JP2010115260A - Image processing apparatus, image processing program, and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus, an image processing program, and an image processing method by which a foam region can be detected at a good precision regardless of the shape of the foam. <P>SOLUTION: This image processing apparatus 10 includes: a region dividing section 151 which divides an intracorporeal endoluminal image showing the inside of an intracorporeal lumen into a plurality of regions; a high-frequency feature value calculating section 152 which calculates a luminance high-frequency feature value for each of the plurality of regions divided by the region dividing section 151; and a foam region detecting section 155 which detects the foam region for each of the regions according to the luminance high-frequency feature value for each region. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, an image processing program, and an image processing method for processing an intraluminal image obtained by imaging an internal lumen.

  Patent Document 1 discloses a technique for detecting a bubble region reflected in an image obtained by imaging the inside of a body lumen. In Patent Document 1, first, the edge strength of a pixel in an image is calculated, and a correlation value between the calculated edge strength and a bubble model set in advance based on the feature of the bubble is calculated. And the part with a high correlation with a bubble model is detected as a bubble area | region.

JP 2007-313119 A

  In Patent Literature 1, a bubble model is set by assuming in advance the shape characteristic of a bubble reflected in an image in a body lumen, for example, the outer shape of the bubble such as a circular shape or an elliptical shape. For this reason, there is a problem that bubbles having a shape that does not match the bubble model cannot be detected. In addition, if it is configured to prepare a plurality of foam models assuming various shapes of foam, it is troublesome and the correlation with the prepared foam models must be calculated respectively, which increases the processing time. There was a problem.

  The present invention has been made in view of the above, and an object of the present invention is to provide an image processing apparatus, an image processing program, and an image processing method capable of accurately detecting a bubble region regardless of the shape of the bubble. .

  In order to solve the above-described problems and achieve the object, an image processing apparatus according to an aspect of the present invention is an image processing apparatus that processes an intraluminal image obtained by imaging a body lumen. A region dividing unit that divides the intraluminal image into a plurality of regions, a high frequency feature amount calculating unit that calculates a luminance high frequency feature amount for each of the plurality of regions, and each region based on the luminance high frequency feature amount of each region And a bubble area detecting means for detecting the bubble area.

  There are convex edges of luminance due to the regular reflection component of the illumination light in the outline of the bubble and inside the bubble, and these become high-frequency luminance high-frequency components. According to the image processing apparatus according to this aspect, the high-frequency luminance feature amount is calculated for each region obtained by dividing the intraluminal image, and the bubble region is detected for each region based on the high-frequency luminance feature amount of each region. it can. Therefore, the bubble region can be detected stably regardless of the shape of the bubble, and the detection accuracy of the bubble region can be improved.

  An image processing program according to another aspect of the present invention is an image processing program for causing a computer to process a body lumen image obtained by imaging a body lumen. A region dividing step for dividing the region into regions, a high frequency feature amount calculating step for calculating a luminance high frequency feature amount for each of the plurality of regions, and a bubble region for each region based on the luminance high frequency feature amount of each region And causing the computer to execute a bubble region detection step.

  An image processing method according to another aspect of the present invention is an image processing method for processing a body lumen image obtained by imaging a body lumen, and divides the body lumen image into a plurality of regions. A region dividing step, a high frequency feature amount calculating step for calculating a luminance high frequency feature amount for each of the plurality of regions divided in the region dividing step, and a luminance high frequency feature amount of each region calculated in the high frequency feature amount calculating step. And a bubble region detecting step for detecting a bubble region for each region.

  According to the present invention, a bubble region can be detected stably regardless of the shape of the bubble, and the detection accuracy of the bubble region can be improved.

  DESCRIPTION OF EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiments. Moreover, in description of each drawing, the same code | symbol is attached | subjected and shown to the same part.

(Embodiment 1)
First, the first embodiment will be described. FIG. 1 is a schematic diagram illustrating an overall configuration of an image processing system including the image processing apparatus according to the first embodiment. As shown in FIG. 1, the image processing system captures an image inside a subject 1 (an intraluminal image), and an intraluminal image transmitted wirelessly from the capsule endoscope 3. A receiving device 5 for receiving, an image processing device 10 for processing and displaying the intraluminal image captured by the capsule endoscope 3 based on the intraluminal image received by the receiving device 5, and the like. Is done. For example, a portable recording medium (portable recording medium) 7 is used for transferring image data between the receiving device 5 and the image processing device 10.

  The capsule endoscope 3 has an imaging function, a wireless function, and the like. The capsule endoscope 3 is swallowed from the mouth of the subject 1 and introduced into the subject 1, and sequentially moves through the body lumen. An inner image is taken. The captured intraluminal image is wirelessly transmitted outside the body.

  The receiving device 5 includes receiving antennas A <b> 1 to An that are dispersedly arranged at positions on the body surface corresponding to the passage route of the capsule endoscope 3 in the subject 1. The receiving device 5 receives the image data wirelessly transmitted from the capsule endoscope 3 via the receiving antennas A1 to An. The receiving device 5 is configured so that the portable recording medium 7 can be freely attached and detached, and sequentially stores the received image data in the portable recording medium 7. In this way, the intraluminal image inside the subject 1 captured by the capsule endoscope 3 is accumulated and stored in the portable recording medium 7 in time series by the receiving device 5.

