshold value ■, and converts them into simple binary image information.

Similarly, the pseudo-halftone processing circuit 12 dithers each of the waveforms I, If, and III using an appropriate dither threshold pattern to convert them into pseudo-halftone image information.

Each piece of converted image information is input to the selector circuit 13.

On the other hand, in the image area determination circuit 14 of FIG. 1, the simple slice processing circuit 15 uses the threshold value ■ to generate the waveform 1.+8) in FIG. Il, I[[ is simply sliced, and the short run length detection circuit 16 uses the result as a threshold value r
Detect runs with run length values smaller than . In the case of FIG. 2(a), waveform I corresponds to this.

The image area determination circuit 14 instructs the selector circuit 13 to select simple binary image information as corresponding to waveform I.

The edge detection circuit 17 also detects waveform 1. This edge detection circuit 17 detects edge portions where the density level rapidly changes for each of n and I[[.

Digital filters using calculations such as Laplacian, gradient, and Sobel can be used. FIG. 2 (In the 810 example, the leading and trailing edges of the waveform ■ are edge-detected,
The image area determination circuit 14 instructs the selector circuit 13 to select simple binary image information.

The image area determination circuit 14 determines that the area other than the detected short run length and edge portion is a grayscale image, and the selector circuit 1
3 to select pseudo halftone binary image information.

In this way, from the selector circuit 13,
As shown in FIG. 2, binary image information is output in the form of a mixture of simple binary image information and pseudo-halftone binary image information.

Here, the leading and trailing edges of the waveform ■ are represented as black levels, but the central part is represented as a gray level depending on the original, so even characters with thick lines have a clear outline. It can be played back as an easy-to-see image.

〔Example〕

FIG. 3 is a block diagram of a binarization processing circuit according to an embodiment of the present invention.

In FIG. 3, 10 is a delay circuit composed of an r-stage shift register, 11 is a simple slice processing circuit using a threshold value ■, 12 is a pseudo halftone processing circuit, 13 is a selector circuit, and 14 is an image area determination circuit. , 15 and 19 are simple slice processing circuits using threshold value ■, 16a is a run length counter, 16b is a run length judgment circuit, 16C is a JK flip-flop, 16d is an AND circuit, 17a is a line buffer, and 17b is an edge detection filter. , 18 is an OR circuit.

Here, in FIG. 1 and FIG. 3, elements referred to by the same numbers have corresponding functions. However, 16 corresponds to 16a to 16d.

17 corresponds to 17a and 17b. Also, the threshold
.

■ and r are common to both figures.

FIG. 4 is a signal waveform diagram for explaining the operation of the embodiment circuit of FIG. 3.

CLK: Input pixel clock A: Signal obtained by binarizing the input pixel with threshold value ■ B: Signal that detects the run of the part of the signal A whose run length is larger than threshold value r C: Signal A Signal delayed by one pixel clock *D: From signal B and signal C, detects a run whose run length is r or more. Signal E: Performs the AND of signal C and signal R, and the run length is smaller than r. Signal F for detecting a run: Signal G detecting an edge portion from a signal delayed by one pixel clock: Image area determination signal based on the logical sum of signal E and signal F.

The operation will be explained with reference to FIGS. 3 and 4.

The input multilevel image signal is applied to a simple slice processing circuit 15 and a delay circuit 10.

The signal A binarized by the simple slice processing circuit 15 is counted by the run length counter 16a to count the run length of black runs, and the result is input to the run length determination circuit 16b, where it is compared with a threshold value r. In the illustrated example, a signal B that becomes "H" when the run length is greater than 0 is output to the J terminal of the flip-flop 16C.
has a length of 5 pixel clocks.

On the other hand, the input multilevel image signal delayed by one pixel clock in the delay circuit 10 is binarized in the simple slice processing circuit 19, and is converted into a signal C by JK in timing with the signal B.
It is applied to the terminal of flip-flop 16C.

The JK flip-flop 16C turns on when the signal levels of the J and J terminals both go to "H" and sets the Q terminal to "L", but when both go to "L", it turns off and sets the Q terminal to " Set to “H” (signal *D).

Signal *D is obtained by extracting only black runs with a run length larger than the threshold value r from signal C.

The AND circuit 16d removes a black run portion with a run length larger than the threshold value r from the signal C by taking the AND of the signal C and the signal *D (signal F).

Simple slice processing circuit 11. pseudo halftone processing circuit 12;
Since the function of the selector circuit 13 is the same as that explained in FIG. 1, the explanation will be omitted here.

The delayed input halftone image signal is stored in the line buffer 17a for the number of lines required for edge detection.

The edge detection filter 17b extracts data of a predetermined plurality of pixels from the inner buffer 17a for each pixel of interest, and performs filter processing to detect an edge portion (signal F). Note that the threshold value T is used as a reference value for detecting edge portions.

The signal E and the signal F are combined by the OR circuit 18 to form the signal G.
becomes. This signal G indicates only a run or a part (edge) of a run belonging to the 2-level image area in the signal C, and is output from the simple slice processing circuit 11 when it is at the 1H'' level to the selector circuit 13. is selected, and when it is at the "L" level, the output of the pseudo halftone processing circuit 12 is selected.

Next, the filtering process of the edge detection filter 17b will be explained.

The edge detection filter 17b is a kind of second-order differential or first-order differential filter, and taking a 3×3 pixel matrix as shown in FIG. 5 as an example.

For example, in the Laplacian method.

12SO-ΣSt/41≧T1...(1)
(i=0.1.2...,8) In the gradient method.

l S8-34 1 + l S2-S6 1 ≧
T2... Performs the calculation of (2), T1. T2 represents a threshold value.

The Laplacian method constitutes a digital filter that passes the second-order differential components of nine images, and the gradient method constitutes a digital filter that passes the -th order differential components.

