JP3115079B2 - Image processing device for light transmitting particles - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to light-transmitting particles on a gray-scale image based on image density information of a gray-scale image obtained by receiving light transmitted through the light-transmitting particles and light from the background. The present invention relates to an image processing apparatus for light-transmitting particles that can be clearly distinguished from a background.

[0002]

2. Description of the Related Art Conventionally, a synthetic tree which is an example of light-transmitting particles
The transparent particles made of fat are suspended in water, and a light source with uniform illuminance is used.
Irradiates the transparent particles with the diffused light of
As a background that does not transmit the transmitted light
The scattered light is imaged with a CCD camera, and
After obtaining the grayscale image related to the
A binary image is created with the specified threshold, and
Image processing method to detect outline shape and particle size of transparent particles
was there. The transparency captured by the conventional image processing method described above.
Fig. 10 shows an example of the original image of the gray image of the bright particles and the background.
You. As shown, transparent particles were imaged using transmitted illumination
In this case, the lens effect unique to this transparent particle works,
The image density at the outer periphery of the child is high and the image density at the center is low
Produces a gray image. A binary image obtained from such an original image
An example of the concept is shown in FIG. In the binary image shown in the figure
Means that, for example, the image of the transparent particles is a dark image area A₀(Hah
Portion) and the dark image area A₀Boundary line L inside
_i(Hereinafter referred to as the inner contour line)
Area H₀(White background portion), and the dark image area A₀Outside
Border L on the side_o(Hereinafter referred to as the outer contour)
Light image area B as image₀~ B_Three(White background) formed
Have been. In the above image processing method,
Light image area H in the transparent particles₀And background light image area B
₀~ B_ThreeAre indistinguishable and are treated the same. That
In fact, spherical transparent particles are actually
It will be recognized as a ridge shape. Therefore,
In the image processing method, for example, each inner contour line L_iAnd outer contour
L_oThe length of the inner contour L_iAnd outer contour line L_oSurrounded by
Light image area H₀And B₀~ B _ThreeThe area and shape of
Therefore, based on these values, the light image area in the transparent particles
Area H ₀And light image area B as background₀~ B_ThreeDistinguish between
It has become so. In addition, the light image area distinguished above
H₀Is numerically filled, for example, as shown in Figure 12.
Above the dark image area A₀By being treated in the same way as
Transparent particles (shown by black background in FIG. 12) and spine on the binary image
The scenery (shown by a white background in FIG. 12) was identified. this
By using the binary image processed as described above,
To facilitate the operation of measuring the size and shape of transparent particles
Has become.

[0003]

In the case where, for example, the transparent particles are arranged in substantially one layer without overlapping in the light transmission direction and their particle diameters are relatively uniform, the conventional image processing method is used. if, it is relatively easy to identify and the light image area B ₀ .about.B ₃ pale image region H ₀ and the background in the transparent particles. However, when the transparent particles are industrially imaged online, it is practically difficult to arrange the transparent particles in a plane perpendicular to the light transmission direction in the imaging section of the apparatus. In many cases, it is necessary to take an image of the transparent particles layered on each other. In such a case, when a binary image is created by the above-described conventional image processing method, there arises a problem that a lower layer transparent particle image distorted by refraction of transmitted light is captured overlapping with an upper layer transparent particle image. actually, the rest without a light image area H ₁ to H ₈ in the transparent particles to be identified as the transparent particles (FIG. 12) is filled, that clearly identify the actual transparent particles and the background in the area could not. Therefore, it has not been possible to improve the measurement accuracy such as the particle size measurement of the transparent particles. SUMMARY OF THE INVENTION The present invention has been made in view of the above-described conventional problems, and is directed to an image processing apparatus for light-transmitting particles capable of clearly distinguishing an image of light-transmitting particles from a background image. It is intended to be provided.