  The image processing apparatus 10 is for a doctor or the like to observe and diagnose an intraluminal image captured by the capsule endoscope 3, and is realized by a general-purpose computer such as a workstation or a personal computer. The image processing apparatus 10 is configured so that the portable recording medium 7 can be freely attached and detached. The image processing apparatus 10 processes the intraluminal image stored in the portable recording medium 7 and, for example, displays it on a display such as an LCD or an EL display. Display sequentially in sequence order.

  The intraluminal image captured by the capsule endoscope 3 and processed by the image processing apparatus 10 shows mucous membranes, contents floating in the intraluminal lumen, bubbles, etc. The part is reflected. The image processing apparatus 10 according to the first embodiment detects a bubble area reflected in the body cavity image. The intraluminal image captured by the capsule endoscope 3 is a color image having a pixel level (pixel value) for each color component of R (red), G (green), and B (blue) at each pixel position. is there.

  Here, the detection principle of a bubble area | region is demonstrated compared with the conventional method demonstrated by background art. 2 and 3 are schematic diagrams illustrating an example of an intraluminal image in which a bubble region is reflected. FIG. 2 shows an intraluminal image in which a plurality of bubbles 21 to 23 whose outer shape is substantially circular are reflected. On the other hand, FIG. 3 shows an intraluminal image in which the bubble 31 is greatly reflected in the lower left region toward the drawing. For example, when attention is paid to the region of the bubble 21 in FIG. 2, high brightness portions 211 and 213 are generated in the contour portion of the bubble 21 and in the bubble 21. Similarly, in the bubble 31 shown in FIG. 3, high brightness portions 311, 313, and 315 are generated in the contour portion of the bubble 31 and in the bubble 31.

  FIG. 4 is a diagram illustrating an example of a foam model 41 used in conventional foam detection (see Patent Document 1). FIG. 4 shows a bubble model 41 having a ring-shaped high-luminance structure portion 411 and a high-luminance structure portion 413 located inside the ring shape. In the conventional method disclosed in Patent Document 1, a bubble region is detected based on a correlation value between the edge strength in the image and the bubble model 41, and the bubbles 21 to 23 shown in FIG. What is similar to the bubble model 41 can be detected.

  On the other hand, it is difficult to detect the bubbles 31 shown in FIG. FIG. 5 is a diagram showing a state where the bubble 31 shown in FIG. 3 is detected using the bubble model 41 of FIG. As shown in FIG. 5, for example, the portion indicated by the broken line 51 can be detected as belonging to the bubble region based on the correlation value with the bubble model 41 indicated by the one-dot chain line in FIG. The portion indicated by is low in correlation with the bubble model 41 and cannot be detected as a bubble region by detection using the bubble model 41.

  In contrast to such a conventional technique, the first embodiment detects a bubble region by paying attention to the luminance high-frequency component of the intraluminal image. FIG. 6 is an explanatory diagram for explaining the detection principle of the bubble region in the first embodiment. In FIG. 6, the bubble region shown in the intraluminal image is a bubble region 61 shown in the uppermost part (a), and the bubble region 61 is detected.

  In Embodiment 1, first, an intraluminal image is divided into regions based on luminance edges. Here, the bubble region 61 has a high-luminance portion 611 of the outline portion and an internal high-luminance portion 613. For this reason, for example, when attention is paid to the line L5 shown in FIG. 6A, the intraluminal image is divided into regions at the boundary position indicated by the alternate long and short dash line.

  Then, a high-frequency luminance feature amount for each divided area is calculated. That is, first, the luminance component of each pixel constituting the intraluminal image is obtained. FIG. 6B shows the luminance component a of each pixel on the line L5. Next, a smoothing process is performed on the luminance component a. FIG. 6C shows the luminance component b after smoothing. Next, the difference between the luminance component a and the smoothed luminance component b is calculated, and only the high luminance component is extracted to obtain the luminance high frequency component ab as shown in FIG.

  Then, a luminance high-frequency feature amount is calculated for each divided region. For example, in FIG. 6E, the average value of the luminance high-frequency component of the pixels constituting the region is calculated and used as the luminance high-frequency feature amount F (x, y). Then, a region where the luminance high frequency feature amount is equal to or greater than a predetermined threshold D5 is set as a bubble region. In FIG. 6 (e), as indicated by hatching in the figure, five areas adjacent on the line L5 are detected as bubble areas.

  According to the method described above, the bubble region can be detected by dividing the intraluminal image in the body based on the luminance edge and calculating the luminance high frequency feature amount for each region. Therefore, the bubble region can be detected stably, and the detection accuracy of the bubble region can be improved. Further, it is not necessary to prepare a bubble model in advance as in the conventional method, and further, processing time does not increase as in the case of preparing a plurality of bubble models assuming various shapes of bubbles. .

