JPH09311936A - Image binarization method and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a correct binary image where a detected image can be correctly discriminated from its background by calculating the occurrence frequency of each luminance of pixels forming a two-dimensional image and binarizing this image with the occurrence frequency decided when the addition value of those occurrence frequency is equal to a prescribed level defined as the threshold. SOLUTION: A frequency table generation part 6 calculates the occurrence frequency N of each luminance I of pixels forming a two-dimensional image and produces a frequency table 12. A frequency rearrangement part 13 rearranges the frequency N contained in the table 12 in the order of smaller value of them and generates a frequency order table 14. A threshold calculation part 15 calculates the threshold NHJ of occurrence frequency based on each frequency order J set on the table 14 and the occurrence frequency N1 of the order J. A binarization processing part 16 binarized the two-dimensional image stored in an image memory 5 into a binary image 18 based on the threshold NHJ of the frequency N calculated at the part 15. Then a display part 17 displays and outputs the image 18.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image binarization method and an image binarization apparatus for binarizing a two-dimensional image.

[0002]

2. Description of the Related Art For example, in a system for recognizing characters written on a label, recognizing the shape of an object, or detecting that a moving object has passed a predetermined place,
A two-dimensional image of a predetermined area including an image of a detection target captured by a camera is binarized, and various image processing such as shape determination is performed.

In many cases, in a two-dimensional image which is the image of such image processing, the detected image area is smaller than a predetermined ratio as compared with the background area. For example, when characters are drawn in different colors on the same color label, the area of the characters is 20% to 30% or less of the area of the background. In addition, for example, when reading the vehicle number on the license plate of a running vehicle, in the case of a passenger car, green characters are described against a white background, but this green character corresponds to a white background. The area ratio is 10% to 20%. further,
When detecting a vehicle passing through a road, the vehicle travels with a certain vehicle-to-vehicle spacing, so if the background is a fence, sky, or road surface, for example, the area of the detected image (vehicle) is compared to the area of the background. And 20% to 30%.

As described above, the p-tile method has been put into practical use as a method for binarizing a two-dimensional image in which the detected image area is less than a predetermined ratio compared with the background area ("Introduction to Computer Image Processing" Tamura. Hideyuki supervised by Soken Publishing).

FIG. 7 is a block diagram showing a schematic configuration of an image binarization apparatus adopting the p-tile method. For example, a two-dimensional image 1 drawn on paper is read by a camera 2, converted into a digital two-dimensional image, and temporarily written in an image buffer 3. The two-dimensional image written in the image buffer 3 is cut out by the next image cutting-out unit 4 into a two-dimensional image having a predetermined area (a predetermined number of pixels) including the detected image and the background and written in the image memory 5. Be done. Each pixel P forming a two-dimensional image stored in the image memory 5 has its own xy coordinate and, for example, 4 or 8
It has a brightness I of a predetermined bit configuration such as bits. Therefore, each pixel is written as P (x, y, I).

The frequency table creating section 6 creates the frequency table by calculating the occurrence frequency of each brightness I in each pixel P (x, y, I) forming the two-dimensional image. This frequency table can also be graphed as shown in FIG.

The binarization processing unit 7 binarizes the two-dimensional image using the occurrence frequency N _I of each luminance I set in the frequency table and displays the binarized image on the display unit 8. FIG. 9 is a flow chart showing a specific binarization processing procedure of the binarization processing unit 7.

The flow chart starts, and the rank K of the brightness I is initialized to 1 at P1. The luminance order K is assumed to be arranged in descending order from the highest luminance. That is, 1 in the brightness rank K indicates the highest brightness. For example, when the brightness is in the range of 0 to 256, the brightness rank K = 1 indicates the brightness 256 and the brightness rank K = 2 indicates the brightness 255. Also, P1
The area S indicated by the number of pixels is initialized to zero.

At P2, the frequency N _K of the brightness I of the brightness rank _K is added to the area S (S = S + N _K ). However, each pixel P
Regarding the frequency of (x, y, I), it is assumed that frequency 1 has one unit area. The area S after the addition in P3 is predetermined 2
If the area S _H corresponding to, for example, 20% of the three-dimensional image area has not been reached, 1 is added to the brightness rank K in P4, and P2 is added.
Return to.

