JP2006155579A - Image processing method and image processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an excellent low-pass filter which has the same characteristics in a peripheral region of an image as a center of the image and ensures an average brightness. <P>SOLUTION: Low-pass filtering is applied only to the outermost periphery of an original image to reduce the noise in the outermost periphery of image, thereby preventing errors in extrapolation. Then, an extended region is provided around the image by extrapolation, thereby eliminating the need for border processing in the periphery of the image when performing the final low-pass filtering. Finally, low-pass filtering is applied to the regions where the original image exists. Thus, a filtering effect can be obtained in which the regions nearer to the center of the image and the preripheral regions of the image have the same characteristics. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing method for applying a low-pass filter to a captured image.

  An image processing method for applying a low-pass filter to an image generally obtains an average value of peripheral pixels of a target pixel and uses this as a new luminance value of the target pixel. At this time, it is possible to perform a powerful low-pass filter process by widening the area for averaging, and it is possible to obtain a macro luminance distribution by removing local features of the image.

Conventional image processing methods are mainly used for the purpose of noise removal. For this reason, a device has been devised so as to retain the pixel edge component to be processed as much as possible, and when taking the average value of the surrounding pixels, a method such as selecting only pixels in a direction with high correlation is employed. (For example, refer to Patent Document 1).
JP 2001-61157 A

  However, the conventional technology does not give consideration to the processing of the peripheral portion of the image. For example, when low pass filter processing is performed by obtaining an average value of three pixels above, below, left, and right of the target pixel, it is necessary to determine an average value of 7 × 7 pixels centered on the target pixel. However, when the target pixel is located at the right end of the image, there is no pixel on the left side. For this reason, in a general method, for the area of three pixels around the image, the original luminance value is used as it is, or a new luminance value is obtained by an average value only for the portion where the pixel image exists, etc. This method is used.

  When the original luminance value is used as it is, there is a problem that the pixel portion is not subjected to low-pass filter processing. In addition, when a new luminance value is obtained based on an average value for only a portion where pixels are present, there is a problem that the luminance level is shifted in addition to the problem that the strength of the low-pass filter varies depending on the location. The reason why the brightness level is shifted is that, for example, when the brightness of the image increases from left to right, when the target pixel is the right end of the image, the average of the target pixel and the pixel located on the left (the pixel below the target pixel) This is because a new luminance is obtained from an average value of only pixels having luminance values), and as a result, has a value lower than the original luminance value at this position.

  SUMMARY OF THE INVENTION The present invention solves the above-described conventional problems, and an object thereof is to provide an image processing method and an image processing apparatus capable of performing a suitable low-pass filter process in which the average luminance is guaranteed for peripheral pixels.

  In the image processing method according to claim 1 of the present invention, when the image is handled as two-dimensional data of pixels having multi-tone luminance, at least low-pass filter processing is performed on the outermost peripheral pixel of the original image or the pixels near the outermost periphery. And expanding the area of the original image and obtaining the luminance of the expanded area by extrapolating pixels included in the low-pass filtered range as a turning point, and including the expanded area And a step of performing low-pass filter processing on the area of the original image.

  The image processing method according to claim 2 of the present invention is the image processing method according to claim 1, wherein at least the low-pass filter processing is performed on the outermost peripheral pixel of the original image or a pixel near the outermost periphery, the outermost pixel of the original image It is characterized by one-dimensional low-pass filter processing.

  According to a third aspect of the present invention, in the image processing method according to the first aspect, in the step of performing at least a low-pass filter process on the outermost peripheral pixel of the original image or a pixel in the vicinity of the outermost periphery, the direction is perpendicular to the extrapolating direction. A low-pass filter process is performed using the pixel located.

  The image processing method according to a fourth aspect of the present invention is the image processing method according to the first aspect, wherein the expansion range for expanding the region of the image is determined according to the degree of low pass filter processing to the region of the original image including the expanded region. It is characterized by doing.

  The image processing method according to claim 5 of the present invention is the image processing method according to claim 4, wherein the size of the surrounding pixel area required for the low-pass filter processing for the area of the original image including the expanded area is vertical A × horizontal B. When the size of the original image is extended, the original image is expanded by A pixels in the vertical direction and B pixels in the horizontal direction.

  An image processing method according to a sixth aspect of the present invention is the second processed image according to the first aspect, wherein a difference between the first processed image obtained by the image processing method according to the first aspect and the original image is obtained. Further comprising the step of obtaining

The image processing method according to claim 7 of the present invention is characterized in that, in claim 6, the second processed image is binarized to detect a defective portion included in the original image.
The image processing apparatus according to claim 8 of the present invention detects the first processed image that is very close to the luminance distribution of the original image by averaging the original image obtained by photographing the flat plate used for the inspection by the inspection target camera. An image processing unit; a subtractor that detects a second processed image that is a difference between the first processed image and the original image; and a binarization circuit that detects a result image obtained by binarizing the second processed image. And a determination circuit for determining whether or not the quality of the inspection target camera is acceptable in comparison with the result image and a quality acceptance standard.

