KR102380560B1 - Corneal Ulcer Region Detection Apparatus Using Image Processing and Method Thereof - Google Patents

Abstract

The present invention relates to a corneal ulcer detection apparatus and method based on image processing. According to the present invention, an image input unit that receives an eye image obtained by photographing the cornea of a subject to be examined, extracts a region of interest from the input eye image, and corresponds to the background using RGB values of each pixel included in the extracted region of interest A preprocessor that performs image pre-processing so that the boundary between the region of interest and the region corresponding to the ulcer is distinguished, derives a threshold value according to the distribution of pixel values in the region of interest An ulcer area detector that detects an ulcer area by applying a flood-fill algorithm that expands the area corresponding to the ulcer, and an outline for the detected ulcer area are generated, and the generated outline is applied to the original eye image. It includes an output unit for displaying and outputting.
As described above, according to the present invention, it is possible to increase the objective and accuracy in detecting the ulcer area by expanding the area using the threshold value derived through the Otsu method around the reference point selected by the random reference point selection algorithm. In addition, according to the present invention, objective results can be derived by minimizing user intervention, and medical treatment results can be provided to medical staff and patients with improved reliability for the purpose of treatment.

Description

Corneal Ulcer Region Detection Apparatus Using Image Processing and Method Thereof

The present invention relates to an apparatus and method for detecting corneal ulcers based on image processing, and more particularly, to a cornea for segmenting a region of interest from an original image captured by the cornea, and detecting the ulcer included in the region of interest in units of pixels It relates to a device for detecting an ulcer and a method therefor.

Image segmentation is recognized as an important process for early diagnosis and analysis of diseases in the medical field. It is used for volume measurement and diagnosis for the detection of diseases in areas that cannot be identified with the .

Among them, since diseases in the field of ophthalmology are directly related to daily sensory organs, early diagnosis through imaging is essential before proceeding with hasty diagnosis and treatment.

Among various ophthalmic diseases, corneal ulcers are diseases that occur in the corneal epidermis and can be divided into bacterial and fungal types depending on the route of infection. Diagnosis of classification and subsequent treatment process is made. At this time, the medical staff diagnoses the size of the ulcer with the naked eye, and the problem is that since the ulcer is measured by subjective judgment, there is a problem that objective data accurately measuring the ulcer area cannot be presented to the patient.

The technology underlying the present invention is disclosed in Korean Patent Application Laid-Open No. 10-2020-0049195 (published on May 8, 2020).

As described above, according to the present invention, it is to provide a corneal ulcer detection apparatus and method for segmenting a region of interest from an original image of the cornea and detecting the ulcer included in the region of interest in units of pixels.

According to an embodiment of the present invention for achieving this technical problem, in the apparatus for detecting corneal ulcers based on image processing, an image input unit for receiving an ocular image obtained by photographing the cornea of a subject to be examined, a region of interest from the input ocular image A preprocessor that performs image preprocessing so that the boundary between the region corresponding to the background and the region corresponding to the ulcer is distinguished by using the RGB values of each pixel included in the extracted region of interest; and pixel values in the region of interest An ulcer area detector that derives a threshold value according to the distribution and detects the ulcer area by applying a flood-fill algorithm that expands the area corresponding to the ulcer around the reference pixel using the derived threshold value; and and an output unit for generating a contour line for the detected ulcer region and displaying the generated contour line on the original eye image.

The preprocessor may convert each pixel corresponding to the region of interest to a gray scale form, and then perform histogram smoothing on the region of interest converted to the gray scale form.

The preprocessor may convert the grayscale form using the following equation.

The preprocessor applies a filter having an arbitrary size to the region of interest to which the histogram smoothing has been applied, arranges pixel values in the filter in ascending or descending order, and then uses an intermediate pixel value of a plurality of pixel values to reduce noise in the region of interest. The boundary of the area corresponding to the ulcer can be simplified by removing and gamma-correcting treatment.

The ulcer region detection unit sets a threshold value according to the distribution of pixel values in the region of interest using the Otsu algorithm, and then acquires a reference point using coordinate information of each pixel included in the region of interest, and the obtained reference point A region corresponding to the ulcer may be expanded by comparing a pixel difference value between a reference pixel corresponding to , and an adjacent pixel and the threshold value.