When the area S after addition reaches the predetermined area S _H at P3, the brightness I corresponding to the brightness rank K at this point is set as the threshold value I _HK (P5). Then, each pixel P (x, y, of the two-dimensional image stored in the image memory 5
The brightness I of I) is binarized by the threshold value I _HK (P6). Then, a binarized image is created (P7), and the created binarized image is displayed and output on the display unit 8 (P8).

[0011]

However, the image binarization method and the image binarization apparatus using the p-tile method described above still have the following problems to be solved. For example, when the two-dimensional image 1 drawn on the paper shown in FIG. 10 (d) is not illuminated with light and is entirely dark, the captured two-dimensional image 10 (base) is shown in FIG. 10 (a). As shown in FIG. 1, when the brightness of the part of the detected image 10a is 4 and the brightness of the part of the background 10b is 1,
As shown in the luminance frequency characteristic of 1 (a), since between the brightness frequency a ₂ of brightness frequency a ₁ and the detection image 10a portion of the background 10b part there is a large difference, the threshold value for binarization I
_HKa exists between the brightness 1 and the brightness 4, and the two-dimensional image 10 (base) shown in FIG. 10D is correctly binarized.

If the two-dimensional image 1 drawn on the paper is illuminated with light and the entire image is bright, the photographed two-dimensional image 10 (base) is as shown in FIG. 10 (b). Detection image 1
When the luminance of the portion of 0a is 8 and the luminance of the portion of the background 10b is 5, as shown in the luminance frequency characteristic of FIG. 11B, the luminance frequency b ₁ of the background 10b portion and the detected image 10a Since there is a large difference between the brightness frequency b _{2 of a} part and the threshold value I _HKb for binarization, brightness 5 and brightness 8 are set.
And the two-dimensional image 10 (b) shown in FIG.
ase) is correctly binarized.

However, if the two-dimensional image 1 drawn on the paper has a portion irradiated with light and a shaded portion, the photographed two-dimensional image 10 (base) is as shown in FIG. 10 (c). In addition, even in the same background 10b part, the part where the brightness is 1 and the brightness is 5
And the same detection image 10a also has a portion having a luminance of 4 and a portion having a luminance of 8. In this case, one of the luminances 4 of the detected image 10a is lower (darker) than the other luminance 5 of the background 10b, and the reverse development occurs.

Therefore, in this case, as shown in the luminance frequency characteristic of FIG. 11C, there are four luminance frequencies c _{1 to} c ₄ . The threshold I _HKc for binarization exists between the luminance 5 and the luminance 8. That is, since one luminance 4 of the detection image 10a exists below the threshold value I _HKc , one portion of this detection image 10a becomes the same as the background, and
An erroneous binarized image 11 is obtained as shown in FIG.

Further, as shown in FIG. 13A, when the brightness range is 0 to 99 and the brightness of the detected image in the two-dimensional image 10 (base) is changing in a narrow range of 0 to 9. In this two-dimensional image 10 (base), [0] and [99]
When binarized with, the threshold value becomes 5, and as shown in FIG. 13 (b), the brightness of 1 to 4 becomes the background 10b, and the detected image 1
0a cannot be detected accurately.

The present invention has been made in view of the above circumstances, and by setting a threshold value based on the frequency of occurrence of luminance, even if the two-dimensional image is not uniformly illuminated with light. Provided are an image binarization method and an image binarization device that can obtain a correct binarized image that can correctly distinguish the detected image from the background even if there is a slight variation in the brightness of the detected image. The purpose is to do.

[0017]

In order to solve the above problems, in the image binarization method of the present invention, each pixel forming a two-dimensional image in which the detected image area is a predetermined ratio or less compared to the background area. Then, the frequency of occurrence of each luminance in is calculated, and the frequency of occurrence of each luminance obtained is added in ascending order of the frequency of occurrence.
The occurrence frequency when the added value reaches a predetermined value is set as a threshold value, and the two-dimensional image is binarized by this threshold value.

Further, in the image binarizing apparatus of the present invention, the occurrence frequency is calculated to find the occurrence frequency of each luminance in each pixel forming a two-dimensional image in which the detected image area is less than a predetermined ratio compared with the background area. Means, a threshold value calculating means for adding the occurrence frequencies of the respective luminances obtained by the occurrence frequency calculating means in ascending order of the occurrence frequency, and a threshold value for the occurrence frequency when the added value reaches a predetermined value, and a threshold value calculation And a binarizing means for binarizing the two-dimensional image with the threshold value obtained by the means.