  In the image processing method according to claim 9 of the present invention, the luminance distribution of the original image is obtained by performing at least a low-pass filter process on the outermost peripheral pixel or the pixels in the vicinity of the outermost periphery of the original image obtained by photographing the flat plate used for inspection by the inspection target camera. Detecting a first processed image P1 that is very close to the first processed image, detecting a second processed image having a pixel-by-pixel difference between the first processed image and the original image, and for the second processed image A step of detecting a result image further binarized by median filtering; a step of comparing the count value of the total number of defective pixels included in the result image with the number of defective pixels of quality standards to determine a defective / non-defective product; It is provided with.

  In the image processing method according to claim 10 of the present invention, at the time of image processing in which an image is handled as two-dimensional data of pixels having multi-tone luminance, at least a low-pass filter is applied to the outermost peripheral pixel or the outermost peripheral pixel of the original image. A first processing image obtained by calculating the luminance of the expanded region by extending the region of the original image and extrapolating a pixel included in the low-pass filtered region as a turning point A step of obtaining a second processed image obtained by subjecting the first processed image to low pass filter processing, and a difference obtained for each pixel by comparing the second processed image and the original image for each corresponding pixel. Generating a low-pass filtered image using only the pixels of the first processed image at a position where is a value smaller than a predetermined threshold value. Characterized in that was.

  The image processing method according to an eleventh aspect of the present invention is the image processing method according to the first aspect, wherein only the pixel of the first processed image at a position where the difference obtained for each pixel of the original image is smaller than a predetermined threshold value. And a step of generating a low-pass filtered image using an image as a calibration image, and further using the calibration image to correct a correction target image different from the original image. And

  The image processing method according to claim 12 of the present invention is characterized in that in claim 11, an image other than the original image and a calibration image are divided for each pixel to correct an image different from the original image. To do.

  The image processing method according to claim 13 of the present invention includes a step of performing at least a low-pass filter process on the outermost peripheral pixel of the original image or a pixel in the vicinity of the outermost periphery, expanding the area of the original image, and applying the low-pass filter. A step of obtaining a first processed image obtained by obtaining the luminance of the expanded area by extrapolating a pixel included in the range as a turning point; and a second processed image obtained by low-pass filtering the first processed image And the second processed image and the original image are compared for each pixel, and the low-pass filter processing is performed again on the first processed image without using a pixel that is different from a predetermined threshold value or more. And a step of performing.

  An image processing method according to a fourteenth aspect of the present invention is the image processing method according to the thirteenth aspect, wherein the second processed image and the original image are compared for each pixel so as not to use a pixel that is different from a predetermined threshold value or more. When the low-pass filter processing is performed again on the first processed image, the luminance of the original pixel is compared with the second processed image for each pixel. A step of marking and a step of setting the marked pixel as a target pixel, examining a value of a pixel used for a low-pass filter located in the vicinity of the target pixel, and performing a process used for an average calculation to obtain a new luminance value of the target pixel It is characterized by comprising.

According to the image processing method of the present invention, it is possible to bring about a preferable low-pass filter result that guarantees the average luminance for pixels near the image.
In addition, according to the image processing apparatus using the image processing method of the present invention, it is possible to realize a camera inspection apparatus that can detect a defect without malfunction even in a peripheral portion of the image, and to accurately remove unevenness in luminance of the image. The fluorescence microscope can be realized.

The image processing method of the present invention will be described below based on specific embodiments.
Example 1
1 to 4 show (Embodiment 1) of the present invention.

FIG. 1 shows a processing flow of an image processing method in Embodiment 1 of the present invention.
In step 101, only a pixel located on the outermost periphery of the target original image P0 is subjected to low-pass filter processing.

In step 102, extrapolation is performed to create an image that is slightly larger than the original image.
In step 103, a low-pass filter process is performed on the entire original image to create a target low-pass image.

Next, the above procedure will be described using a more specific example.
FIG. 2 is an image diagram showing the original image and each pixel after the processing in step 101. In FIG. 2, reference numeral 201 denotes an original image, which is a black and white image in which each pixel has 256 gradation luminance values of 0 to 255. In order to simplify the description, it is assumed that it has a size of 9 horizontal pixels × 7 vertical pixels. The luminance value at each pixel is represented by A (x, y).

First, with respect to this original image 201, the procedure of the low-pass filter processing is shown only for the pixels located at the outermost periphery in step 101.
Here, only the pixels shown by shading are subjected to low-pass filter processing. First, A (1,0) to A (7,0) are low-pass filtered in the horizontal direction. As the low-pass filter in the horizontal direction, basically, an average of the luminance values of the original image of five pixels centered on the target pixel A (x, y) is obtained, and this is used as a new luminance value B ( x, y). For example, when A (4, 0) is the target pixel,
(A (2,0) + A (3,0) + A (4,0) + A (5,0) + A (6,0)) / 5
Is determined as a new luminance value B (4, 0). When the pixel of interest is A (1, 0), there is no pixel on the left side for A (1, 0), and therefore the pixel that does not exist is excluded from the low-pass filter calculation.
(A (0,0) + A (1,0) + A (2,0) + A (3,0)) / 4
Is a new luminance value B (1, 0). Similarly, if the pixel of interest is A (7,0), (A (5,0) + A (6,0) + A (7,0) + A (8,0)) / 4
Is a new luminance value B (7, 0). A (1, 6) to A (7, 6) are also subjected to a low-pass filter process in the horizontal direction in the same manner.