The ulcer region detection unit clusters pixels included in the region of interest into a region corresponding to an ulcer and a region corresponding to a background according to the Otsu algorithm, and the dispersion value of the region corresponding to the ulcer and the dispersion of the region corresponding to the background A pixel value having a maximum difference in values may be set as the threshold value.

The variance value may be calculated using the following equation.

는 임의의 픽셀이 궤양에 해당하는 영역에 속할 확률이고,

은 임의의 픽셀이 배경에 해당하는 영역에 속할 확률이고,

은 궤양에 해당하는 영역내 픽셀값의 평균을 나타내고,

은 배경에 해당하는 영역내 픽셀값의 평균을 나타내며,

은 전체 영상의 픽셀값의 평균을 나타낸다. here,

is the probability that any pixel belongs to the area corresponding to the ulcer,

is the probability that a random pixel belongs to the region corresponding to the background,

represents the average of pixel values in the area corresponding to the ulcer,

represents the average of pixel values in the area corresponding to the background,

represents the average of pixel values of the entire image.

The ulcer region detection unit generates an array storing coordinate information of each pixel included in the region of interest, and then performs a histogram to randomly select one pixel from among a plurality of pixels corresponding to the pixel value with the highest frequency. It can be selected and set as the reference pixel.

The output unit determines whether an undetected region is included in the detected ulcer region, and if the determination result includes an undetected region, generates an inverted image having the same size as the region of interest and inverted background color, A hole corresponding to an undetected area may be filled by applying a flood-fill algorithm having a threshold value of 0 at a point corresponding to each corner of the inverted image.

The output unit alternately performs an erosion operation and a dilation operation corresponding to a morphological operation on the ulcer region to remove noise included in the boundary portion of the ulcer region and smooth the boundary portion, and Canny Edge Detection An algorithm can be used to generate the contour of the ulcer area.

In addition, according to another embodiment of the present invention, in the method for detecting a corneal ulcer using an image processing-based corneal ulcer detection device, the step of receiving an eye image obtained by photographing the cornea of a subject to be examined, the interest from the input eye image extracting a region and performing image preprocessing so that the boundary between the region corresponding to the background and the region corresponding to the ulcer is distinguished using the RGB values of each pixel included in the extracted region of interest; pixels in the region of interest Detecting the ulcer area by deriving a threshold value according to the value distribution, and applying a flood-fill algorithm that expands the area corresponding to the ulcer around the reference pixel using the derived threshold value; and detection and generating an outline for the ulcerated region, and outputting the ulcer region by displaying the generated outline on the original of the eye image.

As described above, according to the present invention, it is possible to increase the objective and accuracy in detecting the ulcer area by expanding the area using the threshold value derived through the Otsu method around the reference point selected by the random reference point selection algorithm. In addition, according to the present invention, objective results can be derived by minimizing user intervention, and medical treatment results can be provided to medical staff and patients with improved reliability for the purpose of treatment.

1 is a configuration diagram for explaining a corneal ulcer detection apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a corneal ulcer detection method using the corneal ulcer detection apparatus according to an embodiment of the present invention.
FIG. 3 is a flowchart for explaining step S230 shown in FIG. 2 .
FIG. 4 is a diagram illustrating a state in which a region of interest is converted to grayscale in step S231 shown in FIG. 3 .
FIG. 5 is a diagram illustrating a state in which histogram smoothing is applied to a region of interest in step S232 shown in FIG. 3 .
FIG. 6 is a diagram illustrating a method of setting an intermediate pixel value in step S233 shown in FIG. 3 .
FIG. 7 is a diagram illustrating a state in which noise included in a region of interest is removed using an intermediate value set in step S233 shown in FIG. 3 .
FIG. 8 is a diagram illustrating a state in which gamma correction is applied to a region of interest in step S234 shown in FIG. 3 .
9 is a flowchart for explaining step S240 shown in FIG.
FIG. 10 is a diagram for explaining the structure of an array for storing position information of pixels in step S242 shown in FIG. 9 .
11A and 11B are diagrams for explaining a method of expanding an area according to a flood fill algorithm in step S243 shown in FIG. 9 .
12 is a flowchart for explaining S250 illustrated in FIG. 2 .
13 is a diagram illustrating a state in which a corrected image is output by filling an undetected area in step S251 shown in FIG. 12 .
14 is a diagram illustrating a state in which a morphology operation is performed in step S252 shown in FIG. 12 .
15 is a view showing the result of displaying the contour line detected in step S253 shown in FIG. 12 on the original image.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In this process, the thickness of the lines or the size of the components shown in the drawings may be exaggerated for clarity and convenience of explanation.