  FIG. 7 is a block diagram illustrating a functional configuration of the image processing apparatus 10 according to the first embodiment. In the first embodiment, the image processing apparatus 10 includes an external interface unit 11, an operation unit 12, a display unit 13, a recording unit 14, a calculation unit 15, and a control unit that controls the operation of the entire image processing device 10. 17.

  The external interface unit 11 is for acquiring image data picked up by the capsule endoscope 3 and received by the receiving device 5. For example, the portable recording medium 7 is detachably attached to the portable recording medium. 7 is constituted by a reader device that reads out the image data stored in the memory 7. Image data read from the portable recording medium 7 via the external interface unit 11 is held in the recording unit 14, processed by the calculation unit 15, and displayed on the display unit 13 under the control of the control unit 17. . The acquisition of the image captured by the capsule endoscope 3 is not limited to the configuration using the portable recording medium 7. For example, a server may be separately installed instead of the portable recording medium 7 and an image captured by the capsule endoscope 3 may be stored in advance in this server. In this case, the external interface unit 11 is configured by a communication device or the like for connecting to the server. Then, data communication with the server via the external interface unit 11 may be performed to acquire an image. Alternatively, an image captured by the capsule endoscope 3 may be stored in advance in the recording unit 14 and read from the recording unit 14 to obtain an image.

  The operation unit 12 is realized by, for example, a keyboard, a mouse, a touch panel, various switches, and the like, and outputs an operation signal to the control unit 17. The display unit 13 is realized by a display device such as an LCD or an EL display, and displays various screens including a display screen of an image captured by the capsule endoscope 3 under the control of the control unit 17. .

  The recording unit 14 is realized by various IC memories such as ROM and RAM such as flash memory that can be updated and recorded, an information recording medium such as a built-in or data communication terminal, a CD-ROM, and a reader thereof. A program for operating the image processing apparatus 10 and realizing various functions of the image processing apparatus 10, data used during execution of the program, and the like are stored. For example, the recording unit 14 includes an image processing program 143 for detecting a bubble region from the image data 141 of the body lumen image captured via the external interface unit 11 and the body lumen image to be processed. Stored.

  The calculation unit 15 processes an image captured by the capsule endoscope 3 and performs various calculation processes for detecting a bubble region from the intraluminal image. The calculation unit 15 includes a region dividing unit 151 as a region dividing unit, a high frequency feature amount calculating unit 152 as a high frequency feature amount calculating unit, and a bubble region detecting unit 155 as a bubble region detecting unit. The area dividing unit 151 divides the intraluminal image into a plurality of areas. The high frequency feature amount calculation unit 152 calculates the luminance high frequency feature amount for each region divided by the region division unit 151. The high-frequency feature amount calculation unit 152 includes a high-frequency component calculation unit 153 as a high-frequency component calculation unit and an in-region statistical value calculation unit 154 as a statistical value calculation unit. The high frequency component calculation unit 153 calculates the luminance high frequency component of each pixel constituting the intraluminal image. The intra-region statistical value calculation unit 154 calculates, for each region, the luminance high-frequency component statistical value calculated by the high-frequency component calculation unit 153 for the pixels constituting the region. The bubble region detection unit 155 detects a bubble region for each region based on the luminance high frequency feature amount of each region. The bubble region detection unit 155 includes a high frequency feature amount determination unit 156 as high frequency feature amount determination means. The high frequency feature amount determination unit 156 determines whether the luminance high frequency feature amount of each region is equal to or greater than a predetermined threshold value.

  The control unit 17 is realized by hardware such as a CPU. The control unit 17 configures the image processing apparatus 10 based on image data acquired via the external interface unit 11, operation signals input from the operation unit 12, programs and data stored in the recording unit 14, and the like. Instructions to each unit, data transfer, and the like are performed, and the overall operation of the image processing apparatus 10 is comprehensively controlled.

  FIG. 8 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus 10 according to the first embodiment. The process described here is realized by the operation of each unit of the image processing apparatus 10 according to the image processing program 143 stored in the recording unit 14.

  As shown in FIG. 8, first, the calculation unit 15 acquires an intraluminal image to be processed (step a1). By the processing here, the calculation unit 15 reads and acquires the intraluminal image to be processed from the image data 141 acquired through the external interface unit 11 and recorded in the recording unit 14.

  Subsequently, the area dividing unit 151 executes an area dividing process (step a3). Various methods are known as the region division method. Here, for example, the method disclosed in WO2006 / 080239 is used to perform region division based on the luminance edge of the intraluminal image. To do. FIG. 9 is a flowchart showing a detailed processing procedure of the region division processing.

In the region dividing process, as shown in FIG. 9, the region dividing unit 151 first calculates a luminance component from the intraluminal image (step b1). Specifically, the luminance component Y of each pixel is calculated according to the following equation (1) based on the color components (R, G, B) of each pixel of the intraluminal image.
Y = 0.3 × R + 0.59 × G + 0.11 × B (1)

  Here, the luminance component is calculated according to the calculation formula of the luminance Y in a general YCbCr color space (reference: CG-ARTS Association, Digital Image Processing, 299P), but other values corresponding to this are calculated. It may be used. Alternatively, the G component value of the intraluminal image may be used as the luminance component. This is because the value of the G component is close to the absorption wavelength band of hemoglobin and the sensitivity is easily obtained, so that the structure of the intraluminal image is well represented.