A (0,1) to A (0,5) are low-pass filtered in the vertical direction. As the low-pass filter in the vertical direction, basically, an average of the luminance values of the original image of five pixels centered on the target pixel A (x, y) is obtained, and this is used as a new luminance value B ( x, y). For example, when A (0,2) is the target pixel,
(A (0,0) + A (0,1) + A (0,2) + A (0,3) + A (0,4)) / 5
Is determined as a new luminance value B (0, 2). When the pixel of interest is A (0,1), since there is no upper pixel for A (0,1), the pixel that does not exist is excluded from the low-pass filter calculation.
(A (0,0) + A (0,1) + A (0,2) + A (0,3)) / 4
Is a new luminance value B (0, 1). A (8, 1) to A (8, 5) are also subjected to a low-pass filter process in the vertical direction in the same manner.

In the case where A (0,0), A (8,0), A (0,6) and A (8,6) are the target pixels, these are located at the corners. Performs low-pass filtering in the horizontal direction. Therefore, when A (0,0) is the target pixel,
(A (0,0) + A (1,0) + A (2,0) + A (0,1) + A (0,2)) / 5
Is determined as a new luminance value B (0, 0), and A (8, 0) is the target pixel.
(A (6,0) + A (7,0) + A (8,0) + A (8,1) + A (8,2)) / 5
Is determined as a new luminance value B (8, 0). The same method is used for A (0,6) and A (8,6).

With the above operation, an image 202 is obtained in which only the pixels located on the outermost periphery are subjected to low-pass filter processing.
In the above description, the luminance average of five pixels centered on the pixel of interest is used as the low-pass filter. Similar effects can be obtained.

Next, details of the processing in step 102 will be described with reference to FIG.
FIG. 3 is an image diagram showing each pixel of the image when the process of step 102 is performed. Reference numeral 301 denotes an image created by the process of step 102, in which a white part is a part created by the process of step 101 (image 202 in which only a pixel located at the outermost periphery is subjected to low-pass filter processing). The portion shown in the figure is a pixel newly created by the processing in step 102.

  That is, the white portion has the same size as the original image, and the shaded portion is an expanded portion. In the white portion, the portion represented by A (x, y) has the same luminance value as that of the original image, and the portion represented by B (x, y) is the luminance value that has become a new value by the processing in step 101. . The width of the area to be expanded is determined according to the area required for the low-pass filter used in step 103.

  Here, the low-pass filter used in step 103 obtains the average of 5 × 5 pixel areas centered on the target pixel, and sets this as a new luminance value of the target pixel. For this reason, for example, in order to obtain the processing result at the position of B (0, 0) with high accuracy, the luminance values of the upper two pixels and the left two pixels are required. The area for two pixels is expanded. The luminance value of the expanded area is obtained by extrapolation.

First, a detailed description will be given of how to obtain the luminance values of the extended areas on the upper and lower sides of the image.
For example, the luminance value of C (0, −1) is obtained by “2 * B (0, 0) −B (0, 1)”, and the luminance value of C (8, −1) is “2 * B (8,0) -B (8,1) ". Further, the luminance value of C (1, −1) is obtained by “2 * B (1, 0) −A (1, 1)”, and C (2, −1) to C (7) are obtained by the same method. , -1) can be obtained. Further, the luminance value of C (0, -2) is obtained by “2 * B (0,0) −B (0,2)”, and the luminance value of C (8, −2) is “2 * B (8,0) -B (8,2) ". Further, the luminance value of C (1, −2) is obtained by “2 * B (1,0) −A (1,2)”. Similarly, each value of C (2, -2) to C (7, -2) can be obtained.

  Further, the luminance value of C (0,7) is obtained by “2 * B (0,6) −B (0,5)”, and the luminance value of C (8,7) is “2 * B (8 , 6) -B (8, 5) ". The luminance value of C (1,7) is obtained by “2 * B (1,6) −A (1,5)”, and C (1,7) to C (7,7) is obtained in the same manner. ) Can be obtained. The luminance value of C (0,8) is obtained by “2 * B (0,6) −B (0,4)”, and the luminance value of C (8,8) is “2 * B (8 , 6) -B (8, 4) ". Further, the luminance value of C (1,8) is obtained by “2 * B (1,6) −A (1,4)”. Similarly, each value of C (2,8) to C (7,8) can be obtained.

Next, how to determine the luminance values of the extended areas on the left side and the right side of the image will be described in detail.
For example, the luminance value of C (−1, 0) is obtained by “2 * B (0, 0) −B (1, 0)”, and the luminance value of C (−1, 6) is “2 * B (0,6) -B (1,6) ". Further, the luminance value of C (−1, 1) is obtained by “2 * B (0, 1) −A (1, 1)”, and C (−1, 2) to C (−) is obtained by the same method. Each value of 1,5) can be obtained.

  In addition, since the upper and lower extended regions of the original image have already been obtained, C (−1, −2) is obtained by the same method using C (0, −2) and C (1, −2). Can do. Further, C (-1, -1), C (-1, 7), and C (-1, 8) can be obtained by the same method. Next, the luminance value of C (−2,0) is obtained by “2 * B (0,0) −B (2,0)”, and the luminance value of C (−2,6) is “2 * B”. B (0,6) -B (2,6) ". Further, the luminance value of C (−2,1) is obtained by “2 * B (0,1) −A (1,2)”, and C (−2,2) to C (−) is obtained by the same method. 2 and 5) can be obtained.