In addition, the terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to intentions or customs of users and operators. Therefore, definitions of these terms should be made based on the content throughout this specification.

Hereinafter, a corneal ulcer detection apparatus according to an embodiment of the present invention will be described in more detail with reference to FIG. 1 .

1 is a configuration diagram for explaining a corneal ulcer detection apparatus according to an embodiment of the present invention.

As shown in FIG. 1 , the corneal ulcer detection apparatus 100 according to an embodiment of the present invention includes an image input unit 110 , a preprocessor 120 , an ulcer region detection unit 130 , and an output unit 140 . .

First, the image input unit 110 receives an ocular image obtained by photographing the cornea of a test subject.

Next, the preprocessor 120 simplifies the surface of the eyeball image in order to segment the ulcer region from the received eyeball image. In other words, the preprocessor 120 extracts the ROI designated by the user. Then, the preprocessor 120 uses the RGB values of each pixel included in the extracted region of interest to be divided into a region corresponding to the background and an region corresponding to an ulcer in the order of grayscale conversion, histogram smoothing, median value filter, and gamma correction. Perform pre-processing.

Then, the ulcer region detection unit 130 detects the ulcer region by applying a flood-fill algorithm to the region of interest on which the pre-processing has been performed. In other words, the ulcer area detection unit 130 sets any one randomly selected pixel within the area corresponding to the ulcer as the reference pixel, and corresponds to the ulcer if the difference between the reference pixel and the adjacent pixel is within a preset threshold. expand to the area At this time, the threshold value is calculated by the Otsu algorithm. The ulcer area detection unit 130 detects the ulcer area by repeatedly performing a flood-fill algorithm until the ulcer area is maximally expanded.

Finally, the output unit 140 generates a contour line for the detected ulcer region, and displays the generated contour line on the original eye image and outputs it. That is, the output unit 140 determines whether an undetected area is included in the detected ulcer area, and when the undetected area is included, the process of filling the undetected area is performed. Then, the ulcer area detection unit 130 smoothes the boundary portion of the ulcer area to generate a contour line, and displays the generated contour line on the original image of the eyeball and outputs it.

Hereinafter, a method for detecting a corneal ulcer using the corneal ulcer detection apparatus 100 will be described in more detail with reference to FIGS. 2 to 15 .

2 is a flowchart illustrating a corneal ulcer detection method using the corneal ulcer detection apparatus according to an embodiment of the present invention.

As shown in FIG. 2 , the apparatus 100 for detecting corneal ulcers according to an embodiment of the present invention receives an eye image obtained by photographing the cornea of a test subject ( S210 ).

In other words, corneal ulcer is a disease that occurs in the epidermis of the cornea, and since the patient cannot directly observe the site of the disease, the diagnosis is made based on images taken with close-up equipment. Accordingly, the image input unit 110 receives the eyeball image photographed through the close-up photographing device.

Next, the preprocessor 120 extracts the region set by the user as the region of interest ( S220 ).

The corneal ulcer detection method according to an embodiment of the present invention detects a corneal ulcer by performing image segmentation in a state where the user's intervention is minimized. do. Then, the preprocessor 120 extracts the ROI through a region designated by the user or a set threshold value.

When the extraction of the region of interest is completed in step S220, the preprocessor 120 performs a preprocessing process on the region of interest (S230).

Hereinafter, step S230 will be described in more detail with reference to FIGS. 3 to 8 .

3 is a flowchart for explaining step S230 shown in FIG. 2 , FIG. 4 is a view showing a state in which the region of interest is converted to grayscale in step S231 shown in FIG. 3 , and FIG. 5 is S232 shown in FIG. It is a diagram showing a state in which the histogram smoothing is applied to the region of interest in step S233, FIG. 6 is a diagram showing a method of setting an intermediate pixel value in step S233 shown in FIG. It is a diagram illustrating a state in which noise included in the region of interest is removed using an intermediate value, and FIG. 8 is a diagram illustrating a state in which gamma correction is applied to the region of interest in step S234 shown in FIG. 3 .