  Subsequently, the region dividing unit 151 calculates edge information of the intraluminal image based on the calculated luminance component Y (step b3). Specifically, a spatial filtering process using a Sobel filter is performed on the luminance component Y of the intraluminal image in the body to perform edge extraction (reference: CG-ARTS Association, Digital Image Processing, 116P). Note that edge extraction may be performed by using a first-order differential filter other than the Sobel filter, for example, a first-order differential filter such as a Prewitt filter, or a second-order differential filter such as a Laplacian filter or a LOG (Laplacian of Gaussian) filter (reference: CG -ARTS Association, Digital Image Processing, 114P).

  Subsequently, the region dividing unit 151 performs a smoothing process on the calculated edge information of the intraluminal image (step b5). This smoothing process is performed for the purpose of noise removal, and can be realized by, for example, a spatial filtering process using a Gaussian filter. A bilateral filter, an averaging filter, or the like may be used (reference: CG-ARTS Association, Digital Image Processing, 108P).

Subsequently, the area dividing unit 151 calculates the downward direction of the luminance gradient in each pixel (step b7). The procedure is briefly described. Here, the luminance value of the target pixel is Px, and the luminance values of eight adjacent pixels adjacent to the target pixel with the target pixel as the center are P_i (i = 1 to 8). First, the difference between the luminance value Px of the target pixel and the luminance value P_i (i = 1 to 8) of the adjacent pixel is calculated according to the following equation (2).
dP = P_i-Px (2)

  Next, an adjacent pixel having a minimum difference dP, that is, a maximum negative value is selected, and a direction from the target pixel to the selected adjacent pixel is set as a downward direction of the luminance gradient. When all the differences dP are positive values, the target pixel is set as a minimum value pixel. The minimum value pixel is a pixel that finally arrives when moving along the downward direction of the luminance gradient. By performing the above processing with all pixels as the target pixel, the downward direction of the luminance gradient in each pixel can be calculated.

  Subsequently, the area dividing unit 151 calculates the coordinates of the minimum value pixel in each pixel (step b9). For example, first, adjacent pixels in the downward direction of the luminance gradient set for the target pixel are selected. Next, adjacent pixels in the downward direction of the luminance gradient set for the selected adjacent pixels are selected. This process is repeated, and the finally reached pixel is set as the minimum value pixel of the target pixel, and the coordinate value is obtained. By performing the above processing using all pixels as the target pixel, the coordinates of the minimum value pixel in each pixel can be calculated.

  Then, the region dividing unit 151 divides the intraluminal image into regions by attaching the same label to the pixels having the same coordinates of the reaching minimum value pixels (step b11). The area of the pixel that reaches the same minimum value pixel is surrounded by a luminance edge. For this reason, as a result of the region division processing, region division with the luminance edge as a boundary can be performed. If the intraluminal image is divided into regions, the process returns to step a3 in FIG. 8, and then proceeds to step a5.

  As another area segmentation method, a known watershed algorithm (reference: Luc Vincent and Pierre Soille. Watersheds in digital spaces: An efficient algorithm based on immersion simulations. Transactionson Pattern Analysis and Machine Intelligence, Vol. 13, No. 6, pp.583-598, June 1991) may be used. The watershed algorithm is a method of dividing an image so that a boundary is formed between water accumulated in different depressions when water is filled in the terrain where the pixel value information of the image is regarded as altitude. For this reason, by applying the watershed algorithm to the edge information, an area division equivalent to the above-described area division can be realized.

  In addition, here, the case where the region is divided based on the luminance edge has been described, but the image may be divided into a grid of a predetermined size without considering the edge. In this case, each processing step described with reference to FIG. 9 becomes unnecessary, and the processing time can be shortened.

  Subsequently, in step a5 shown in FIG. 8, the high frequency feature quantity calculation unit 152 executes a high frequency feature quantity calculation process, and calculates a luminance high frequency feature quantity for each region divided in step a3. FIG. 10 is a flowchart showing a detailed processing procedure of the high-frequency feature amount calculation processing.

  In the high frequency feature amount calculation process, as shown in FIG. 10, first, the high frequency component calculation unit 153 of the high frequency feature amount calculation unit 152 calculates the luminance high frequency component of each pixel based on the luminance component of the intraluminal image. (Step c1). Specifically, first, the luminance component of the intraluminal image (luminance component a in FIG. 6B) is smoothed, and the smoothed image (luminance component b after smoothing in FIG. 6C). ). The filter used for the smoothing process is not limited. For example, a bilateral filter, a Gaussian filter, an averaging filter, or the like can be used. Subsequently, a difference between the luminance component of the intraluminal image and the smoothed image is calculated for each pixel, and the calculated difference is used as the luminance high-frequency component of each pixel (the luminance high-frequency component ab in FIG. 6D). To do. Here, since the luminance high-frequency component of the entire intraluminal image is first calculated, the processing speed can be increased.