  In addition, since the upper and lower extended regions of the original image have already been obtained, C (−2, −2) is obtained by the same method using C (0, −2) and C (2, −2). Can do. Further, C (−2, −1), C (−2, 7), and C (−2, 8) can be obtained by the same method.

  Further, the luminance value of C (9,0) is obtained by “2 * B (8,0) −B (7,0)”, and the luminance value of C (9,6) is “2 * B (8 , 6) -B (7, 6) ". The luminance value of C (9,1) is obtained by “2 * B (8,1) −A (7,1)”, and C (9,2) to C (9,5) are obtained in the same manner. ) Can be obtained. In addition, since the upper and lower extended regions of the original image have already been obtained, C (9, -2) can be obtained by the same method using C (8, -2) and C (7, -2). it can. Further, C (9, -1), C (9, 7), and C (9, 8) can be obtained by the same method.

  Next, the luminance value of C (10,0) is obtained by “2 * B (8,0) −B (6,0)”, and the luminance value of C (10,6) is “2 * B ( 8,6) -B (6,6) ". The luminance value of C (10,1) is obtained by “2 * B (8,1) −A (6,1)”, and C (10,2) to C (10,5) is obtained in the same manner. Each value of can be obtained. Since the upper and lower extended regions of the original image have already been obtained, C (10, -2) can be obtained by the same method using C (8, -2) and C (6, -2). it can. Further, C (10, -1), C (10, 7), and C (10, 8) can be obtained by the same method. As described above, the luminance values of all the pixels of the image 301 can be obtained.

Finally, the process in step 103 will be described.
Since only the area of the original image is necessary for the final processed image, only the white part in FIG. 3 is calculated here. As described above, the low-pass filter performed in step 103 obtains the average of 5 × 5 pixel areas centered on the target pixel, and sets this as a new luminance value of the target pixel. Therefore, in FIG. 3, the processed luminance value at the pixel position of B (0,0) is an average of a 5 × 5 pixel area centered on this,
(C (-2, -2) + C (-1, -2) + C (0, -2) + C (1, -2) + C (2, -2) + C (-2, -1) + C (-1 , -1) + C (0, -1) +
C (1, -1) + C (2, -1) + C (-2,0) + C (-1,0) + B (0,0) + B (1,0) + B (2,0) + C (-2 , 1) + C (-1,1) + B (0,1) + B (1,1) + B (2,1) + C (-2,2) + C (-1,2) + B (0,2) + B ( 1,2) + B (2,2)) / 25
Is required. Thereafter, the same processing is performed on all the white portions in FIG. 3 to obtain a target processed image. Here, as an example of the low-pass filter, the case where the average of the 5 × 5 pixel region is used is shown, but even when the range of the low-pass filter is different, another means such as using a weighted average is used. Even if it is used, the same effect can be obtained.

  Here, since the pixel is created around the original image by the process of step 102, it is not necessary to perform exception processing specialized for the end part in the process of step 103, and the center part and the end part The same processing can be performed. The effect at this time will be described with reference to FIG.

  Here, the description assumes that the luminance of the original image increases from left to right, but for the sake of simplicity, the vertical direction of the image is assumed to have uniform luminance, and the horizontal direction This will be described in a one-dimensional manner limited to this process.

  FIG. 4A shows one row of pixels in the third row from the top of the original image in FIG. 2 and expresses the luminance value by height. 401 to 409 are A (0,2) to A ( The brightness value of 8, 2) is indicated by height. As can be seen from the figure, the brightness increases as going to the right. In the conventional low-pass filter, the average value of five pixels centered on the target pixel is used as the processing result. However, since there is no right pixel at the A (8,2) position, the luminance values 407 to 409 are 3. The average value of pixels is obtained. As can be seen from the figure, this result is almost equal to the height of the luminance value 408 and is lower than the original luminance value 409.

On the other hand, FIG. 4B shows an example of image processing according to the present invention.
FIG. 4B is a diagram in which pixels in the fifth row from the top of the image in FIG. 3 are extracted for one row, and the luminance values are expressed in height. 401 to 408 are A (0,2) to A (7 2), 410 indicates B (8,2), and 411-412 indicate the luminance values of C (9,2) -C (10,2) in height.

  The luminance value 410 is created by the low-pass filter in the vertical direction in step 101. Here, since it is assumed that the same luminance value is arranged in the vertical direction of the original image, the luminance value 410 is the luminance value 409. Has the same value as In addition, the luminance values 411 and 412 are generated by extrapolation in step 102, and the luminance increases at the same rate as the luminance values 407 to 410. In step 103, the process at the B (8, 2) position is to obtain an average value of five pixels having luminance values 407 to 412. As can be seen from the figure, this result is almost equal to the height of the luminance value 410. Become. From this, it can be seen that according to the present invention, a value close to the average value of the original image can be obtained even in the peripheral portion of the image.

