JPS63220375A - Binarization processing method for image - Google Patents

Abstract

PURPOSE:To accurately extract an object from a background by finding out a mean value of luminance and its standard deviation and finding out a threshold based on an operation formula by using the obtained values and a parameter. CONSTITUTION:The mean value X of the luminance of an image and its standard deviation alpha are found based on image date in an image memory, the threshold Th is determined by using the obtained values X, sigma and a previously set parameter K based on the operation formula of X+Ksigma and the image data read out from the image memory are binarized on the basis of the threshold Th. Even when the background luminance of an image is changed due to the change of illuminating conditions or a measuring object, when there is no object, or when the area of an object is unclear, the threshold is automatically set up to a proper value and the extracting accuracy of the object from the background can be improved.

Description

[Detailed Description of the Invention] (Industrial Application Field) This invention relates to an image binarization processing method used when extracting an object in an image from the background, and in particular, relates to an image binarization processing method that automatically sets a binarization threshold. This invention relates to an improvement of an image binarization processing method that allows settings to be made.

(Prior art and its problems) The 21a processing method has traditionally been widely used as a method for extracting the object in an image from the background, but as a method for setting a threshold in such a binarization processing method, teeth,
There are two main types: fixed type and automatic type. In the fixed type, a predetermined threshold value is determined in advance, and image data is binarized based on this threshold value to distinguish between the object, the back, and the other. However, in the case of a fixed type, the threshold value is always kept constant, so if the illumination conditions change or the degree of backfill of the image changes due to changes in the sample being measured, etc., the object will disappear from the background. There was a problem in that the extraction accuracy decreased.

On the other hand, the automatic method takes into account the above problems, and even if the background brightness of the image changes due to changes in lighting conditions or the measurement target, the optimal threshold value is automatically set according to these changes. This method uses the mode method and P-
Various methods such as the tile method have been proposed. Among these methods, the mode method is a method in which the brightness distribution of an image is determined and a threshold value is determined between two peaks (the bottom of a valley) corresponding to the object and the background. However, this method has the problem that when there is no target object in the image, if there is density unevenness in the background, the density unevenness may be mistaken for the target object and an incorrect threshold value may be set. was. Further, the P-tile method is a method in which a brightness distribution is determined in the same manner as the mode method, and a brightness such that the proportion of points having the brightness or higher becomes a predetermined value is determined as a threshold value. however,
This method requires that the area of the desired object be known in advance, and cannot be used if the area of the object is unknown.

(Purpose of the Invention) This invention has been made to solve the above problem, and is applicable when the background brightness of an image changes due to changes in illumination conditions or the measurement target, when the target does not exist, or when the target It is an object of the present invention to provide an image binarization processing method that can automatically set a threshold value to an appropriate value even when the area is unknown, thereby improving the accuracy of extracting an object from the background.

(Means for Achieving the Object) In order to achieve the above object, the image binarization processing method of the present invention calculates the average value X and standard deviation σ of the brightness of the image based on the image data in the image memory, The present invention is characterized in that a threshold value is determined by the arithmetic expression of X+kσ using a preset parameter k, and the image data read from the image memory is binarized based on this threshold value.

(Operation) FIG. 1 shows a schematic flow of the present invention. Input image data is converted into a digital value corresponding to its brightness and stored in the image memory. A binarization threshold Th for converting this into binarized image data is determined as follows.

■Calculation of average value X and standard deviation σ of brightness (Step S1 in Figure 1): From the brightness level x of N pixels of the entire screen or a partially extracted screen, (i-1...N) , the average value X and standard deviation σ are determined by the following equations.

(2) The number of pixels N used in this calculation is set to be large enough to sufficiently represent the luminance distribution of the background.

■Calculation of binarization threshold Th (step S2 in Figure 1)
: Binarization threshold T using preset parameter k
Find h using the following formula.

Th-X+kσ (3) In this case, the value of k changes its sign depending on the luminance relationship between the background and the detected object (target object). That is, as shown in the density histogram of Fig. 2, when the luminance of the detected object a is higher than the luminance of the background, k>0, and conversely, as shown in the density histogram of Fig. 3, when the luminance of the detected object a is higher than the luminance of the background. If the brightness is lower than the brightness of the background, k<O. Further, the optimum value of k depends on the contrast between the background and the detected object, that is, when the contrast is large, the value of k is set large, and conversely, when the contrast is small, the value of .degree. is set small. By determining k as above, the detected object a
And back! A threshold value T between the two peak values corresponding to I6
h can be determined accurately.

(2) Image processing (step S3 in FIG. 1): The image data read from the image memory is subjected to binarization processing using the binarization threshold Th determined by the above equation (3). This allows the detected object to be extracted from the background with high accuracy.

(Example) An example in which the binarization processing method of the present invention is applied to a foreign matter detection device in a plastic sheet will be described based on FIG. 4 □. As shown in the figure, sample 1 is
It is mounted on Y stage 2. The sample image magnified by the microscope 3 is captured by the TV camera 4 and stored in the video memory 5 as a digital image.

Computer 6 reads image data from video memory 5 and performs binarization processing according to the flowchart shown in FIG. Specifically, based on the image data of all pixels in the screen extracted by the TV camera 4, the above equation (1) is calculated to obtain the average luminance value X, and the above (2)
Compute the formula to find the standard deviation σ (step 81)
. Next, by calculating the above equation (3), the binarization threshold T
Determine h (step 82). At this time, the parameters are set from the keyboard of the computer 6.

