JP2009136655A - Image processing device, image processing method, and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing device, an image processing method, and an image processing program which allow the speedy determnation of image areas to be processed. <P>SOLUTION: The image processing device includes: a DCT coefficient-obtaining part 51 which regarding the imaging areas in a predetermined range included in an interested image, obtains a DCT coefficient as frequency information representing a plurality of frequency components of the image areas; a generator for feature quantity of DCT 52 which calculates a power value representing the strength of an AC component of the DCT coefficient as a power value representing a strength of one frequency component out of a plurality of frequency components represented by the frequency information and which generates a feature quantity of DCT indicating a feature of the image areas to which each DCT coefficient correspond from this power value; and an area-determining part 53 which determines areas imaged in each imaging part based on the feature quantity of DCT. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, an image processing method, and an image processing program for performing processing for determining the type of an area captured in an image portion of a predetermined range of each image.

  2. Description of the Related Art Conventionally, a technique for discriminating an area of an image has a wide range of applications such as use for separating a main object in an image from a background, and various methods have been proposed. In particular, the method for determining the area of an image based on the feature amount of a set of pixels in a predetermined range has a larger amount of information than the method for determining the area based on a feature amount in units of pixels. Improvement can be expected.

  Here, a representative feature amount of a set of pixels is a texture feature amount. A texture is a repeated pixel value pattern. In general, an image processing apparatus performs numerical processing by two-dimensional Fourier transform, a co-occurrence matrix, or the like in order to handle a texture feature amount (see, for example, Non-Patent Document 1).

Digital Image Processing Editorial Board, “Digital Image Processing”, Second Edition, CG-ARTS Association, March 2006, p.192-195

  By the way, the conventional image processing apparatus has a problem that it cannot perform region discrimination based on the texture feature amount at a high speed because the calculation load for the calculation by the two-dimensional Fourier transform and the co-occurrence matrix is very large.

  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 that can quickly determine the type of an area captured in a processing target image.

  In order to solve the above-described problems and achieve the object, an image processing apparatus according to an aspect of the present invention provides a frequency representing a plurality of frequency components of an image portion of a predetermined range included in the target image. An acquisition unit for acquiring information, and a power value indicating the intensity of one frequency component among a plurality of frequency components represented by the frequency information is calculated for each predetermined frequency band, and the frequency information is calculated based on the power value. A feature amount generation unit that generates a feature amount indicating the feature of the associated image portion; and an area determination unit that determines a type of an area captured in the image portion based on the feature amount. Features.

  According to this aspect, the power value indicating the intensity of one frequency component among the plurality of frequency components represented by the frequency information is calculated, and the feature amount of each image portion is generated based on the power value. Can be obtained by a simple process. Therefore, there is an effect that the type of the region captured in the processing target image can be determined at high speed.

  An image processing method according to another aspect of the present invention includes an acquisition step of acquiring frequency information representing a plurality of frequency components of an image portion of a predetermined range included in a target image, and the frequency information represents A power value indicating the intensity of one frequency component among a plurality of frequency components is calculated for each predetermined frequency band, and a feature amount indicating a feature of an image portion associated with the frequency information based on the power value is calculated. A feature amount generating step to be generated; and a determination step of determining a type of a region imaged in the image portion based on the feature amount.

  An image processing program according to still another aspect of the present invention includes an acquisition step of acquiring frequency information representing a plurality of frequency components of an image portion of a predetermined range included in a target image, and the frequency information A power value indicating the intensity of one frequency component among a plurality of frequency components to be expressed is calculated for each predetermined frequency band, and a feature amount indicating a feature of an image portion associated with the frequency information based on the power value And a determination step of determining a region captured in the image portion based on the feature amount.

  It should be noted that any combination of the above-described constituent elements and a representation of the present invention converted between a system, a recording medium, etc. are also effective as another aspect of the present invention.

  ADVANTAGE OF THE INVENTION According to this invention, the technique which can perform area | region discrimination | determination at high speed can be provided.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of an image processing apparatus according to the invention will be described with reference to the accompanying drawings. As an image processing apparatus in each embodiment and each modification, an image processing apparatus that uses an in-vivo image captured by a capsule endoscope and compressed by DCT encoding as a processing target image will be described as an example. However, the image that can be processed by the image processing apparatus according to the present invention is not limited to the in-vivo image compressed by DCT encoding. Further, the present invention is not limited by the present embodiment. In the description of the drawings, the same or corresponding parts are denoted by the same reference numerals.

  Here, a system that captures and stores an in-vivo image using a capsule endoscope will be described. Capsule endoscopes are equipped with an imaging device and a data transmission device. After being swallowed by a subject and before being spontaneously discharged, the capsule endoscope is moved through a tubular organ such as the stomach, small intestine, large intestine, etc. In-vivo images are captured at a predetermined imaging rate while moving, and image data of the captured in-vivo images are sequentially transmitted to a receiving device installed outside the body of the subject. The number of in-vivo images that the capsule endoscope captures in the subject includes the imaging rate (usually 2 to 4 frames / sec) and the stay time in the subject of the capsule endoscope (usually 8 hours = 2880 sec). It is shown by multiplying and becomes a large number of sheets of about 6 to 120,000 sheets. For this reason, the in-vivo image captured by the capsule endoscope is normally compressed using DCT encoding and stored in a portable storage medium included in the receiving device.