Further, when the luminance value of the pixel in the extended area is obtained by the processing in step 102, it is obtained by extrapolation. In extrapolation, for example, C (2, -1) = 2 * B (2,0) -A (2,1)
As can be seen from the equation, when B (2, 0) has an error due to noise or the like, the value of C (2, −1) includes an error twice that error. Become. Similarly, C (2, −2) has the same error. When the luminance at the position of B (2,0) is obtained in the processing of 103, C (2, −1) and C (2, -2) is used, this error greatly affects the result. That is, it can be seen that when there is an error such as noise in the luminance of the outermost periphery of the original image, the final processing result is greatly affected. The processing in step 101 prevents this, and a low-pass filter is applied in advance to the outermost peripheral pixels of the original image to prevent noise.

Therefore, it is possible to realize a low-pass filter that retains the luminance average of the original pixels.
As described above, in the first embodiment, when the low-pass filter process is performed, the low-pass filter process is performed after the periphery of the image is expanded. A suitable low-pass filter result in which the average luminance is guaranteed for the peripheral pixels can be obtained.

(Example 2)
FIG. 5 shows an image processing apparatus that executes image processing based on the image processing method described in the first embodiment. FIG. 9 shows a specific processing method of the defect detection apparatus 502.

  This image processing apparatus can be used for inspection of an image photographing device using a lens or a CCD (charge-coupled device) such as a digital still camera or a digital video. Here, an example applied to an inspection apparatus for inspecting a lens or CCD defect in a digital still camera to be inspected will be described in detail below.

  In FIG. 5, a camera 500 is a digital still camera to be inspected, and a plane plate 501 used for inspection is photographed using the camera. The flat plate 501 is a light-colored plain plate, and the entire surface is painted with a single color of gray, for example. The original image P0 is an image obtained by photographing the flat plate 501 with the camera 500, and is digital image data. In this embodiment, it is assumed that the original image P0 has a luminance of 256 gradations per pixel of the CCD of the camera 500 and a size of 1024 × 768 pixels. A method for detecting the lens or CCD defect of the camera 500 using the original image P0 using the defect detection device 502 will be described below.

  Reference numeral 503 denotes an image processing unit that realizes the image processing method according to the first embodiment, and the image processing unit 503 performs a process of averaging pixels in a range of 100 × 100. The size of the area of the image processing unit 503 is a sufficiently large area for performing the averaging process according to the present invention because the assumed defective pixels are about 50 × 50 at the maximum. Here, for each pixel in the defect area of maximum 50 × 50, when averaging using the surrounding 100 × 100 pixels, the defect portion is maximum 1/4, and at least 3/4 is Has a background luminance value. Therefore, the luminance value of the pixel after averaging is much closer to the luminance value of the background than the luminance value of the defective portion. As a result, the influence of the defective portion is almost eliminated, and the output of the image processing unit 503 becomes the first processed image P1 that is very close to the luminance distribution of the original image P0.

  Reference numeral 504 denotes a subtracter that calculates the difference between the first processed image P1 and the original image P0 output from the image processing unit 503. In the second processed image P2 output from the subtractor 504, only the defective portion which is the difference between them has a value other than zero.

  Next, the output of the subtracter 504 is input to the binarization circuit 505. The binarization circuit 505 evaluates the output from the subtractor 504 for each pixel. When the absolute value of the luminance value of the target pixel is larger than a predetermined threshold value, “1” is set for the pixel, and the threshold value is less than the threshold value. In this case, “0” is output for the pixel. Here, the threshold value is set to “1” in order to detect a very fine defect. However, if the defect is clear, the threshold value can be set to a larger value. The output from the binarization circuit 505 outputs different pixels of the original image P0 and the original luminance distribution of the original image P0 as “1”, and as a result, only defective pixels are “1”. As a result image P3 can be obtained.

  The result image P3 includes all pixels in the defective portion. However, depending on the quality acceptance standard of the product to be inspected, a non-defective product is determined according to the level of the defect, and therefore the result image P3 should not be used as it is. Therefore, the evaluation value calculation circuit 506 and the determination circuit 507 evaluate the result image P3 and compare with the quality acceptance standard input from the shipping range input unit 508 to determine whether the quality of the camera 500 to be inspected is acceptable. .

  For example, when the criterion that the defect is accepted if the total number of pixels determined to be defective is 50 pixels or more is input in advance to the shipping range input means 508, the determination circuit 507 has a value of “1” in the result image P3. The total number of pixels having the pixel number is obtained. When the number is less than 50 pixels, a pass determination is performed, and when it is 50 pixels or more, a pass determination is performed, and this is output to the result display device 509. In addition, a criterion that the quality acceptable standard input from the shipping range input unit 508 is rejected if the total of the maximum number of pixels of the continuous defective pixel aggregate is 30 pixels or more is input in advance. In this case, the determination circuit 507 searches for pixels having the value “1” in the result image P3, and searches for the number of pixels having the maximum size. When the number of pixels is smaller than the number of pixels, the determination of acceptance is made, and when the number of pixels is 30 pixels or more, the determination of failure is made, and this is outputted to the result display device 509. In the result display device 509, the determination result from the determination circuit 507 is displayed on the monitor, and the pass / fail of the product is communicated to a person.