  Note that the technique for calculating the luminance high-frequency component is not limited to the exemplified technique. For example, the high-frequency luminance component may be calculated by applying a known high-pass filter process. As a representative example of the high-pass filter processing, for example, first, an image is converted into a frequency space by fast Fourier transform (FFT). And the process which calculates using a window function which passes only a high frequency component, and calculates a high frequency component by performing inverse FFT is known.

  Subsequently, the process of loop A is sequentially performed with the regions divided in step a3 in FIG. 8 as regions of interest (step c3 to step c7). That is, the intra-region statistical value calculation unit 154 calculates the statistical value of the attention region based on the high-frequency luminance component of each pixel calculated in step c1 (step c5). Specifically, first, the absolute value of the value of the high-frequency luminance component in each pixel constituting the region of interest is calculated. Next, an average value of the calculated absolute values is calculated and used as a statistical value in the attention area. In addition, although the case where an average value was calculated as a statistical value was illustrated here, it is not limited to the average value. For example, the maximum value, median value, mode value, and the like may be calculated as statistical values.

  Then, the intra-region statistical value calculation unit 154 performs a loop A process using all regions as the attention region, and calculates a statistical value for each region. The statistical value calculated for each region in this way is set as the luminance high-frequency feature amount of each region (luminance high-frequency feature amount F (x, y) in FIG. 6E). If the processing of the loop A is performed for all the regions, the process returns to step a5 in FIG. 8, and then proceeds to step a7.

  And in step a7 of FIG. 8, the bubble area | region detection part 155 performs a bubble area | region detection process. As described above, the convex edge of the luminance due to the regular reflection component of the illumination light exists in the outline portion of the bubble and the inside of the bubble, and these become luminance high frequency components with high amplitude. For this reason, in the area where bubbles are reflected, the value of the high-frequency luminance feature value calculated in step a5 in FIG. 8 is high. In the bubble region detection process, a region where the luminance high frequency feature amount of each region is equal to or greater than a predetermined threshold value (threshold value D5 in FIG. 6E) is detected and set as a bubble region. The threshold value can be set as appropriate. The threshold value may be a value that is variably set according to a user operation.

  FIG. 11 is a flowchart showing a detailed processing procedure of the bubble area detection process in the first embodiment. In the bubble area detection process, the process of loop B is sequentially performed with the area divided in step a3 in FIG. 8 as the attention area (step d1 to step d9). That is, first, the high frequency feature quantity determination unit 156 of the bubble area detection unit 155 performs threshold processing on the luminance high frequency feature quantity of the attention area (step d3), and if the luminance high frequency feature quantity of the attention area is equal to or greater than a predetermined threshold ( Step d5: Yes), the bubble area detection unit 155 detects the attention area as the bubble area (step d7). Here, since the bubble region can be detected by a simple process of performing threshold processing on the luminance high frequency feature amount of the region of interest, the processing speed can be increased. If the process of loop B is performed for all the regions, the process returns to step a7 in FIG. 8, and then proceeds to step a9.

  In step a9 in FIG. 8, the calculation unit 15 outputs the detection result of the bubble area in the intraluminal image to be processed, and the processing in the calculation unit 15 of the image processing apparatus 10 ends. For example, based on the detection result of the bubble region detected for each region in step a7, the calculation unit 15 sets flag information indicating whether the region is a bubble region for each pixel of the intraluminal image, creates an image, and performs control. The display unit 13 outputs a display via the unit 17.

(Embodiment 2)
Next, a second embodiment will be described. FIG. 12 is a block diagram illustrating a functional configuration of the image processing apparatus 10b according to the second embodiment. In addition, the same code | symbol is attached | subjected about the structure same as the structure demonstrated in Embodiment 1. FIG. As shown in FIG. 12, the image processing apparatus 10b is a control that controls the operation of the external interface unit 11, the operation unit 12, the display unit 13, the recording unit 14b, the calculation unit 15b, and the entire image processing apparatus 10b. Unit 17.

  The recording unit 14b stores image data 141 of the intraluminal image captured through the external interface unit 11 and an image processing program 143b for detecting a bubble region from the intraluminal image to be processed. The

  The calculation unit 15b includes a region dividing unit 151 as a region dividing unit, a color feature amount calculating unit 161b as a color feature amount calculating unit, a luminance feature amount calculating unit 162b as a luminance feature amount calculating unit, and a high-frequency feature. A high frequency feature amount calculation unit 152 as a quantity calculation unit and a bubble region detection unit 155b as a bubble region detection unit are included.