[0004]

In order to achieve the above object, a first aspect of the present invention is to receive light transmitted through light-transmitting particles and light from a background, and to adjust the amount of received light. An image sensor that outputs a generated image, a density image generating means for generating a density image from an output signal from the image sensor, and a density image area based on the density information of the generated density image. An image processing apparatus comprising: a boundary line extracting means for extracting a boundary line to be divided into a light image region; and a method for storing image density gradient data in a direction crossing the boundary line in a pair with the extracted boundary line. 1
And a boundary between the light-transmitting particles and the background, a dark image area and a light image area in the light-transmitting particles, based on a predetermined first threshold value relating to the image density gradient data. The image processing apparatus is configured as an image processing apparatus for light-transmitting particles according to a point including a first boundary line separation unit that separates the boundary line from the boundary line. In order to achieve the above object, a second invention is directed to a light image surrounded by the extracted boundary instead of the image density gradient data in a direction crossing the extracted boundary in the first invention. A boundary between the light-transmitting particles and the background and a boundary between a dark image region and a light image region in the light-transmitting particles are distinguished by using image density distribution data in the region. It is. In order to achieve the above object, a third aspect of the present invention provides a light image surrounded by the extracted boundary instead of the image density gradient data in the direction crossing the extracted boundary in the first aspect. Using the statistical value of the data related to the image density in the area,
The boundary between the light-transmitting particles and the background is distinguished from the boundary between the dark image area and the light image area in the light-transmitting particles.

[0005]

In the image processing apparatus according to the first aspect of the present invention, the boundary line separating the dark image region and the light image region extracted from the gray image obtained from the light transmitted through the light transmitting particles and the light from the background. The image density gradient data in the direction crossing these boundaries is stored in the first storage means. In this case, the boundary between the light-transmitting particles and the background and the boundary inside the light-transmitting particles have a magnitude in the image density gradient data. Therefore, the first boundary line classification means
For example, based on a predetermined first threshold relating to the image density gradient data set manually or automatically in advance,
A boundary line between the light transmitting particles and the background and a boundary line between a dark image region and a light image region in the light transmitting particles are distinguished. Thus, the light image area in the light transmitting particles is distinguished without being regarded as the light image area indicating the background. On the other hand, in the image processing apparatus according to the second aspect, in order to distinguish the light image area in the light-transmitting particles from the light image area in the background, a pair with the boundary line is used. Image density distribution data in the enclosed light image area is stored by the second storage means. The reference pattern of the image density distribution data is different between the image density distribution data of the light image area in the light transmitting particles and the image density distribution data of the light image area representing the background. Therefore, the second boundary line discriminating means distinguishes the light image area in the light transmitting particles from the light image area representing the background based on predetermined reference distribution data on the image density distribution data. Further, in the image processing apparatus according to the third aspect, in order to distinguish the light image area in the light-transmitting particles from the light image area in the background, the boundary line is paired with the boundary image. The statistical value of the data relating to the image density in the enclosed light image area is stored by the third storage means. The statistical value of the data relating to the image density of the light image area in the light transmitting particles is an output value different from the statistical value of the data relating to the image density of the light image area representing the background. Therefore, the third boundary line discriminating means sets a light image area in the light transmitting particles and a light image area representing the background on the basis of a predetermined second threshold value relating to the statistical value of the data relating to the image density. To distinguish.

[0006]