  In the above, the second processed image P2 uses the output of the subtractor 504 as it is. However, since the noise component included in the original image P0 actually has a value other than 0, In order to remove this, the output of the subtractor 504 obtained by applying a median filter can be used as the second processed image P2. The median filter is generally used in image processing. The median filter arranges the luminance values of the pixels in the peripheral region of the target pixel in order of size, and the one located in the center is used as the luminance value of the target pixel. The median filter can eliminate a noise portion while leaving only a luminance change region having a certain size.

  In the conventional image processing apparatus, since the average luminance is not guaranteed in the peripheral portion, if the conventional image processing method is used for the image processing unit 503, there is a possibility that defect detection in the peripheral portion cannot be performed normally. On the other hand, according to this method, since the average value of the peripheral part of the low-pass filter is close to the original image, it is possible to accurately extract only the defect.

  As described above, in this embodiment, the original image P0 and the first processed image P1 representing the luminance distribution inherent in the original image P0 obtained by the image processing method according to the first embodiment are subtracted and binarized. As a result, only defects can be accurately extracted.

Next, an example of a processing method when the defect detection apparatus 502 described above is realized by software will be described in detail with reference to FIG.
FIG. 9 is a flowchart of an embodiment of a method for detecting a lens or CCD defect in an inspection of a digital still camera or the like.

  In step 901, the low pass filter shown in the first embodiment is first applied to the original image P0 photographed by a camera or the like. Here, it is assumed that the pixels in the range of 100 × 100 are averaged, and this is a sufficiently large region with respect to the assumed maximum defect region (50 × 50). For this reason, the influence of the defective portion is almost eliminated by the same principle as that of the image processing unit 503 described above, and the processing result is the first processed image P1 very close to the luminance distribution of the original image P0.

  In step 902, a pixel-by-pixel difference between the first processed image P1 and the original image P0 is obtained. As a result of this processing, only the defective portion, which is ideally the difference between the two, has a value other than zero. However, in the actual result, the noise component included in the original image P0 also has a value other than 0. Therefore, the median filtering process is performed on the processing result in step 902 in step 903.

  The median filter is generally used in image processing. The median filter arranges the luminance values of the pixels in the peripheral region of the target pixel in order of size, and the one located in the center is used as the luminance value of the target pixel. The median filter can eliminate a noise portion while leaving only a luminance change region having a certain size. For this reason, in the processing result P2 of step 903, only the defective part has a value other than zero.

  In step 904, the processing result P2 is binarized. In this binarization, the processing result P2 is evaluated for each pixel. When the absolute value of the luminance value of the target pixel is larger than a predetermined threshold value, “1” is set for the pixel. “0” is output to the pixel. Here, the threshold value is set to “1” in order to detect a very fine defect. However, if the defect is clear, the threshold value can be set to a larger value. As a result of the binarization, different pixels between the original image P0 and the original luminance distribution of the original image P0 are output as “1”, and as a result, only defective pixels are expressed as “1”. A result image P3 can be obtained.

  In step 905, the total number of defective pixels included in the result image P3 is counted. This can be realized by counting pixels having a value of “1” in the entire result image P3.

  In step 906, the value counted in step 905 is compared with the quality standard number of defective pixels 50 based on a quality criterion given in advance, for example, a criterion “a product having a defective pixel number of 50 or more is defective”. If the value counted in step 905 is 50 or more, information indicating failure is output in step 907. If the value counted in step 905 is less than 50, information indicating that the product is non-defective is output in step 908.

(Example 3)
6 to 8 show a fluorescence microscope image acquisition apparatus that executes image processing based on the image processing method described in (Example 1).

  In FIG. 6, reference numeral 601 denotes an observation target as a correction target image, for example, a bacterial cell stained with a fluorescent dye or the like. When the laser light from the laser light source 602 is reflected by the dichroic mirror 603 and applied to the observation object 601, the fluorescent dye of the observation object 601 is excited by the laser and emits light at a specific wavelength. This light is captured by the CCD camera 604, digitized as image data, and processed by the host computer 605.

  Here, it is difficult to make all the laser beams irradiated to the observation object 601 uniform in the observation plane, and usually has an illuminance distribution in which the central part is bright and the peripheral part is dark. When the fluorescent image excited by the laser light is displayed as it is, the original distribution of the fluorescent dye is not accurately displayed, so that the illuminance distribution of the laser light needs to be calibrated.

Calibration is performed as follows using a calibration image prepared separately from an image obtained by photographing the observation object 601.
The calibration image is acquired by photographing a standard sample on which a fluorescent dye is uniformly applied instead of the observation object 601 in FIG. Since the acquired image represents the illuminance distribution of the laser light, the calibration image is divided from the image obtained by photographing the observation target for each corresponding pixel, and calibration is performed to cancel the influence of the illuminance distribution of the laser light. be able to.

  However, it is difficult to ideally create a standard sample for use in acquiring a calibration image, and in fact, in many cases, small dust is mixed. When calibration is performed using the calibration image in which the dust is mixed as it is, the influence of the dust appears as a shadow in the displayed image.

Therefore, in (Embodiment 3), the host computer 605 is configured to execute the following image processing.
The host computer 605 includes a calibration image creation routine 606 and an observation image correction routine 607. A calibration image creation routine 606 for creating an ideal calibration image from the photographed standard sample image is configured as shown in FIG.