First, as shown in FIG. 3 , the preprocessor 120 according to the embodiment of the present invention converts the region of interest into a gray scale form ( S231 ).

In other words, the preprocessor 120 converts the region of interest into grayscale using Equation 1 below.

That is, the preprocessor 120 multiplies the R, G, and B values of the pixels included in the region of interest by a specific constant to map the pixel values to values between 0 and 255. Then, as shown in FIG. 4 , the region of interest having the color value has the same luminance as the original image, and is converted to grayscale while maintaining the shading of the color.

Next, the preprocessor 120 performs a histogram smoothing process on the grayscale-converted region of interest (S232).

In other words, the value of the pixel included in the region of interest changed to grayscale represents the contrast value. At this time, if the distribution of light and dark values is skewed to one side or the distribution is not uniform, the boundary for the ulcer area cannot be clearly distinguished. Accordingly, the preprocessor 120 applies histogram smoothing to clearly distinguish the region corresponding to the ulcer and the region corresponding to the background.

In other words, the preprocessor 120 generates a histogram using pixel values of the ROI. Then, the preprocessor 120 obtains an accumulated value, that is, the frequency count based on the acquired histogram, and normalizes the ROI using the obtained frequency count.

)를 산출한다. At this time, the preprocessor 120 uses a mapping function (

) is calculated.

는 히스토그램의 누적값이고,

는 누적 히스토그램의 최소값이고, M 및 N은 관심영역의 크기값을 나타낸다. 또한, L은 정규화된 누적 히스토그램 범위의 크기값을 의미하므로 회색조 영상에서는 256의 값을 가진다. here,

is the cumulative value of the histogram,

is the minimum value of the cumulative histogram, and M and N represent the size values of the ROI. In addition, since L means the size value of the normalized cumulative histogram range, it has a value of 256 in the grayscale image.

Next, the preprocessor 120 maps the ROI to the ROI using the calculated mapping function.

8이고, 히스토그램의 누적값이 34이고, 누적 히스토그램의 최소값이 1이라고 가정한다. 그러면, 전처리부(120)는 수학식 2에 의해 133.5의 사상함수를 획득할 수 있다. 그 다음, 전처리부(120)는 획득한 사상함수를 이용하여 픽셀값이 48인 모든 픽셀에 133.5라는 값으로 맵핑한다. 그러면, 도 5에 도시된 바와 같이, 전처리부(120)는 명암 대비가 높은 영상을 획득할 수 있다. For example, if the size of the region of interest is 8

It is assumed that 8, the cumulative value of the histogram is 34, and the minimum value of the cumulative histogram is 1. Then, the preprocessor 120 may obtain a mapping function of 133.5 by Equation (2). Next, the preprocessor 120 maps all pixels having a pixel value of 48 to a value of 133.5 by using the obtained mapping function. Then, as shown in FIG. 5 , the preprocessor 120 may acquire an image with high contrast.

When step S232 is completed, the preprocessor 120 removes the noise included in the region of interest by using the intermediate pixel value (S233).

The region of interest on which the histogram smoothing has been performed may be divided into a region corresponding to an ulcer and a region corresponding to a background according to contrast. However, since noise is included in the boundary of the region corresponding to the ulcer, the preprocessor 120 removes the noise by applying a median filter to the region of interest. In detail, the preprocessor 120 applies a filter having an arbitrary size to the region of interest and sorts pixel values in the filter in ascending or descending order. Then, the preprocessor 120 removes noise in the ROI by using a pixel value located in the middle among the aligned pixel values.

For example, as shown in FIG. 6 , it is assumed that a filter having a size of 3×3 is applied to a region of interest having a size of 4×4. Then, the preprocessor 120 performs the pixel values included in the filter, that is, “4”, “4”, “5”, “4”, “3”, “6”, “3”, “1”, “2”. " can be arranged in ascending or descending order to obtain a pixel value "4" located in the middle. Then, the preprocessor 120 maps the obtained intermediate pixel value to the region extracted by the filter.

As shown in FIG. 7 , the preprocessor 120 repeats the process of applying the intermediate pixel value to obtain an image from which noise has been removed from the contour line of the ulcer region.

Next, the preprocessor 120 simplifies the boundary by performing a gamma correction process on the area corresponding to the ulcer ( S234 ).