Embodiments of the present invention will be described below with reference to the accompanying drawings to provide an understanding of the present invention. The following embodiment is an example embodying the present invention and is not intended to limit the technical scope of the present invention. Here, FIG. 1 is a schematic configuration diagram showing a transparent particle inspection apparatus according to one embodiment of the present invention, and FIG. 2 is a diagram showing an outer contour line and an inner contour line in a grayscale image of a transparent particle imaged by the transparent particle inspection apparatus. FIG. 3 is a graph showing a characteristic curve of a Mar-Hildress filter used to obtain the outer contour line and FIG. 3 is an explanatory diagram for explaining a zero crossing point method applied to obtain the outer contour line and the inner contour line. (A) is a graph showing an image density curve in the x direction, (b) is a graph showing a first-order differential curve of the image density curve in (a), and (c) is a graph showing the image density curve in (a). FIG. 4 is a graph showing a second derivative curve, FIG. 4 is a flowchart showing a processing procedure of operations from imaging of a transparent particle by the transparent particle inspection apparatus to measurement of a particle diameter, and FIG. Get processed by State diagram illustrating the binary image, FIG. 6 is graph showing the frequency distribution of the mean values for the contour line the entire length of the image gradient direction transverse to Kakusoto contour and the inside contour,
FIG. 7 is an explanatory diagram showing a binary image in which a light image region in a transparent particle distinguished from a light image region representing a background by the image processing device of the transparent particle inspection device is treated as a dark image region, and FIG. FIG. 9 is a state explanatory view showing an embodiment in which an image of each transparent particle is separated from the binary image of FIG. 9 and the respective particle diameters are measured. FIG. It is a graph figure showing the frequency distribution of the image density for every pixel in the light image area used for distinguishing each image of the background, and (a) shows the light image area in the light image area surrounded by the outer contour line. FIG. 4B is a graph showing an image density distribution, and FIG. 4B is a graph showing an image density distribution in a light image area within the inner contour.

As shown in FIG. 1, a transparent particle inspection apparatus 1 according to the present embodiment generates spherical transparent particles 7 and is disposed near the outer periphery of a visible portion 9 of a transparent particle generation apparatus 8 which blows out with water and diffuses. A diffused light from the transmitted illumination device 4 which is disposed on the opposite side of the transmitted illumination device 4 that irradiates the light and the transmitted illumination device 4 with the visible portion 9 interposed therebetween, and receives the diffused light from the transmitted illumination device 4 according to the amount of received light. An imaging unit 2 including a CCD camera 5 (image sensor) having an electronic shutter function for outputting an image data signal, and a grayscale image based on the image data signal from the CCD camera 5. It mainly comprises an image data calculation unit 3 for converting the binary image into an image of the transparent particles 7 from the background image and detecting the particle size and shape of the transparent particles 7 from the background image. The image data calculation unit 3 includes a camera amplifier for amplifying an image data signal from the CCD camera 5 and an image memory IM for temporarily storing the amplified image data signal as a gray image, and stores the gray image from the image memory IM. An image processing device 6 that converts the image into a binary image and identifies a background image and a transparent particle image in the binary image, a monitor that displays a grayscale image or a binary image from the image processing device 6 on a screen,
It is mainly composed of a video hard copy device that prints out a grayscale image or a binary image from the image processing device 6. In order to capture a still image of the transparent particles 7 flowing in the visible portion 9 of the transparent particle generator 8, the electronic shutter was used as described above, but the shutter speed of the electronic shutter was reduced to approximately 1/1000 second. For example, even if the moving speed of the transparent particles 7 is about 5 cm / sec, no problem occurs in terms of image quality. Further, since the light from the transmission illumination device 4 is diffused light,
The image obtained by the CCD camera 5 is an image having an image density gradient. On the other hand, it is well known that the output value of the differential filter differs depending on the density gradient of the image density input.

Therefore, the image processing device 6 of the image data calculating unit 3 uses the contour line (partitioning) for dividing into a dark image area and a light image area based on the image density data of the gray image stored in the image memory IM. The outer contour and the inner contour;
A differential filter was adopted to obtain the boundary line). In particular,
It is well known that a closed contour can always be obtained with a second derivative filter. Therefore, a secondary differential filter using a Laplacian operator was used. In the present embodiment, the Mard-Hildress (Marr-
(Hillreth) filter (boundary extraction means). In the above-mentioned Mahl-Hildes filter, Laplacian of Gaussian function is used. The characteristics of the Mar-Hilless filter are shown in the following equation (1).

(Equation 1) Here, σ indicates the standard deviation of the Gaussian function. The Gaussian function G (y, x) is shown in the following equation (2).

