JPH04220080A - Binarization method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention relates to a method for converting a multi-valued image into a binary image, for example, an adaptive binarization method that distinguishes thin line parts such as characters in an original image from image parts, and performs processing suitable for each. It is about the method.

0002

[Prior Art] A conventional adaptive binarization method uses nxm of an original image.
The image area of the original image was determined based on the size of the difference between the target pixel and the pixel average.

0003

[Problems to be Solved by the Invention] However, in the conventional method, noise-like signals are easily collected, which is a major cause of erroneous image area determination. As a result, erroneous determination portions called white spots and black spots occur in the binary image, which greatly deteriorates the image quality.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a binarization method using an image area discrimination method that reliably determines character portions in a multilevel image and does not emphasize noise-induced portions. do.

[0005]

[Means and effects for solving the problems] In order to solve the above problems, the binarization method of the present invention divides an original image into NxM windows, and uses a line drawing detection pattern prepared in the NxM window and a line drawing detection pattern prepared in the NxM window. The method is characterized in that the average image data is compared, the line drawing area of the original image is discriminated, and the binarization threshold value is adaptively changed according to the line drawing-likeness using the discrimination result.

[0006]

[Embodiment] In order to solve the above problem, the embodiment of the present invention described below calculates the continuous amount centering on the point of interest since the character part basically has a certain continuity, and the continuous amount is A method is used to determine thin line image areas such as characters and symbols based on the size. Even if there is only one black or white point due to noise, the likelihood of a thin line portion decreases when compared with data that matches a continuous amount. This makes it possible to clearly distinguish thin line parts such as continuous characters compared to conventional methods.

FIG. 1 shows a window for detecting thin line portions. Here, one embodiment of the present invention will be described using a 3x3 window as an example.

FIG. 2 shows an example of a pattern for detecting thin line portions. Here, it is assumed that the point of interest is at position a22.

FIG. 3 shows the relationship between the original image, the window, and the point of interest. 302 is a diagram in which the original image is divided into pixels, and 301 is a 3×3 window. 303 indicates the pixel position of interest.

FIG. 4 shows an image area discrimination circuit when a 3×3 window is used and the line detection pattern of FIG. 2 is used.

Reference numeral 401 denotes a photoelectric conversion element, such as a CCD. 402 is an amplifier, and 418 is for digitizing the analog signal. Therefore, the data after 418 becomes, for example, 8-bit data and is output. 403-405
form a 3-line buffer memory of 8 bits each.

Therefore, the data from 401 are A1, A2,
A3 . The number of CCD elements in 401 and 403 to 405 are different because they are calculated in 3x3 (2 different), and the original pixel is A/D converted 1.
Line drawing detection is obtained with a clock delay. 406 to 409 are adders, 406 is a part that calculates (A) in FIG.
7 is a circuit that calculates (B), 408 (C), and 409 (D). 415 is a circuit for calculating the sum of image data within a 3x3 window, and 417 is a circuit for subtracting data three times the value of the pixel of interest since the number of pixel points of interest is four times as much with four inputs of 415.

Reference numeral 419 is for converting the data values of 9 pixels of 3×3 into 3 data values.

410 to 413 calculate the difference between each pattern component and the window average value, and equivalently obtain the result of calculating the line drawing component. 414 is a circuit that obtains the maximum absolute value of each component (A) to (D), and 416 outputs the final line drawing amount.

As can be seen from the above description, in the case of a noise signal existing as an isolated point, the energy thereof is reduced to 1/3 in the case of 3×3, and erroneous judgments due to noise are reduced. As the window size increases to 5x5 and 7x7, it becomes less susceptible to noise.

FIG. 5 shows a method for adaptively binarizing a multivalued image using the line drawing detection method described above.

501 is the above-mentioned CCD, 502 is an amplifier,
503 is A that converts analog data to digital data
/D converter 511 is a line drawing detection circuit section shown in FIG. The output 416 in FIG. 4 is compared with the data set by the setters S1, S2, and S3 to control the changeover switch 504. When it is determined that the 511 output is most likely to be a line drawing (
When it is in the middle of the character image, it is switched to the position (b), and when it seems to be an image area, it is switched to the position (c).