In other words, the preprocessor 120 performs a correction operation using a gamma correction filter so that the difference between the region corresponding to the background and the region corresponding to the ulcer included in the region of interest can be more clearly indicated.

The preprocessor 120 calculates a gamma value according to Equation 3 below, and non-linearly transforms the line by using the calculated gamma value.

는 관심영역에서 (i,j)에 위치한 픽셀의 보정된 밝기값을 나타내고,

는 픽셀값을 나타낸다. here,

represents the corrected brightness value of the pixel located at (i, j) in the region of interest,

represents a pixel value.

Accordingly, as shown in FIG. 8 , the preprocessor 120 simply outputs the bright part of the ROI, that is, the part corresponding to the ulcer, using the calculated gamma value.

When the pre-processing process is completed according to step S230, the ulcer area detection unit detects the ulcer area by applying a flood-fill algorithm (S240).

Hereinafter, step S240 will be described in more detail with reference to FIGS. 9 to 11B .

9 is a flowchart for explaining step S240 shown in FIG. 2 , FIG. 10 is a diagram for explaining the structure of an arrangement for storing pixel position information in step S242 shown in FIG. 9 , FIGS. 11A and 11B are It is a diagram for explaining a method of expanding an area according to the flood fill algorithm in step S243 shown in FIG. 9 .

As shown in FIG. 9 , the ulcer region detection unit 130 according to an embodiment of the present invention sets a threshold value according to the distribution of pixel values in the region of interest using the Otsu algorithm ( S241 ).

The Otsu algorithm is an algorithm that can derive an appropriate threshold according to the distribution of pixel values in an image. Accordingly, the ulcer region detection unit 130 clusters pixels included in the region of interest into two regions according to the Otsu algorithm, and sets a threshold value using a variance value of each clustered region. In detail, the ulcer area detection unit 130 classifies the area corresponding to the ulcer and the area corresponding to the background by using a specific pixel value as a threshold based on the histogram. That is, the threshold value represents a reference value for dividing the area corresponding to the ulcer and the area corresponding to the background. is the optimal threshold.

In this case, the variance value is calculated using Equation 4 below.

는 임의의 픽셀이 궤양에 해당하는 영역에 속할 확률이고,

은 임의의 픽셀이 배경에 해당하는 영역에 속할 확률이고,

은 궤양에 해당하는 영역내 픽셀값의 평균을 나타내고,

은 배경에 해당하는 영역내 픽셀값의 평균을 나타내며,

은 전체 영상의 픽셀값의 평균을 나타낸다. here,

is the probability that any pixel belongs to the area corresponding to the ulcer,

is the probability that a random pixel belongs to the region corresponding to the background,

represents the average of pixel values in the area corresponding to the ulcer,

represents the average of pixel values in the area corresponding to the background,

represents the average of pixel values of the entire image.

으로 풀이된다. 즉, 궤양영역 검출부(130)는 궤양에 해당하는 영역의 픽셀 평균값과 배경에 해당하는 영역의 픽셀 평균값의 차이가 최대가 되어 분산 값이 최대가 될 때의 기준 값을 최적의 임계값으로 설정한다. According to Equation 4, the variance value is

is interpreted as That is, the ulcer region detection unit 130 sets the reference value when the difference between the average pixel value of the region corresponding to the ulcer and the pixel average value of the region corresponding to the background becomes the maximum and the dispersion value becomes the maximum as the optimal threshold value. .

When the threshold value setting is completed in step S241, the ulcer area detection unit 130 sets a reference pixel to expand the area corresponding to the ulcer using a flood-fill algorithm (S242).

In other words, the flood-fill algorithm is an area expansion algorithm that widens the area around a specified reference value in a two-dimensional or more array. It is an algorithm that expands an area, sets the expanded pixel as a pixel again as a reference point, and repeats until the area is no longer expanded. Therefore, in order to perform a flood-fill algorithm, it is necessary to set a starting reference pixel.

Looking at the method of setting the reference pixel, first, the ulcer region detector 130 obtains a histogram H in the region of interest.