(Equation 2) Note that a Mar.-Mars function employing a Gaussian function as in the above equation (1)
According to the hillless filter, the grayscale image data is smoothed by the Gaussian function, so that the image becomes extremely resistant to noise. FIG. 2 shows an example of the characteristic curve of the Mar-Hillless filter. Then, when filtering a digital image using the above-mentioned Mar-Hildress filter, discretization processing of gray image data is performed as follows. Here, assuming that the image density of the pixel (i, j) on the i-th row and the j-th column on the input image is u (i, j) and each of the parameters ζ, η and the filter radius R are integers, the pixel on the output image is The output value v (i, j) of (i, j) is given by the following equation (3).

(Equation 3) here,

(Equation 4)

(Equation 5) Laplacian of Gaussian function is when x and y are real numbers.

(Equation 6) Satisfy the equation. Here, if the range of the parameters ζ and η is limited to the area A as in the equation (4), the equation (6) is no longer satisfied, and thus the offset h is set in the above equation (4), thereby obtaining Σ∇ ² G (ζ, η) has been corrected.

Using the above-mentioned Mar-Hildress filter, as shown in FIG. 3, the Laplacian output of the image density curve in the x-direction on the grayscale image shown in FIG. At the zero crossing point in the second derivative curve, the density gradient value by the first differentiation shown in (b) shows the maximum value, and the pixel at this zero crossing point can be considered as a contour point. The set of contour points determined in this way is determined as each of the contour lines (outer contour line, inner contour line). By the way, as described above, it is known that there is a difference between the outer contour line and the inner contour line in the density gradient of the image density in the direction crossing the respective contour lines. The grayscale image stored in the image memory IM is directly
Alternatively, primary differential filtering was performed on an output image composed of the Laplacian output of the Mar-Hilless filter. A so-called pre-bit filter is used as the primary differential filter. The density gradient value at the contour point at this time corresponds to, for example, the maximum density gradient value shown in FIG. Then, the image density gradient values calculated for each pixel related to each contour line extracted by the Mar-Hildes filter are summed, and this sum value is divided by the total number of pixels related to the contour line to obtain the sum. The average value of the image density gradient for the contour is obtained. The contour line thus determined and the average value of the image density gradient relating to the contour line are stored in pairs in the memory M of the image processing device 6. The following description discloses an example in which R = 10 is used as the filter radius R of the Mar-Hilless filter, and σ = 2 is used as the standard deviation σ of the Gaussian function. The transparent particle inspection apparatus 1 according to the present embodiment is configured as described above.