The operation principle of the image binarization method and the image binarization apparatus thus configured will be described. In a two-dimensional image in which the detected image area is equal to or less than a predetermined ratio as compared with the background area, the frequency of occurrence of luminance in the background portion is significantly smaller than the frequency of occurrence of luminance in the detected image portion. In this case, even if the two-dimensional image is not uniformly irradiated with light and the luminance of the background portion is divided into a plurality of portions, the frequency of each of the plurality of divided luminances is equal to the frequency of occurrence of the luminance of the detected image portion. Big enough in comparison.

Further, when the luminance of the background portion is divided into a plurality of pieces and the luminance of the detected image portion is also divided into a plurality of pieces,
Each frequency of each luminance of the detection image portion divided into a plurality is sufficiently smaller than each frequency of each luminance of the background portion divided into a plurality.

Therefore, in the present invention, the frequency of occurrence of each luminance in each pixel forming a two-dimensional image is sequentially added in ascending order, and the frequency of occurrence at the time when the added value reaches a predetermined value is set as a threshold value. , The two-dimensional image is binarized with this threshold value.

That is, the pixel corresponding to each luminance of each occurrence frequency below this threshold is set to one of the two values, and the pixel corresponding to each luminance of each occurrence frequency exceeding this threshold is set to the other of the two values. By doing so, it is possible to reliably distinguish the detected image portion including the luminance having the low occurrence frequency and the background portion including the luminance having the high occurrence frequency.

[0023]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the schematic arrangement of an image binarization apparatus that employs the image binarization method according to the embodiment. The same parts as those of the conventional image binarization apparatus shown in FIG. 7 are designated by the same reference numerals. Therefore, detailed description of the overlapping portions is omitted.

The two-dimensional image 1 is read by the camera 2, converted into a digital two-dimensional image, and temporarily written in the image buffer 3. The two-dimensional image written in the image buffer 3 is cut out by the next image cutting-out unit 4 into a two-dimensional image having a predetermined area (predetermined number of pixels) including the detected image and the background, as shown in FIG. It is written in the image memory 5 shown. Each pixel forming the two-dimensional image stored in the image memory 5 is represented by a pixel P (x, y, I) with coordinates (x, y) and brightness I as shown in FIG.

The frequency table creating unit 6 calculates the occurrence frequency N of each brightness I in each pixel P (x, y, I) forming the two-dimensional image, and the frequency table 12 shown in FIG. 2 (b).
Create Therefore, in the frequency table 12, the occurrence frequency N (N = N ₁ , N ₂ , N ₃ , ..., N _m ) indicated by the number of pixels for each luminance I (I = ₁ , ₂ , ₃ , ..., M _). ) Is set.

The frequency-order rearranging unit 13 includes a frequency table 12
The respective occurrence frequencies N are rearranged in the ascending order of the occurrence frequency N to create the frequency order table 14 shown in FIG. Therefore, in the frequency order table 14, each occurrence frequency rank J (J = 1, 2,
, ..., M), the occurrence frequencies N _J1 , N _J2 , N _J3 , ..., N _Jm corresponding to the corresponding frequency rank J are set.

The threshold value calculation unit 15 calculates the threshold value N _{HJ of the} occurrence frequency based on each frequency rank J set in the frequency order table 14 and the occurrence frequency N _J of each frequency rank J. The binarization processing unit 16 binarizes the two-dimensional image stored in the image memory 5 using the threshold value N _HJ of the occurrence frequency N calculated by the threshold value calculation unit 15, and the binarized image shown in FIG. Create 18. The display unit 17 displays and outputs the binarized image 18 created by the binarization processing unit 16.

FIG. 3 shows the above-described frequency table creating unit 6, frequency order rearranging unit 13, threshold value calculating unit 15, and binarization processing unit 16.
9 is a flowchart showing a specific processing operation of the display unit 17.
The flow chart starts, and in Q1, the frequency N of occurrence of each brightness I in each pixel P (x, y, I) of the two-dimensional image stored in the image memory 5 is calculated to create the frequency table 12. Next, at Q2, each occurrence frequency N of the frequency table 12
Are sorted in ascending order of occurrence frequency and the frequency order table 14
Create

At Q3, the occurrence frequency rank J is initialized to 1. It should be noted that the occurrence frequency rank J is an order in which the luminance occurrence frequency N is low. That is, 1 of the occurrence frequency rank J indicates the lowest occurrence frequency N _J1 . Furthermore, the area S indicated by the number of pixels is initialized to 0 in Q3.