)가 각 픽셀값의 확률(

)이 된다. 관심영역에서는 명도가 낮은 배경에 해당하는 영역과 명도가 높은 조명영역을 제외하면 궤양에 해당하는 영역에 속한 픽셀의 값이 가장 높은 확률을 가지게 된다. 따라서, 궤양영역 검출부(130)는 배경에 해당하는 영역

과 조명영역

을 제거하여 수학식 5와 같이 새로운 히스토그램(H')을 획득한다. As shown in FIG. 10 , the ulcer region detector 130 creates an array storing coordinate information of each pixel included in the region of interest. At this time, the number of pixels stored in each array (

) is the probability of each pixel value (

) becomes In the region of interest, the value of the pixel belonging to the region corresponding to the ulcer has the highest probability, except for the region corresponding to the background with low brightness and the illuminated region with high brightness. Accordingly, the ulcer region detection unit 130 is a region corresponding to the background.

and lighting area

to obtain a new histogram (H') as in Equation (5).

Then, the ulcer area detector 130 obtains the pixel value with the highest frequency based on the acquired new histogram H' by using Equation (6).

The ulcer region detector 130 randomly selects one pixel from among a plurality of pixels corresponding to the acquired pixel value.

When step S242 is completed, the ulcer area detector 130 expands the area corresponding to the ulcer by comparing the pixel difference value between the selected reference pixel and the adjacent pixel with the threshold value obtained in step S241 ( S243 ).

The ulcer region detection unit 130 performs flood-fill in four or eight directions, up, down, left, and right around the selected reference pixel.

For example, as shown in FIG. 11A , it is assumed that a flood-fill region is expanded in four directions and the threshold value is 3. As shown in FIG. Then, the ulcer region detection unit 130 detects neighboring pixels with a pixel value of “1” as the center, that is, pixels having “6”, “2”, “3”, and “5”, respectively. A difference value with respect to an adjacent pixel is calculated, and a region is extended in a direction of a pixel having a pixel difference value less than 3 from the reference pixel. Here, the extended pixels correspond to pixels having pixel values “2” and “3” in FIG. 11A .

In addition, as shown in FIG. 11B , it is assumed that the flood-fill area is expanded in 8 directions and the threshold value is 3. Then, the ulcer region detection unit 130 calculates a difference value from a neighboring pixel in all directions around a reference pixel having a pixel value of “1”, and expands the region in the direction of a pixel having a difference value less than 3. Here, the expanded pixel corresponds to pixels having pixel values “3”, “2” and “1” in FIG. 11B .

The ulcer area detection unit 130 repeatedly performs flood-fill until the area is no longer expanded, and detects the ulcer area in a state where the expansion is completed.

When step S240 is completed, the output unit 140 generates a contour line for the detected ulcer region, and displays the generated contour line on the original eye image (S250).

Hereinafter, step S250 will be described in more detail with reference to FIGS. 12 to 15 .

12 is a flowchart for explaining S250 shown in FIG. 2 , FIG. 13 is a view showing a state in which the corrected image is output by filling the undetected area in step S251 shown in FIG. 12 , and FIG. 14 is shown in FIG. It is a view showing a state in which the morphology operation is performed in step S252 shown, and FIG. 15 is a view showing the result of displaying the contour line detected in step S253 shown in FIG. 12 on the original image.

As shown in FIG. 12 , the output unit 130 first determines whether an undetected area is included in the detected ulcer area, and if the undetected area is included as a result of the determination, the hole corresponding to the undetected area is filled. (S251).

In other words, depending on the image, the contrast value of the center of the ulcer may not be detected as if there was a hole compared to other parts. Accordingly, the output unit 130 determines whether an undetected area is included in the detected ulcer area.

When it is determined that the undetected region is included, the output unit 130 generates an inverted image having the same size as the ROI and inverted background color. In addition, the output unit 130 expands the area by applying a flood-fill algorithm having a threshold value of 0 at points corresponding to corners of the inverted image. The output unit 130 fills the hole remaining inside the ulcer area by repeatedly performing flood-fill until the area is no longer expanded. Then, as shown in FIG. 13 , the output unit 130 fills in the hole remaining inside the ulcer area, so that a more accurate result image can be derived.

Next, the output unit 130 removes noise included in the boundary portion of the ulcer region and performs a smoothing process (S252).

The output unit 130 performs a morphological operation of alternately performing an erosion operation for reducing the range of the ulcer area and a dilation operation for expanding the range of the ulcer area, thereby removing noise at the boundary without changing the size of the ulcer area.