Next, the operation of the transparent particle inspection apparatus 1 will be described with reference to the flowchart of FIG. First, the operation clock of the image processing device 6, the memory M,
A predetermined initial setting process for the image memory IM and the program is executed (S1). Thus, the light transmitted through the transparent particles 7 flowing through the visible portion 9 of the transparent particle generation device 8 and the diffused light as the background of the transparent particles 7 are received as a still image by the CCD camera 5 by the action of the electronic shutter ( S2). Then, the grayscale image data signal corresponding to the received light amount from the CCD camera 5 is amplified by a camera amplifier, and then temporarily stored in the image memory IM of the image processing device 6 as, for example, a grayscale image as shown in FIG. Is performed (S3). That is, means for realizing the function of creating a grayscale image from the output signal from the CCD camera 5 by the camera amplifier and the image processing device 6 in step S3 is an example of the grayscale image generating means of the present invention. And
The image processing device 6 applies a Mar-Hilless filter to the grayscale image data stored in the image memory IM,
A contour line as a set of zero-crossing points (pixels) of Laplacian output as shown in FIG. 5 is extracted based on the grayscale image data (S4). As a result, the gray image is divided into a binary image of a dark image region (here, a white background) and a light image region (here, a black background) by the extracted outline. At the same time, the image processing device 6 obtains an image density gradient value in a direction crossing the extracted contour line based on a primary differential value at each zero-crossing point of the Laplacian output, and obtains an image density gradient average value of pixels related to the entire contour line. A value is calculated (S5). The image density gradient average value is stored in the memory M of the image processing device 6 as a pair with the corresponding contour line (S6). That is, the memory M of the image processing device 6
It is an example of the first storage means according to the present invention for storing image density gradient data in a direction crossing the boundary line in a pair with the extracted boundary line. Subsequently, the image processing device 6 calculates the average value of the image density gradient for each of the extracted contour lines as shown in FIG.
The frequency distribution is calculated as shown in FIG. This frequency distribution is
For example, it is displayed and output on a monitor of the image data calculation unit 3 or a video hard copy device (S7). Then, based on the displayed frequency distribution, the operator determines whether the extracted contour is an outer contour which is a boundary between the transparent particles 7 and the background, or a dark image area in the transparent particles 7. A threshold value (first threshold value) for discriminating whether the line is an inner contour line that is a boundary line with the light image region is set and input (S8). Here, it is assumed that a value of 90 has been set as a threshold value for the image density gradient average value by the operator. As described above, the image density gradient on the outer contour line is large and the image density gradient on the inner contour line is small. A contour line corresponding to the image density gradient average value equal to or less than the threshold value is classified as an inner contour line (S9). That is, the arithmetic processing in step S9 executed by the image processing device 6 is referred to as
A first method for distinguishing a boundary between the light transmitting particles and the background and a boundary between a dark image region and a light image region in the light transmitting particles based on a predetermined threshold regarding the image density gradient data. Is an example of the boundary line classification means.

The light image area surrounded by the separated inner contours is treated in the same manner as the dark image area at the periphery of the image relating to the transparent particles 7, and is numerically filled as shown in FIG. And processed (S10). According to the binary image filled in this way, the transparent particles 7 in the visible portion 9
Even if there are multiple layers in the light transmission direction,
As in the binary image obtained by the conventional image processing technique, in fact without causing different light image areas H ₁ to H ₈ (FIG. 12) and an image of very accurately transparent particles 7 of the image and the background Can be identified. Subsequently, the binary image shown in FIG. 7 is subjected to the separation processing of the overlapping image of the transparent particles 7 as shown in FIG. 8 using a well-known inscribed circle filling method (curvature K = 6) as shown in FIG. The particle size and the like are detected for each image of the separated transparent particles (S12) and stored in the memory M. The particle size of the image of the transparent particles detected as described above is used for calculating the particle size distribution, the average value of the particle size, and other statistical values (S13). As described above, the apparatus according to the present embodiment includes not only the shape information relating to the dark image area and the light image area bounded by the boundary line obtained based on the image density information of the captured grayscale image, but also Using the image density information in the direction crossing the boundary line that separates the light image area,
Since the contour extracted by the hillless filter is determined, the contour between the image of the transparent particles 7 and the image of the background can be clearly identified, and in particular, the transparent particles 7 overlapping in the light transmission direction can be identified. It is an apparatus suitable for performing the image processing described above. In the above-described embodiment, an example is disclosed in which the threshold value regarding the image density gradient average value for discriminating whether the extracted contour line is the outer contour line or the inner contour line is manually set and input. However, the present invention is not limited to this. The evaluation criterion (discrimination criterion) of the threshold is determined in advance, and the threshold is automatically changed by a so-called discrimination criterion method in which a threshold that maximizes the threshold is selected. It is also possible to configure.