At Q4, the luminance occurrence frequency N _J of the occurrence frequency rank _J is added to the area S (S = S + N _J ). However, it is assumed that the frequency of each pixel P (x, y, I) is frequency 1 having one unit area. The area S _HJ corresponding to, for example, 20% of the predetermined two-dimensional image area S after addition in Q5
If it has not reached, the occurrence frequency rank J is incremented by 1 in Q6, and the process returns to Q4.

When the area S after addition reaches a predetermined area S _HJ in Q5, a threshold value for binarizing the occurrence frequency N _J corresponding to the occurrence frequency rank J at this point in time for binarizing the two-dimensional image. N _HJ (Q7). Then, each pixel P (x, y, I) of the two-dimensional image stored in the image memory 5 is binarized by the threshold value _NHHJ of the occurrence frequency (Q8). That is, the calculated threshold value N _{HJ of} each pixel P (x, y, I)
The brightness I of each occurrence frequency N below is set to the [1], [H], or [black] level indicating the detected image of the binary values, and each brightness I of each occurrence frequency N exceeding this threshold value _NHJ. Is set to the [0], [L], or [white] level indicating the background of the two values.

In this way, the threshold value N for the occurrence frequency is set.
_By binarizing each pixel P (x, y, I) with _HJ , it is stored in the image memory 5 and the binarized image 18 is converted from the two-dimensional image.
(Q9), and the created binarized image 18 shown in FIG.
Is output to the display unit 8.

In the image binarizing device thus constructed, the occurrence frequency N of each luminance I in each pixel P (x, y, I) forming the two-dimensional image 10 photographed by the camera 2 is obtained. Then, the frequency table 12 is created. Then, the occurrence frequency N of each luminance I set in the frequency table 12 is sorted in the ascending order of the occurrence frequency and registered in the frequency order table 14.

Then, the occurrence frequencies N registered in the frequency order table 14 are sequentially added in the ascending order of the occurrence frequencies, and the added value S becomes, for example, 20% determined by the ratio of the detected image area to the background area. The occurrence frequency N _J at the time when the predetermined value set on the basis is reached is set as the threshold value N _HJ . As can be understood from the brightness occurrence frequency characteristic shown in FIG. 5, each brightness I of each occurrence frequency N equal to or lower than the threshold value N _HJ is set to one of two values (black), and each occurrence frequency N exceeding this threshold value N _HJ. The respective brightnesses I of are set to the other (white) of the two values.

That is, the two-dimensional image 10 (base) is not binarized by the luminance value, but the luminance frequency N is 2
By digitizing, the detected image portion including the luminance I with the low occurrence frequency N and the background portion including the luminance with the high occurrence frequency N can be reliably distinguished.

For example, in the luminance occurrence frequency characteristic of the two-dimensional image 10 shown in FIG. 10C, in which the two-dimensional image 1 drawn on paper has a portion irradiated with light and a shaded portion,
As shown in FIG. 11C, there are four luminance frequencies c _{1 to} c ₄ .

In the embodiment, the threshold value N _HJ of the luminance occurrence frequency N is c ₁ , c ₃ and c ₂ . Since it exists between c ₄ and c ₄ , the occurrence frequency corresponding to the detected image 10a is the threshold value N _HJ
Each pixel P (x, y, I) having the following luminance 4.8 is at the [black] level, and each pixel P (having luminance 1, 5 whose occurrence frequency corresponding to the background exceeds the threshold value N _HJ x,
Both y and I) are at the [white] level. Therefore, the correct binary image 18 shown in FIG. 4 is obtained.

The two-dimensional image 1 shown in FIG.
Even when no light is radiated on the background and the whole is dark, the luminance frequency a ₁ of the background portion and the luminance frequency a of the detected image portion a
_Since there is a large difference between ₂ and ₂ , the threshold value N _HJ is located between them, and the detected image and the background can be clearly distinguished.