  The area dividing unit 151 divides the intraluminal image into a plurality of areas in the same manner as in the first embodiment. The color feature amount calculating unit 161b calculates a color feature amount for each region divided by the region dividing unit 151. The luminance feature amount calculation unit 162b calculates a luminance feature amount for each region. The high frequency feature amount calculation unit 152 calculates the luminance high frequency feature amount for each region in the same manner as in the first embodiment. Although not shown, the high-frequency feature amount calculation unit 152 includes the high-frequency component calculation unit 153 and the in-region statistical value calculation unit 154 described in the first embodiment.

  The bubble area detection unit 155b is based on the color feature amount calculated by the color feature amount calculation unit 161b, the luminance feature amount calculated by the luminance feature amount calculation unit 162b, and the luminance high frequency feature amount calculated by the high frequency feature amount calculation unit 152. , Detect the bubble area. The bubble area detection unit 155b includes a color feature amount determination unit 163b as a color feature amount determination unit, a luminance feature amount determination unit 164b as a luminance feature amount determination unit, and a high frequency feature amount determination unit as a high frequency feature amount determination unit. 156b. The color feature amount determination unit 163b determines that a region where the color feature amount is within a predetermined range is a non-bubble region. The luminance feature amount determination unit 164b determines that a region where the luminance feature amount is equal to or less than a predetermined threshold is a bubble-free region. The high-frequency feature amount determination unit 156b determines whether or not the luminance high-frequency feature amount is greater than or equal to a predetermined threshold for a region that has not been determined as a bubble-free region by the color feature amount determination unit 163b or the luminance feature amount determination unit 164b.

  FIG. 13 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus 10b according to the second embodiment. The process described here is realized by the operation of each unit of the image processing apparatus 10b according to the image processing program 143b stored in the recording unit 14b. In FIG. 13, the same reference numerals are assigned to the same processing steps as those in the first embodiment.

In the second embodiment, after the region division process (step a3), the color feature amount calculation unit 161b calculates a color feature amount for each region divided in step a3 (step e41). There are various values as the color feature amount. Here, for example, the color differences Cb and Cr in the YCbCr color space are used as the color feature amount. First, based on the color components (R, G, B) of each pixel of the intraluminal image, the color differences Cb, Cr of each pixel are calculated according to the following equations (3), (4).
Cb = 0.17 × R−0.33 × G + 0.50 × B (3)
Cr = 0.50 × R−0.42 × G + 0.08 × B (4)

  Next, the average value of the color difference Cb value of each pixel for each region and the average value of the color difference Cr value of each pixel for each region are calculated and used as the color feature amount of each region. Here, the color feature amount is two-dimensional feature amount data including an average value of Cb and an average value of Cr in each region. Here, although the color difference is used as the color feature amount, the present invention is not limited to this. For example, hue, saturation, color ratio, and the like may be used as the color feature amount. Further, not only the average value but also the maximum value, median value, mode value, etc. may be calculated as the color feature amount.

  Subsequently, the luminance feature amount calculation unit 162b calculates a luminance feature amount for each region divided in step a3 (step e43). Specifically, first, the luminance component Y of the intraluminal image is calculated. Regarding the calculation of the luminance component Y, the method described as the process of step b1 in FIG. 9 in the first embodiment can be used. This value can be used when the luminance component Y has already been calculated in the region division processing. Next, the average value of the luminance component Y value of each pixel for each region is calculated and set as the luminance feature amount of each region. Here, the average value of the luminance component Y is calculated as the luminance feature amount. However, the average value is not limited to the average value. For example, the maximum value, the median value, the mode value, and the like may be calculated as the luminance feature amount. Good.

  In the second embodiment, after the high frequency feature amount calculation process (step a5) performed by the high frequency feature amount calculation unit 152, the bubble region detection unit 155b performs the bubble region detection process (step e7). FIG. 14 is a flowchart showing a detailed processing procedure of the bubble region detection processing in the second embodiment.

  In the bubble area detection process, the process of loop C is sequentially performed with the area divided in step a3 in FIG. 13 as the attention area (step f1 to step f19). That is, first, when the color feature amount determination unit 163b of the bubble region detection unit 155b performs threshold processing on the color feature amount of the attention region (step f3), and the color feature amount of the attention region is within a predetermined range set in advance. (Step f5: Yes), the process proceeds to step f7 to set the attention area as a non-bubble area.

  In the method of calculating the high-frequency luminance feature and discriminating the bubble area, there is a rare case where the high-frequency luminance component resulting from bleeding or redness is erroneously detected as the bubble area. Therefore, the color feature amount is discriminated and an area such as bleeding or redness is defined as a non-foaming area, so that it is excluded from the discrimination target of the foam area, and erroneous detection is suppressed. Specifically, a plurality of sample images are prepared in advance, and in step f7, for example, color feature amounts of areas such as bleeding and redness that are excluded from the determination target of the bubble area as a non-bubble area are acquired as teacher data. Then, based on the acquired color feature amount, the range of the non-bubble region in the color feature amount space is determined in advance.

  Here, the case where the range of the non-foaming region in the color feature amount space is set as a threshold has been described, but the range of the non-foaming region in the color feature amount space is set as a distribution function or a teacher data point. Alternatively, it may be determined whether or not it is within a predetermined range by applying a known statistical determination process, a neighborhood method, or the like.