In detail, the output unit 130 removes noise between the area corresponding to the background and the ulcer area by first performing an erosion operation and then an expansion operation. Then, the output unit 130 fills a small gap in the ulcer area by performing an erosion operation after performing an expansion operation. As shown in FIG. 14 , the output unit 130 provides a smoothing effect to the boundary portion of the ulcer region through repeated execution of the operation and fills in minute gaps therein, thereby deriving a result suitable for detecting the boundary of the ulcer region.

Finally, the output unit 130 detects a contour line from the ulcer region on which the morphological calculation has been completed, and displays the detected canal line on the original eye image (S253).

As shown in FIG. 15 , the output unit 130 detects the contour of the ulcer area using a Canny Edge Detection algorithm. Then, the output unit 130 displays the detected outline on the original of the eyeball image and outputs it.

As described above, according to the present invention, it is possible to increase the objective and accuracy in detecting the ulcer area by expanding the area using the threshold value derived through the Otsu method around the reference point selected by the random reference point selection algorithm. In addition, according to the present invention, objective results can be derived by minimizing user intervention, and medical treatment results can be provided to medical staff and patients with improved reliability for the purpose of treatment.

Although the present invention has been described with reference to the embodiment shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible therefrom. will be. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the following claims.

100: corneal ulcer detection device
110: video input unit
120: preprocessor
130: ulcer area detection unit
140: output unit

Claims (20)

In the corneal ulcer detection device based on image processing,
An image input unit for receiving an ocular image obtained by photographing the cornea of a subject to be examined;
A region of interest is extracted from the inputted eye image, and the image preprocessing process is performed so that the boundary between the region corresponding to the background and the region corresponding to the ulcer is distinguished using the RGB values of each pixel included in the extracted region of interest. preprocessing unit,
A threshold value is derived according to the distribution of pixel values in the region of interest, and a flood-fill algorithm is applied to expand the region corresponding to the ulcer around the reference pixel using the derived threshold to detect the ulcer region. an ulcer area detection unit, and
and an output unit for generating an outline for the detected ulcer region, and displaying the generated outline on an original eye image. According to claim 1,
The preprocessor is
A corneal ulcer detection apparatus for converting each pixel corresponding to the region of interest into a gray scale form, and then performing histogram smoothing on the region of interest converted to the gray scale form. 3. The method of claim 2,
The preprocessor is
A corneal ulcer detection device that converts to a grayscale form using the following equation:

3. The method of claim 2,
The preprocessor is
After applying a filter having an arbitrary size to the region of interest to which the histogram smoothing has been applied, the pixel values in the filter are arranged in ascending or descending order, and then noise in the region of interest is removed using an intermediate pixel value of a plurality of pixel values, and gamma A corneal ulcer detection device that simplifies the boundary of the area corresponding to the ulcer by correcting it. According to claim 1,
The ulcer area detection unit,
After setting a threshold according to the distribution of pixel values in the region of interest using the Otsu algorithm, a reference point is obtained using coordinate information of each pixel included in the region of interest, and a reference pixel corresponding to the obtained reference point is A corneal ulcer detection apparatus for expanding an area corresponding to an ulcer by comparing the pixel difference value between adjacent pixels with the threshold value. 6. The method of claim 5,
The ulcer area detection unit,
According to the Otsu algorithm, pixels included in the region of interest are clustered into an area corresponding to an ulcer and an area corresponding to the background, and the difference between the variance value of the area corresponding to the ulcer and the variance value of the area corresponding to the background is maximum. A corneal ulcer detection device for setting the pixel value to be the threshold value. 7. The method of claim 6,
The variance value is
Corneal ulcer detection device calculated using the following equation:

here,

is the probability that any pixel belongs to the area corresponding to the ulcer,

is the probability that a random pixel belongs to the region corresponding to the background,

represents the average of pixel values in the area corresponding to the ulcer,

represents the average of pixel values in the area corresponding to the background,