In step 701, only the outermost peripheral pixel of the photographed standard sample image is low-pass filtered.
In step 702, a slightly larger image is created by extrapolation.

In step 703, low-pass filter processing is performed on the range of the original image using this. Up to this point, the same method as in the first embodiment is taken.
Finally, in step 704, the image created in step 703 and the original image are compared for each pixel, and the pixel generated by step 702 is not used by using pixels that differ by a predetermined threshold or more. The low-pass filter process is performed again.

FIG. 8 shows the process in step 704 in more detail.
In step 801, the result of step 703 is compared with the luminance of the original pixel for each pixel. As a result, if the luminance value is different by “10” or more, the corresponding pixel as a result of step 703 is set to −1 in step 802, and the marking means that the error is large. Here, there is a high possibility that a portion having a large error is a dust portion, and the marking here has a meaning of marking a dust portion.

Here, the threshold value of the luminance value is set to “10” or more, but is not limited to “10”, and may be set to any other value.
This process is performed for all the pixels of the original image, and if not processed in step 803, steps 801 to 802 are repeated. When marking of a pixel having a large difference is completed for all pixels, the process proceeds to step 804 in order to perform the low-pass filter again.

In step 804, the pixel designated by the order is set as the target pixel. In step 805, the value of the pixel used for the low-pass filter located in the vicinity of the target pixel is checked.
When this value is −1, this pixel is ignored, and when it is not −1, processing for using this value for averaging is performed at step 806. For example, when the number of pixels used for the low-pass filter is 25, and there are two of them that are -1, the average value of the remaining 23 is obtained, and that value is used as the new luminance of the target pixel. Value. In step 807, it is determined whether or not all of the neighboring pixels used for the low-pass filter have been processed in the current pixel of interest. If unprocessed, steps 805 to 806 are repeated. If it is determined in step 807 that all processing has been completed, the result of the low-pass filter is recorded for one pixel in step 808.

  In step 809, it is determined whether or not the above processing has been completed for all the pixels. If the processing has not been completed, steps 804 to 808 are repeated. If all the pixels have been processed, this processing is terminated.

  By this processing, a suitable calibration image that is not affected by dust can be obtained by performing low-pass filter processing without using pixels that deviate from the average brightness due to dust or the like. The created calibration image is stored in the host computer 605.

  After the calibration image creation routine 606 of the host computer 605 is executed in this way and the acquisition of the calibration image is completed, the observation object 601 is stained from the standard sample with a final observation object such as a fluorescent dye. When the observation image correction routine 607 is instructed to be exchanged for the bacterial cell or the like, the observation image correction routine 607 performs low-pass filter processing without using pixels that deviate from the average brightness due to dust or the like. By using a suitable calibration image that is not affected and performing division for each corresponding pixel, calibration is performed to accurately cancel the influence of the illuminance distribution of the laser beam, and the distribution of the original fluorescent dye is determined. It can be displayed accurately.

  In each of the above embodiments, the low pass filter is applied only to the outermost peripheral pixel of the original image P0 in the initial stage of processing. However, this is because the low pass filter is applied only to the pixels near the outermost peripheral pixel of the original image P0. It is also possible to configure such that image processing in the subsequent process is executed by discarding the pixels outside that.

  In each of the above embodiments, at the initial stage of processing, only the outermost peripheral pixel of the original image P0 or only the pixels near the outermost peripheral pixel of the original image P0 is subjected to the low pass filter. Can also be configured to simultaneously perform low-pass filter processing, and the object can be achieved by performing at least low-pass filter processing on the outermost peripheral pixel of the original image or pixels near the outermost periphery.

  In the present specification, “low-pass filter processing using pixels positioned in a direction perpendicular to the extrapolation direction” specifically refers to pixel B (2, 0) and pixel A ( 2 and 0), when extrapolating pixel C (2, -1) and pixel C (2, -2) vertically, pixel B (2, 0) is the turning point. The pixel B (1, 0) and the pixel B (3, 0) in the horizontal direction with respect to the extrapolation direction are pixels positioned in the vertical direction. When extrapolating the pixel C (−1,2) and the pixel C (−2,2) horizontally to the pixel B (0,2) and the pixel A (1,2), the pixel B (0,2) ) Is a turning point. In this case, the pixel B (0, 3) in the vertical direction with respect to the horizontal extrapolation direction and the pixel B (0, 1) are pixels positioned in the vertical direction with respect to the extrapolation direction.

  Since the image processing method according to the present invention provides an accurate processing result that guarantees the average luminance of the peripheral portion, a highly accurate low-pass filter can be realized. By using an image processing apparatus provided with this image processing method for inspection of a digital camera, it is useful that an excellent inspection method without erroneous detection can be realized.

  Further, by using the image processing method of the present invention for calibration of a fluorescence microscope, it is useful that accurate calibration can be realized and accurate display is possible.

Process Flowchart of Image Processing Method in (Embodiment 1) of the Present Invention Pixel arrangement diagram of original image in the same embodiment Pixel arrangement diagram of intermediate image in the same embodiment Image diagram explaining the effect of the embodiment The block diagram of the image processing apparatus in (Example 2) of this invention Block diagram of a fluorescence microscope image system according to (Example 3) of the present invention Processing flowchart for creating a calibration image of the embodiment Flowchart of step 704 in the same embodiment Flowchart when the defect detection apparatus 502 according to the second embodiment of the present invention is realized by software.