  If the color feature amount of the attention area is not within the predetermined range set in advance (step f5: No), then the luminance feature amount determination unit 164b performs threshold processing on the luminance feature amount of the attention area (step f9), When the luminance feature amount of the attention area is equal to or less than a predetermined threshold value set in advance (step f11: Yes), the process proceeds to step f7 to set the attention area as a non-bubble area.

  In the method of calculating the high-frequency luminance feature and discriminating the bubble region, there is a rare case where the high-frequency luminance component caused by the noise in the dark portion is erroneously detected as the bubble region. Therefore, the luminance feature amount is discriminated and the noise part is set as a non-foaming region, so that it is excluded from the discrimination target of the foaming region, and erroneous detection is suppressed. Specifically, a plurality of sample images are prepared in advance, and in step f7, for example, the luminance feature amount of a region such as noise that is excluded from the determination target of the bubble region as a non-bubble region is acquired as teacher data. Based on the acquired luminance feature amount, the range of the bubble-free region in the luminance feature amount space is determined in advance.

  When the color feature amount of the attention area is not within the predetermined range (step f5: No), and the luminance feature amount of the attention area is larger than the predetermined threshold (step f11: No), the high frequency feature amount The determination unit 156b performs threshold processing on the luminance high-frequency feature value of the attention area in the same manner as in the first embodiment (step f13). If the luminance high-frequency feature value of the attention area is equal to or greater than the predetermined threshold value (step f15: Yes). The bubble area detection unit 155b detects the attention area as the bubble area (step f17). If the processing of the loop C is performed for all the regions, the process returns to step e7 in FIG. 13, and then proceeds to step e9.

  In step e9 in FIG. 13, the calculation unit 15b outputs the detection result of the bubble region in the intraluminal image to be processed, and ends the processing in the calculation unit 15b of the image processing apparatus 10b.

  As described above, according to the second embodiment, the effect similar to that of the first embodiment is achieved, and the region where the color feature amount is in the color range of bleeding or redness is defined as a non-foam region, and the detection target of the bubble region Therefore, it is possible to reduce false detection of bleeding and redness. Furthermore, based on the luminance feature amount, the noise portion can be excluded from the detection target of the bubble region as a non-bubble region, and erroneous detection for noise occurring in the dark portion can be reduced.

  In the second embodiment described above, first, the color feature amount determination unit 163b performs threshold processing on the color feature amount of each region to determine a bubble-free region, and then the luminance feature amount determination unit 164b performs a color feature amount. The determination unit 163b determines the non-foam area by performing threshold processing on the luminance feature amount of the area that is not determined as the non-foam area. That is, the non-foaming region based on the color feature amount and the non-foaming region based on the luminance feature amount are combined into a non-foaming region.

  On the other hand, the luminance feature amount discriminating unit 164b discriminates the bubble-free region by performing threshold processing on the luminance feature amount of each region, and then the color feature amount discriminating unit 163b performs the bubble-free region by the luminance feature amount discriminating unit 164b. The non-foam area may be determined by performing threshold processing on the color feature amount of the area that has not been determined to be.

  In addition, the color feature amount determination unit 163b performs threshold processing on the color feature amount of each region to determine a non-foam region, and the luminance feature amount determination unit 164b performs threshold processing on the luminance feature amount of each region. The area may be determined.

  Further, only the non-foaming region based on the color feature amount may be determined, and the non-foaming region based on the luminance feature amount may not be determined. Alternatively, only the non-foaming region based on the luminance feature amount may be determined, and the non-foaming region based on the color feature amount may not be determined.

  In addition, the luminance feature amount calculation unit 162b may calculate the luminance feature amount for a region that has not been determined as a bubble-free region by the color feature amount determination unit 163b. In this case, first, the color feature amount calculation unit 161b calculates the color feature amount of each region, and then the color feature amount determination unit 163b performs threshold processing on the color feature amount of each region to determine the bubble-free region. To do. Next, the luminance feature amount calculation unit 162b calculates the luminance feature amount of the region that has not been determined as the bubble-free region by the color feature amount determination unit 163b, and the luminance feature amount determination unit 164b performs threshold processing on the luminance feature amount. Determine the no-bubble area.

  Alternatively, the color feature amount calculation unit 161b may calculate the color feature amount for an area that has not been determined as a bubble-free area by the luminance feature amount determination unit 164b. In this case, first, the luminance feature amount calculation unit 162b calculates the luminance feature amount of each region, and then the luminance feature amount determination unit 164b performs threshold processing on the luminance feature amount of each region to determine a bubble-free region. To do. Next, the color feature amount calculation unit 161b calculates the color feature amount of the region that has not been determined as the bubble-free region by the luminance feature amount determination unit 164b, and the color feature amount determination unit 163b performs threshold processing on the color feature amount. Determine the no-bubble area.

  In the embodiment described above, the case where the bubble region is detected from the intraluminal image captured by the capsule endoscope 3 has been described. However, the image to be processed is captured by the capsule endoscope 3. However, the present invention is not limited to such an image, and can be similarly applied to the case where a bubble region is detected from an image obtained by imaging the inside of a body lumen.