represents the average of pixel values of the entire image. 6. The method of claim 5,
The ulcer area detection unit,
An array for storing the coordinate information of each pixel included in the region of interest is created, and then a histogram is performed to randomly select one pixel from among a plurality of pixels corresponding to the pixel value with the highest frequency and set it as a reference pixel Corneal ulcer detection device. According to claim 1,
the output unit,
It is determined whether an undetected region is included in the detected ulcer region, and if the determination result includes an undetected region, an inverted image having the same size as the region of interest and inverted background color is generated,
A corneal ulcer detection device that fills a hole corresponding to an undetected area by applying a flood-fill algorithm with a threshold value of 0 at a point corresponding to each corner of the inversion image. 10. The method of claim 9,
the output unit,
By alternately performing an erosion operation and a dilation operation corresponding to the morphological operation on the ulcer region, noise included in the boundary portion of the ulcer region is removed and the boundary portion is smoothed, and the Canny Edge Detection algorithm is used to A corneal ulcer detection device that creates the outline of the ulcer area. In the corneal ulcer detection method using a corneal ulcer detection device based on image processing,
receiving an ocular image obtained by photographing the cornea of a test subject;
A region of interest is extracted from the inputted eye image, and the image preprocessing process is performed so that the boundary between the region corresponding to the background and the region corresponding to the ulcer is distinguished using the RGB values of each pixel included in the extracted region of interest. step to do,
A threshold value is derived according to the distribution of pixel values in the region of interest, and a flood-fill algorithm is applied to expand the region corresponding to the ulcer around the reference pixel using the derived threshold to detect the ulcer region. step, and
and generating an outline for the detected ulcer area, and outputting the ulcer area by displaying the generated outline on an original eye image. 12. The method of claim 11,
The step of performing the pre-processing is,
converting each pixel corresponding to the region of interest into a gray scale form; and
and performing histogram smoothing on the region of interest converted to the grayscale form. 13. The method of claim 12,
The step of converting to the grayscale form is,
A corneal ulcer detection method that calculates each pixel value using the following equation:

13. The method of claim 12,
The step of performing the pre-processing is,
applying a filter having an arbitrary size to the region of interest on which the histogram smoothing has been performed, arranging the pixel values in the filter in ascending or descending order, and then removing noise in the region of interest using an intermediate pixel value of a plurality of pixel values , And
and gamma-correcting the region of interest from which the noise has been removed to simplify the boundary of the region corresponding to the ulcer. 12. The method of claim 11,
The step of detecting the ulcer area comprises:
setting a threshold according to the distribution of pixel values in the region of interest using the Otsu algorithm;
obtaining a reference point using coordinate information of each pixel included in the region of interest; and
and comparing a pixel difference value between a reference pixel corresponding to the obtained reference point and an adjacent pixel with the threshold value to expand an area corresponding to the ulcer. 15. The method of claim 14,
The step of detecting the ulcer area comprises:
According to the Otsu algorithm, pixels included in the region of interest are clustered into an area corresponding to an ulcer and an area corresponding to the background, and the difference between the variance value of the area corresponding to the ulcer and the variance value of the area corresponding to the background is maximum. A corneal ulcer detection method for setting the pixel value to be the threshold value. 17. The method of claim 16,
The variance value is
Corneal ulcer detection method calculated using the following formula:

here,

is the probability that any pixel belongs to the area corresponding to the ulcer,

is the probability that a random pixel belongs to the region corresponding to the background,

represents the average of pixel values in the area corresponding to the ulcer,

represents the average of pixel values in the area corresponding to the background,

represents the average of pixel values of the entire image. 16. The method of claim 15,
The step of obtaining the reference point is
An array for storing the position information of each pixel included in the region of interest is created, and then a histogram is performed to randomly select one pixel from among a plurality of pixels corresponding to the pixel value with the highest frequency and set it as a reference pixel A method for detecting corneal ulcers. 12. The method of claim 11,
The step of outputting the ulcer region comprises:
determining whether an undetected area is included in the detected ulcer area, and if the determination result includes an undetected area, generating an inverted image having the same size as the area of interest and inverted background color; and
and filling a hole corresponding to an undetected area by applying a flood-fill algorithm having a threshold value of 0 at each corner point of the inversion image. 19. The method of claim 18,
The step of outputting the ulcer region comprises:
The erosion operation and dilation operation corresponding to the morphological operation are alternately performed on the ulcer region to remove noise included in the boundary part of the ulcer region and smooth the boundary part, and the ulcer using Canny Edge Detection algorithm Corneal ulcer detection method further comprising the step of generating an outline of the region.

Images