On the other hand, between the light image area in the transparent particle 7 and the light image area in the background, for example, x
The density distribution of the image density data f (x) in the direction is different. For example, the density distribution shown in FIG. 9A shows an image density distribution in a light image area as a closed area surrounded by the outer contour line, and has a sharp bimodal shape. . On the other hand, the image density distribution in the light image area as a closed area within the inner contour line of the transparent particle 7 shown in (b) is an image related to light transmitted through the particle, and therefore has a gentle mountain-like shape. I have. Each of these image density distributions is stored in the memory M in advance as reference distribution data,
Further, the image density data f (x) for each pixel in the x direction in the light image area surrounded by the contour line extracted this time by the Mar-Hildress filter is used as image density distribution data and paired with these contour lines. Is stored in the memory M (second storage means) to determine which of the reference density data of FIG. 9A or FIG. 9B the image density distribution data in each light image area is similar to. By comparing and determining, it is possible to discriminate whether the extracted ring boundary line is an outer contour line or an inner contour line.

That is, as described above, based on the predetermined reference distribution data relating to the image density distribution data, the boundary between the light transmitting particles and the background, the dark image area in the light transmitting particles and the light Means for realizing the function of discriminating the boundary line of the image area is an example of the second boundary line separation unit according to the present invention. On the other hand, the total value as the statistical value of the image density data for all the pixels in the light image region as the closed region surrounded by the outer contour line is the above-mentioned value for the light transmitted through the transparent particles 7 as described above. The sum differs from the sum of the image density data for all the pixels in the light image area as a closed area surrounded by the inner contour line, and a large value (ie,
(Large value in the bright direction). Therefore, a threshold value (second threshold value) relating to the total value of the image density data for separating the outer contour line and the inner contour line is set in advance and stored in the memory M. An outline is extracted by the Mar-Hildress filter, and a total value of image density data for every pixel in a light image area surrounded by the extracted outline is calculated. Memory M (third
Storage means). These contour lines are classified as outer contour lines when the corresponding sum value is larger than the threshold value, and are classified as inner contour lines when the corresponding sum value is smaller than the threshold value. That is, as described above, the boundary between the light-transmitting particles and the background is determined based on a predetermined second threshold value related to the statistical value of the data relating to the image density.
The means for realizing the function of separating the boundary between the dark image area and the light image area in the light transmitting particles is an example of the third boundary line separation means according to the present invention. In the above example, the total value of the image density data for all pixels in the light image area is used as the statistical value of the data relating to the image density.
However, the present invention is not limited to this. As the statistical value, for example, an average value obtained by dividing the total value by the total number of pixels in the area may be used. In the above-described embodiment, data relating to the image density data f (x) as it is is applied as the image density data and the reference distribution data in the light image area. However, the present invention is not limited to this. First order differential value f '(x) or 2 of data f (x)
The image density data and the reference distribution data for the secondary differential value f ″ (x) may be used. Naturally, the primary differential value of the image density data related to the sum value and the average value used as the statistical value is also used. It is also possible to use a second derivative value.

[0014]

The present invention is configured as described above. Thereby, not only the shape information represented by the dark image area and the light image area separated by the boundary line extracted by the boundary line extraction means but also the image density information relating to the extracted boundary line are used together. A boundary between the light-transmitting particles and the background and a boundary between a dark image region and a light image region in the light-transmitting particles can be distinguished. Therefore, it is possible to clearly distinguish the image of the light-transmitting particles from the image of the background, which is particularly suitable for analyzing the images of the light-transmitting particles overlapping in the light transmitting direction, and is extremely practical.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram showing a transparent particle inspection apparatus according to one embodiment of the present invention.

FIG. 2 is a graph showing a characteristic curve of a Mahl-Hildes filter used for obtaining an outer contour and an inner contour in a grayscale image of a transparent particle captured by the transparent particle inspection apparatus.

3A and 3B are explanatory diagrams for explaining a zero-crossing point method applied to obtain the outer contour and the inner contour, wherein FIG. 3A is a graph showing an image density curve in an x direction;
(B) is a graph showing a first derivative curve of the image density curve in (a), and (c) is a graph showing a second derivative curve of the image density curve in (a).

FIG. 4 is a flowchart showing a processing procedure of operations from imaging of a transparent particle to measurement of a particle diameter and the like by the transparent particle inspection apparatus.