505 is a single threshold comparison circuit; for example, if the input data is data in the range of 8 bits 0 to 255, 12
8 is the single threshold.

Reference numeral 508 denotes a concentration-preserving binarization method circuit section that includes a so-called error diffusion method, and has a configuration as explained in FIG. Reference numerals 507 and 509 are threshold tables, in which the threshold matrix is increased in clear image areas, and decreased in areas where it is difficult to distinguish between line drawings and image areas. As a result, when the matrix size is large, the dot arrangement of the concentration-preserving binarization method approaches a periodic structure, and when the matrix size is small, the periodic structure is broken, thereby increasing randomness and improving the reproduction of fine line parts.

The data thus binarized is output to a binary printer 510 and recorded.

[0021] The values at 507 and 509 are not threshold values, but
This shows the order of threshold values.

FIG. 6 shows a binarization circuit, and 615 is a multi-value data input section. 610 is the input multivalued data and the error diffusion accumulation table which will be explained later, and the x mark position is the pixel data of interest and is 6.
Add by 10. The added data is binarized in 608. The threshold value consists of a value drawn from the 616 or 617 table. 616 and 617 are X-axis and Y-axis counters,
The values of the cells indicated by the counters 604, 603, 605, and 606 are selected. 603 and 604 are counted by the scanner's main scanning direction clock 601;
05 and 606 are counted by the sub-scanning direction clock 602. The switch 607 is the switch selected according to FIG. As can be seen from FIG. 6, even if the size of the threshold table changes, error diffusion is preserved, making it possible to reproduce density faithful to the original image. In the table 612, an x mark indicates the pixel position of interest, and a single mark indicates a position that has already been binarized. 613 is a line memory that stores the cumulative value of errors.

(Other Embodiments) In the above embodiments, the line drawing detection method using a 3x3 pattern, a single fixed threshold, a 3 x 3 matrix threshold, and a 5 x 5 matrix threshold were basically explained, but line drawing detection can be performed using any pattern using NXM pixels. It is within the scope of the present invention that the threshold value matrix may also be composed of a PxQ threshold value matrix.

It is clear that even if the line drawing detection pattern in FIG. 2 is reversed, with the background being white and the line being solid, it can be handled by preparing a similar pattern.

As described above, according to the above embodiment of the present invention, the following effects are produced.

(1) Line drawings such as characters can be detected stably, and good line drawings can be reproduced with a single threshold value.

(2) Misjudgment of line drawings due to noise is extremely reduced, and image quality deterioration is eliminated. (3) Image areas close to line drawings, that is, image areas with high frequency components, are binarized with a small threshold matrix size, so the binarized data array has randomness, making it suitable for high-resolution recording.

(4) Since the areas detected as image areas are binarized using a large threshold matrix, the binarized data array becomes periodic and becomes a high-gradation recording (not a Bayer array but a spiral array). Because of the arrangement, the overlap of the dots is uniform).

(5) As a result of the above, the binarized data is arranged relatively randomly in line drawings or parts close to line drawings, but in photographic image parts, because they are arranged in a periodic structure, they cannot be binarized using this method. Similar to the systematic dither method, this method has a better compression effect than the error diffusion method. Moreover, the image quality is better than that of the organized dither method because the line drawing portion is adaptively processed.

[0030]

[Effects of the Invention] As explained above, according to the present invention, an image area discrimination method is used that reliably determines character parts in a multivalued image and does not emphasize noise-like parts.
A value conversion method can be provided.

[Brief explanation of the drawing]

[Figure 1] Window for detecting thin line parts

[Figure 2] Pattern for detecting thin line parts

[Figure 3] Diagram showing the relationship between the original picture, window, and points of interest

[
Figure 4: Image area discrimination circuit

[Figure 5] Diagram showing the binarization method of the present invention

[Figure 6] Binarization circuit diagram of the present invention

[Explanation of symbols]

401 Photoelectric conversion element 402 Multiplier

Claims (1)

[Claims] [Claim 1] Divide the original picture into NxM windows,
Line drawing detection pattern prepared in M window and NxM
The average image data within the window is compared, the line drawing area of the original image is determined, and the discrimination result is used to determine the line drawing area of the original image.
A binarization method characterized by adaptively changing a digitization threshold.