Figure 5-) is a block diagram of an embodiment of an edge detection filter using the Laplacian method. Line buffer A in Figure 5 stores the line data of one line before, and line buffer B stores line data of two lines before. Data ■ is stored.

The ΣSt calculation unit calculates the current line data ■ and the line data ■ and ■ in each line buffer A and B.

Input each, calculate ΣSi, +2SO−ΣSi
/41 calculation unit performs difference calculation with So.

The result of this operation is expressed in a comparator (not shown).
It is compared with the threshold value T1, and the result of equation (l) is obtained.

An edge detection filter using the Laplacian method.

Since high spatial frequency components are extracted, in the case of an original image such as a halftone photograph, a large number of edges are detected in the photograph, which may deteriorate the quality of one image.

To improve this problem, it is effective to use an edge detection filter based on Sobel calculation.

Edge detection filter using Sobel calculation.

l S1+2S8 +57-(S3+2S4 +S5
') 1+ l S1+2S2 +53-(S7+25
6 +55) I≧T3 (3) is calculated, indicating that the difference calculation is performed after smoothing processing is performed in units of 2×2 pixels. Note that T3 is a threshold value. By including this smoothing process, the influence of noise at high frequencies can be reduced.

〔Effect of the invention〕

As described above, according to the present invention, input multilevel image information is accurately separated into a grayscale image area and a two-level image area, and optimal simple slice binarization and pseudo halftone binarization are performed, respectively. In particular, the edges of characters and line drawings with thick lines are clearly reproduced, and even grayscale images such as original photographs are reproduced with image quality comparable to that of conventional methods.

6 to 8 are examples of binarization processing for the same original image to show the performance difference between the present invention and the conventional example,
Figure 6 shows a reproduced image obtained by conventional simple slice binarization, Figure 7 shows a reproduced image obtained by conventional pseudo-halftone binarization (dither method), and Figure 8 shows a reproduced image obtained by the present invention. Each shows a reproduced image when image area determination is performed and binarized.

As is clear from the comparison of the reproduced images in these figures, the reproduced images according to the three embodiments of the present invention maintain good image quality in both the photographic portion and the character/line drawing portion.

Further, the present invention has the advantage that the amount of hardware such as an inset sofa can be reduced compared to conventional ones, and it can be easily operated in real time.

[Brief explanation of the drawing]

FIG. 1 is a basic configuration diagram of the present invention, FIG. 2 is a signal waveform diagram explaining the present invention in detail, FIG. 3 is a configuration diagram of a binarization processing circuit according to an embodiment of the present invention, and FIG. 4 3 is a signal waveform diagram of the embodiment circuit of FIG. 3, FIG. 5 is an explanatory diagram of an embodiment of an edge detection filter, and FIG. 6 is a diagram of a reproduced image by conventional simple slice binarization processing. FIG. 7 is a diagram of a reproduced image obtained by the conventional halftone binarization process, FIG. 8 is a diagram of a reproduced image obtained by the binarization process of the present invention, and FIG.
The figure is an explanatory diagram of document scanning by COD. Fig. 10 is a density signal waveform diagram of an example of image information containing a mixture of characters/line drawings and photographs, etc. Fig. 11 is a signal waveform diagram of image information obtained by converting the density signal waveform of Fig. 10 into simple slice binarization.
FIG. 2 is a configuration diagram of a conventional example of a binarization processing circuit. In Figure 1. 10: Delay circuit 11.15': Simple slice processing circuit 12: Pseudo halftone processing circuit 13: Selector circuit 14: Image area judgment circuit 16: Small run length detection circuit 17: Edge detection circuit 11 yuan θIroi story world 4'51115 2
Illustration May 18, 1986

Claims (1)

[Claims] 1. Input multi-valued image information obtained by reading a gray-scale image such as a photograph or a two-level image such as a character or line drawing in multiple gradations and convert it into binary image information in real time. A simple slicing processing circuit (11) that slices input multi-valued image information pixel by pixel using a predetermined threshold value and converts the input multi-valued image information into simple binary image information; a pseudo-halftone processing circuit (12) that performs dither processing on a pixel basis using a predetermined dither threshold pattern and converts it into pseudo-halftone binary image information; and simple binary image information and pseudo-halftone binary image information. a selector circuit (13) that selects one of them pixel by pixel according to an instruction, and a selector circuit (13) that selects one of them pixel by pixel according to an instruction, and determines whether each pixel of the input multivalued image information belongs to a two-level image area or a grayscale image area. an image area determination circuit (14) that determines whether the image information is simple binary image information or pseudo halftone binary image information and instructs the selector circuit (13) to select simple binary image information or pseudo halftone binary image information; calculates the run length by slicing the input multilevel image information using an appropriate threshold, determines the pixels included in the run whose run length is smaller than a predetermined value as belonging to the two-level image area, and Binary image information, characterized in that it is configured to detect an edge portion of a run whose run length is greater than the predetermined value, and to determine that pixels included in the edge portion also belong to a two-level image area. processing circuit. 2. In claim 1, the simple slice processing circuit (11)
and a pseudo-halftone processing circuit (12), the input multivalued image information is delayed by a delay circuit (10) by the time required for processing in the image area determination circuit (14) and then provided to the pseudo halftone processing circuit (12). Binarization processing circuit. 3. In claims 1 and 2, the image area determination circuit (14):
A binarization processing circuit for image information, characterized in that edge portions are detected by Laplacian calculation. 4. In claims 1 and 2, the image area determination circuit (14):
A binarization processing circuit for image information, characterized in that edge portions are detected by gradient calculation. 5. In claims 1 and 2, the image area determination circuit (14):
A binarization processing circuit for image information, characterized in that edge portions are detected by Sobel calculation.