Then, using this 21I conversion threshold T, the image data corresponding to each pixel read out from the video memory 5 is sequentially binarized to separate the background and the foreign object (detected object) (step S3). . Once the foreign object is extracted, the area, shape, etc. of the foreign object can be easily determined based on the number of pixels, coordinate values, etc. of the region that is the foreign object (detected object).

In this way, when the image processing of the first screen captured by the TV camera 4 is completed, the computer 6 issues a movement command to the XY stage controller 7, moves the automatic XY stage 2, and processes the next screen in the same manner as above. Inspect.

In this way, the average luminance value X and standard deviation σ are determined, and the parameter k is used to determine the threshold value Th from the calculation of equation (3) above. Therefore, the value of k can be adjusted appropriately. By setting the threshold Th, it is possible to accurately determine the threshold Th between the two peak values corresponding to the detected object a and the background as shown in Figs. 2 and 3, and to accurately distinguish the detected object from the background. It becomes possible to extract. Moreover, this threshold llTh can be determined without any problem even when the background brightness of the image changes due to changes in illumination conditions or the object to be measured, when there is no detection object, or when the area of the detection object is unknown.

It has already been mentioned that the preferable value of k depends on the contrast between the background and the detected object, but the specific value of k is -5 to -5 when the contrast and last are large.
10 is desirable, and -2 if the contrast is small.
~7 was appropriate.

FIG. 5 is an example in which the parameter k is measured by the size and area of a foreign object. In this case, the microscope 3 is set to dark-field illumination, so the background is dark and the foreign object is bright (corresponding to Figure 2). Therefore, the value of is set as a positive value.As can be seen from Figure 5, The background and foreign matter are accurately distinguished in the range <k8.

In the above embodiment, the sample image is divided into a plurality of regions and extracted, and the threshold value Th is determined by the above equation (3) every time the sample image is extracted, and the binarization process is performed.
When the background brightness of sample 1 is almost the same, the above threshold With obtained for the image of the first divided area is
It may also be used as a threshold value for images of divided areas after the th.

In this case, the time required to calculate the threshold values for the images of the second and subsequent partitioned areas can be omitted, and the image processing time can be shortened.

In addition, if the height of the height within the sample changes gradually, the threshold value T of the above equation (3) for each divided area
It is desirable to obtain h once and then further weight it (ie, smooth it) with the threshold value of the adjacent partitioned area and use it as the actual binarization threshold value. The smoothing method is, for example, based on the following equation.

Th (i) = α・Th (i) + (1-α)・T, (i-'+) ... (4) Here, Th (i): M + screen threshold α: smoothing The coefficient is (0 × α × 1).

By the way, the binarization processing method of the present invention is particularly effective when the detected object is sufficiently small compared to the background. This is because when the detected object becomes smaller, the average value X and standard deviation σ of the brightness determined by equations (1) and (2) above become almost equal to those values (average value and standard deviation) for the background area. It is. However, if the smoothing method shown in equation (4) above is used, a good threshold value can be determined even when the detected object is larger than the background.

In addition to detecting foreign objects in plastic sheets, this invention
It can be applied to detect foreign objects in glass containers and liquids, and can also be used to recognize marks and small characters on electronic components and printed circuit boards.

(Effects of the Invention) As described above, according to the image binarization processing method of the present invention, the average luminance value Since Th is calculated, even if the background brightness of the image changes due to changes in illumination conditions or the measurement target, if the target does not exist, or if the area of the target is unknown,
The effect is that the threshold value is automatically set to an appropriate value, and the target object can be extracted from the background with high accuracy.

[Brief Description of the Drawings] Fig. 1 is a schematic flow diagram of the present invention, Fig. 2 is a luminance distribution diagram when the detected object is brighter than the background, and Fig. 3 is a luminance distribution diagram when the detected object is darker than the background. 4 is a schematic configuration diagram of a foreign object detection device to which an embodiment of the present invention is applied, and FIG. 5 is a diagram showing the relationship between parameters and detected object area (measured value) in the above device. X... Average value of luminance σ... Standard deviation k... Parameter Th... Threshold

Claims (3)

[Claims] (1) In an image data processing method that stores input image data in an image memory, reads out the image data from the image memory, and performs image processing, the average value of image brightness @X@ is determined based on the image data in the image memory. Determine the standard deviation σ, use the preset parameter k to determine a threshold value using the formula @X@+kσ, and convert the image data read from the image memory into a binary value based on this threshold value. An image binarization processing method characterized by carrying out a conversion process. (2) When processing an image partitioned into multiple regions consecutively in the order of the partitions, the threshold value obtained for the image of the first partitioned area is used as the threshold value for the images of the second and subsequent partitioned areas. An image binarization processing method according to claim 1, wherein the image binarization processing method is also used. (3) When continuously processing an image partitioned into multiple areas in the order of partitions, the threshold value obtained for the image of the i-th (i is an integer) partition area is The weighted value of the threshold value for the image of the th partitioned area is
The image binarization processing method according to claim 1, wherein the image binarization processing method is used as a threshold value for the image of the i-th partitioned area.