Explanation of symbols

101-103 Processing steps of image processing flow 201 Original image 202 Image obtained by performing low-pass filter processing only on pixels located on the outermost periphery 301 Region expanded intermediate image 401-412 Luminance value of each pixel 500 Camera (inspection object)
501 Flat plate 502 Defect detection device 503 Image processing unit 504 Subtractor 505 Binary circuit 506 Evaluation value calculation circuit 507 Determination circuit 601 Observation target 602 Laser light source 603 Dichroic mirror 604 CCD camera 605 Host computer 606 Calibration image creation routine 607 Observation Image correction routine

Claims (14)

When handling images as two-dimensional data of pixels with multi-tone brightness,
At least low-pass filtering the outermost peripheral pixel of the original image or a pixel near the outermost periphery;
Expanding the area of the original image and obtaining the brightness of the expanded area by extrapolating pixels included in the range subjected to the low-pass filter as folding points;
And a step of low-pass filtering the original image area including the expanded area. The image processing method according to claim 1, wherein in the step of performing at least a low-pass filter process on the outermost peripheral pixel of the original image or a pixel near the outermost periphery, a one-dimensional low-pass filter process is performed on the outermost pixel of the original image. 2. The image processing method according to claim 1, wherein, in the step of performing at least a low-pass filter process on the outermost peripheral pixel of the original image or a pixel in the vicinity of the outermost periphery, a low-pass filter process is performed using pixels positioned in a direction perpendicular to the extrapolation direction. . The image processing method according to claim 1, wherein an expansion range for expanding the image area is determined according to a degree of low-pass filter processing performed on the original image area including the expanded area. When the size of the surrounding pixel area required for the low-pass filter processing including the expanded area is low-pass filter processing in the area of the original image has a size of vertical A × horizontal B, the vertical length of the original image is A pixel. The image processing method according to claim 4, wherein the width is expanded by B pixels. The image processing method according to claim 1, further comprising: obtaining a second processed image by obtaining a difference between the first processed image obtained by the image processing method according to claim 1 and the original image. The image processing method according to claim 6, wherein the second processed image is binarized to detect a defective portion included in the original image. An image processing unit for detecting a first processed image that is extremely close to the luminance distribution of the original image by averaging the original image obtained by photographing the flat plate used for the inspection by the inspection target camera;
A subtractor for detecting a second processed image of the difference between the first processed image and the original image;
A binarization circuit for detecting a result image obtained by binarizing the second processed image;
An image processing apparatus provided with a determination circuit that determines the quality of the inspection target camera by comparing the result image with a quality acceptance standard. A step of detecting a first processed image P1 very close to the luminance distribution of the original image by performing at least a low-pass filter process on the outermost peripheral pixel of the original image obtained by photographing the flat plate used for the inspection by the inspection target camera or a pixel near the outermost periphery. When,
Detecting a second processed image of a pixel-by-pixel difference between the first processed image and the original image;
Detecting a result image obtained by performing median filtering on the second processed image and further binarizing;
An image processing method comprising: comparing a count value of the total number of defective pixels included in the result image with a quality-standard defective pixel number to determine a defective / non-defective product. When processing an image as two-dimensional data of pixels having multi-tone luminance,
Performing at least a low-pass filter on the outermost peripheral pixel or the outermost peripheral pixel of the original image;
Expanding a region of the original image and obtaining a first processed image obtained by determining the luminance of the expanded region by extrapolating pixels included in the low-pass filtered range as a turning point;
Obtaining a second processed image obtained by low-pass filtering the first processed image;
The second processed image and the original image are compared for each corresponding pixel, and only the pixel of the first processed image at a position where the difference obtained for each pixel is smaller than a predetermined threshold value is used. An image processing method comprising: generating a low-pass filtered image. Calibrate the image output by the step of generating the low-pass filtered image using only the pixel of the first processed image at a position where the difference obtained for each pixel of the original image is smaller than a predetermined threshold value The image processing method according to claim 1, further comprising a step of correcting a correction target image different from the original image using the calibration image. The image processing method according to claim 11, wherein an image other than the original image is divided for each pixel to correct an image different from the original image. At least low-pass filtering the outermost peripheral pixel of the original image or a pixel near the outermost periphery;
A step of obtaining a first processed image obtained by expanding the region of the original image) and obtaining the brightness of the expanded region by extrapolating pixels included in the low-pass filter as a turning point;
Obtaining a second processed image by low-pass filtering the first processed image;
Comparing the second processed image with the original image for each pixel and performing low-pass filter processing again on the first processed image without using a pixel that is different from a predetermined threshold or more. Image processing method. The second processed image and the original image are compared for each pixel, and the low-pass filter process is performed again on the first processed image without using a pixel that is different from a predetermined threshold value or more.
Comparing the brightness of the second processed image and the original pixel for each pixel and marking the corresponding pixel if the brightness value differs by more than a specified value;
Claiming that the marked pixel is a pixel of interest, a value of a pixel used for a low-pass filter located in the vicinity of the pixel of interest is examined, a process used for averaging is performed, and a new luminance value of the pixel of interest is obtained. Item 14. The image processing method according to Item 13.

Images