1 is a schematic diagram illustrating an overall configuration of an image processing system including an image processing apparatus according to a first embodiment. It is a schematic diagram which shows an example of the intraluminal image in which a bubble area | region is reflected. It is a schematic diagram which shows the other example of the intraluminal image in which a bubble area | region is reflected. It is a figure which shows an example of the foam model used by the conventional foam detection. It is a figure which shows a mode that the bubble shown in FIG. 3 is detected using the bubble model of FIG. 3 is an explanatory diagram for explaining a detection principle of a bubble region according to Embodiment 1. FIG. 2 is a block diagram illustrating a functional configuration of the image processing apparatus according to Embodiment 1. FIG. 3 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus according to the first embodiment. It is a flowchart which shows the detailed process sequence of an area | region division process. It is a flowchart which shows the detailed process sequence of a high frequency feature-value calculation process. 4 is a flowchart showing a detailed processing procedure of bubble area detection processing in the first embodiment. 6 is a block diagram illustrating a functional configuration of an image processing apparatus according to Embodiment 2. FIG. 10 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus according to the second embodiment. 10 is a flowchart showing a detailed processing procedure of bubble area detection processing in the second embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 3 Capsule endoscope 5 Receiving device A1-An Receiving antenna 7 Portable recording medium 10, 10b Image processing apparatus 11 External interface part 12 Operation part 13 Display part 14, 14b Recording part 141 Image data 143, 143b Image processing program 15 , 15b arithmetic unit 151 region dividing unit 152 high frequency feature amount calculating unit 153 high frequency component calculating unit 154 in-region statistical value calculating unit 155, 155b bubble region detecting unit 156, 156b high frequency feature amount determining unit 161b color feature amount calculating unit 162b luminance feature Quantity calculation unit 163b Color feature quantity discrimination unit 164b Luminance feature quantity discrimination unit 17 Control unit 1 Subject

Claims (8)

An image processing apparatus for processing an intraluminal image obtained by imaging the internal lumen,
Region dividing means for dividing the intraluminal image into a plurality of regions;
High-frequency feature amount calculating means for calculating a luminance high-frequency feature amount for each of the plurality of regions;
Based on the luminance high frequency feature amount of each region, a bubble region detection means for detecting a bubble region for each region,
An image processing apparatus comprising: The high frequency feature amount calculating means includes:
High-frequency component calculating means for calculating a luminance high-frequency component of each pixel constituting the intraluminal image;
For each region, a statistical value calculation unit that calculates a statistical value of the high-frequency luminance component calculated by the high-frequency component calculation unit for the pixels constituting the region;
With
The image processing apparatus according to claim 1, wherein a statistical value calculated for each region is used as a luminance high-frequency feature amount for each region.   The image processing apparatus according to claim 1, wherein the bubble area detection unit includes a high frequency feature amount determination unit that determines a region where the luminance high frequency feature amount is equal to or greater than a predetermined threshold as a bubble region. Color feature amount calculating means for calculating a color feature amount for each region;
A color feature amount discriminating means for discriminating a region in which the color feature amount is within a predetermined range as a non-foam region;
With
The image processing according to any one of claims 1 to 3, wherein the bubble region detection unit detects a bubble region for a region that has not been determined as a non-bubble region by the color feature amount determination unit. apparatus. A luminance feature amount calculating means for calculating a luminance feature amount for each region;
A luminance feature amount determining means for determining a region where the luminance feature amount is equal to or less than a predetermined threshold as a non-foam region;
With
5. The image processing according to claim 1, wherein the bubble area detection unit detects a bubble area for an area that has not been determined as a non-bubble area by the luminance feature amount determination unit. apparatus. A luminance feature amount calculating means for calculating a luminance feature amount of an area that has not been determined as a non-foam area by the color feature amount determining means;
A luminance feature amount determining means for determining a region where the luminance feature amount is equal to or less than a predetermined threshold as a non-foam region;
With
The image processing apparatus according to claim 4, wherein the bubble region detection unit detects a bubble region for a region that has not been determined as a non-bubble region by the luminance feature amount determination unit. An image processing program for causing a computer to process a body lumen image obtained by imaging the body lumen,
A region dividing step of dividing the intraluminal image into a plurality of regions;
A high frequency feature amount calculating step for calculating a luminance high frequency feature amount for each of the plurality of regions;
Based on the luminance high frequency feature amount of each region, a bubble region detection step for detecting a bubble region for each region,
An image processing program for causing the computer to execute. An image processing method for processing a body lumen image obtained by imaging the body lumen,
A region dividing step of dividing the intraluminal image into a plurality of regions;
A high-frequency feature amount calculating step for calculating a luminance high-frequency feature amount for each of a plurality of regions divided in the region dividing step;
A bubble region detection step for detecting a bubble region for each region based on the luminance high frequency feature amount of each region calculated in the high frequency feature amount calculation step;
An image processing method comprising:

Images