Similarly, when the two-dimensional image 1 shown in FIG. 10 (b) is illuminated with light and the whole is bright, the luminance frequency b ₁ of the background portion and the luminance frequency b of the detected image portion b
_Since there is a large difference between ₂ and ₂ , the threshold value N _HJ is located between them, and the detected image and the background can be clearly distinguished.

That is, the two-dimensional image 10 is correctly binarized under any of the light irradiation conditions of FIGS. 10A, 10B, and 10C. Further, as shown in FIG. 13A, the range of the brightness I is 0 to 99, and each pixel P of the detected image is
-Dimensional image 1 whose brightness changes in a narrow range of 0-9
Even if it is 0, if the threshold value _HJ of the occurrence frequency N in the embodiment is binarized, the pixel P having the luminance 1, 2, 3.4 has a low occurrence frequency. It is classified into [99] on the same side as the luminances 5, 6, 7, 8, and 9 of the image portion. Therefore, as shown in FIG. 6, a correct binarized image 18 is obtained.

[0041]

As described above, in the image binarization method and the image binarization apparatus of the present invention, the occurrence frequency of each luminance in each pixel forming a two-dimensional image is obtained, and this occurrence frequency is reduced. The two-dimensional images are binarized by sequentially adding and using the occurrence frequency when the added value reaches a predetermined value as a threshold.

Therefore, even if the two-dimensional image is not uniformly illuminated with light, and even if there is a slight variation in the brightness of the detected image, the detected image can be correctly distinguished from the background. A binary image can be obtained.

[Brief description of drawings]

FIG. 1 is an image 2 to which an image binarization method of the present invention has been applied.
Block diagram showing the schematic configuration of the quantizer

FIG. 2 is an image memory incorporated in the image binarization apparatus,
The figure which shows each storage content of a frequency table and a frequency order table

FIG. 3 is a flowchart showing a binarizing operation of the image binarizing apparatus.

FIG. 4 is a diagram showing a binarized image obtained by the image binarization apparatus.

FIG. 5 is a luminance frequency characteristic diagram for explaining the operation of the image binarization apparatus.

FIG. 6 is a diagram showing a binarized image obtained by the image binarization apparatus.

FIG. 7 is a block diagram showing a schematic configuration of an image binarization apparatus to which a conventional image binarization method is applied.

FIG. 8 is a luminance frequency characteristic diagram for explaining the operation of the conventional image binarization apparatus.

FIG. 9 is a flowchart showing a binarizing operation of the conventional image binarizing apparatus.

FIG. 10 is a diagram showing a plurality of two-dimensional images with different light irradiation conditions for the binarized image.

FIG. 11 is a luminance frequency characteristic diagram corresponding to each two-dimensional image in FIG.

FIG. 12 is a diagram showing a binarized image obtained by the conventional image binarization apparatus.

FIG. 13 is a diagram showing a two-dimensional image and a binarized image for explaining problems of the conventional image binarization apparatus.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Image 2 ... Camera 3 ... Image buffer 4 ... Image cut-out part 5 ... Image memory 6 ... Frequency table creation part 10 ... Two-dimensional image 11, 18 ... Binary image 12 ... Frequency table 13 ... Frequency order rearranging part 14 ... Frequency order table 15 ... Threshold value calculation unit 16 ... Binarization processing unit 17 ... Display unit

Claims (2)

[Claims] 1. The frequency of occurrence of each luminance in each pixel forming a two-dimensional image in which the detected image area is less than a predetermined ratio compared to the background area is determined, and the frequency of occurrence of each luminance thus determined is low in frequency of occurrence. An image binarization method in which the two-dimensional images are binarized by using the threshold as the occurrence frequency when the addition value reaches a predetermined value by sequentially performing addition. 2. An occurrence frequency calculation means (6) for obtaining an occurrence frequency of each luminance in each pixel forming a two-dimensional image in which a detected image area is smaller than a predetermined ratio in comparison with a background area, and the occurrence frequency calculation means. Threshold value calculation means (15) that adds the occurrence frequency of each luminance obtained in step 1 in ascending order of occurrence frequency, and sets the occurrence frequency when the added value reaches a predetermined value as a threshold value (15)
An image binarization device comprising: and a binarization unit (16) for binarizing the two-dimensional image with the threshold value obtained by the threshold value calculation unit.