FIG. 5 is a state explanatory view showing a binary image obtained by processing the grayscale image by a Mar-Hildress filter.

FIG. 6 is a graph showing the frequency distribution of the average value of the image density gradient in the direction crossing each outer contour line and each inner contour line over the entire length of the contour line.

FIG. 7 is an explanatory diagram showing a binary image in which a light image region in a transparent particle distinguished from a light image region representing a background by the image processing device of the transparent particle inspection device is treated as a dark image region.

FIG. 8 is a state explanatory view showing an aspect in which an image of each transparent particle is separated from the binary image of FIG. 7 and each particle size is measured.

FIG. 9 is a graph showing a frequency distribution of image densities for each pixel in a light image area used for distinguishing an image between a transparent particle and a background by the transparent particle inspection apparatus according to the second embodiment of the present invention. 5A is a graph showing an image density distribution in a light image area surrounded by the outer contour, and FIG. 5B is a graph showing an image density distribution in a light image area within the inner contour. FIG.

FIG. 10 is an explanatory diagram showing an original image of a grayscale image represented by the respective amounts of light transmitted through transparent particles and received from a background, which are captured by a general transparent particle inspection apparatus.

FIG. 11 is a conceptual diagram showing the concept of a binary image including a dark image region and a light image region obtained from the gray image of FIG.

FIG. 12 is an explanatory diagram showing a binarized image corresponding to FIG. 7 obtained from a grayscale image by a conventional image processing method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Transparent particle inspection apparatus 2 ... Image pick-up part 3 ... Image data calculation part 4 ... Transmission illuminating device 5 ... CCD camera 6 ... Image processing apparatus 7 ... Transparent particle 9 ... Visible part M ... Memory

Claims (3)

(57) [Claims] 1. An image sensor that receives light transmitted through light-transmitting particles and light from a background and outputs a signal corresponding to the amount of received light, and a gray image that generates a gray image from an output signal from the image sensor. An image processing apparatus comprising: a creating unit; and a boundary line extracting unit that extracts a boundary line that divides the grayscale image into a dark image region and a light image region based on the image density information of the grayscale image created. A first storage means for storing image density gradient data in a direction crossing the boundary line in a pair with the extracted boundary line, and the light source based on a predetermined first threshold value relating to the image density gradient data. A boundary line between the transparent particles and the background, and a first boundary line separation unit for separating a boundary line between a dark image region and a light image region in the light transparent particles. Processing of moving light-transmitting particles apparatus. 2. An image sensor for receiving light transmitted through the light-transmitting particles and light from the background and outputting a signal corresponding to the amount of received light, and a gray image for forming a gray image based on an output signal from the image sensor. An image processing apparatus comprising: a creating unit; and a boundary line extracting unit that extracts a boundary line that divides the grayscale image into a dark image region and a light image region based on the image density information of the grayscale image created. A second storage unit for storing image density distribution data in a light image area surrounded by the extracted boundary line in pairs with the extracted boundary line, and based on predetermined reference distribution data relating to the image density distribution data. And a second boundary line discriminating means for separating a boundary line between the light transmitting particles and the background and a boundary line between a dark image region and a light image region in the light transmitting particles. Light transmission characterized by Image processing device for conductive particles. 3. An image sensor for receiving light transmitted through the light-transmitting particles and light from the background and outputting a signal corresponding to the amount of light received, and a gray image for forming a gray image from an output signal from the image sensor. An image processing apparatus comprising: a creating unit; and a boundary line extracting unit that extracts a boundary line that divides the grayscale image into a dark image region and a light image region based on the image density information of the grayscale image created. A third storage means for storing a statistical value of data relating to image density in a light image area surrounded by the extracted boundary line in pairs with the statistical value of data relating to the image density; A third boundary line for separating a boundary line between the light transmitting particles and the background and a boundary line between a dark image region and a light image region in the light transmitting particles based on a predetermined second threshold value. That it has separation means Characteristic image processing device for light transmitting particles.