(Embodiment 1)
The image processing apparatus according to the first embodiment will be described with reference to FIG. FIG. 1 is a block diagram illustrating a schematic configuration of the image processing apparatus 1 according to the first embodiment. The image processing apparatus 1 includes an input unit 10 that inputs various types of information, a storage unit 20 that stores various types of information including in-vivo images, an output unit 30 that outputs various types of information, and a control unit that controls the operation of each unit of the image processing apparatus 1. 40. Furthermore, the image processing apparatus 1 includes an image processing unit 50 that performs an area determination process for processing an in-vivo image and determining the type of the area captured in each image portion of the in-vivo image.

  The input unit 10 is realized by a keyboard, a mouse, a data communication interface, and the like. The input unit 10 receives input of various information directly from an observer, and stores in-vivo image information from portable storage media such as various memory cards, CDs, and DVDs. Accept input.

  The storage unit 20 is realized by a hard disk, a ROM, a RAM, and the like, and stores various information such as various processing programs executed by the control unit 40 and processing results. The storage unit 20 also stores in-vivo image storage unit 21 for storing in-vivo image compression data compressed by DCT encoding and in-vivo image restoration data decoded and restored by the image processing unit 50; And a threshold value storage unit 22 that stores a threshold value used in the area determination process performed by the processing unit 50.

  The output unit 30 is realized by a liquid crystal display or the like, and displays various information such as a result of the region discrimination process performed by the image processing unit 50 and an in-vivo image. The output unit 30 displays a GUI (Graphical User Interface) screen that requests the observer to input various processing information.

  The control unit 40 is realized by the CPU, and controls processing and operation of each unit included in the image processing apparatus 1 by executing various processing programs stored in the storage unit 20. In particular, the control unit 40 executes the image processing program to cause the image processing unit 50 to perform region determination processing and cause the output unit 30 to display the result of region determination processing.

  The image processing unit 50 is realized by a CPU, for example, and includes a DCT coefficient acquisition unit 51, a DCT feature amount generation unit 52, and a region determination unit 53, and performs image processing on the in-vivo image stored in the in-vivo image storage unit 21. .

  The DCT coefficient acquisition unit 51 corresponds to an acquisition unit that acquires frequency information, and is calculated as the frequency information of the image portion in a predetermined range when decoding the compressed data of the in-vivo image compressed by the DCT encoding method. DCT coefficients to be obtained are acquired. When the in-vivo image is displayed on the output unit 30, the control unit 40 controls the DCT coefficient acquisition unit 51 to restore the compressed image and acquire the original image of the in-vivo image.

  Here, the DCT coefficient will be described. The DCT coefficient is frequency information obtained by discrete cosine transform for each image portion of 8 × 8 pixels, and indicates a frequency component of each image portion. (Hereinafter, the image portion of 8 × 8 pixels is referred to as a pixel block.) The DCT coefficient of each pixel block is a set of 64 coefficients represented by an 8 × 8 matrix. As shown in FIG. 2, the 8 × 8 matrix of DCT coefficients is configured by values indicating a DC (direct current) component and 63 values indicating an AC (alternating current) component, and the DC component is the upper left of the 8 × 8 matrix. Arranged in the corner, the AC component is arranged in the other part. As shown in FIG. 3, from the upper left to the lower right in the 8 × 8 matrix, AC components having low-frequency frequency characteristics to AC components having high-frequency frequency characteristics are arranged in order. In JPEG processing, for each pixel block, a total of three DCT coefficients of a DCT coefficient of a luminance component (Y value) and two DCT coefficients of a color difference component (U value, V value) are calculated, and each DCT coefficient is calculated. The DC component indicates a value corresponding to the average luminance or average color difference of the pixel block.

  The DCT feature value generation unit 52 corresponds to a feature value generation unit that generates a feature value of each image portion. As a power value of one frequency component among a plurality of frequency components represented by frequency information, the DCT feature value generation unit 52 is configured for each predetermined frequency band. Then, the power value of the AC component of the DCT coefficient is calculated, and a DCT feature value, which is a feature value indicating the feature of each pixel block, is generated based on this power value. The power value is information indicating the intensity of each frequency component. In the first embodiment, the power value indicates the intensity of the AC component in each frequency band, and does not include information indicating the texture directionality. FIG. 4 is a diagram showing the frequency band to which each AC component belongs and the power value of each frequency band. In FIG. 4, the same numbers are assigned to the matrix of AC components belonging to the same frequency band. As shown in FIG. 4, the AC component is parallel to the diagonal line connecting the lower left corner and the upper right corner of the 8 × 8 matrix for each AC component belonging to each of the 14 frequency bands H1 to H14. Are classified into AC components arranged on the 14 lines. In the embodiment, a case where the frequency band is divided into 14 frequency bands will be described. However, the present invention is not limited to this, and may be divided into an arbitrary number of bands as another example. In FIG. 4, the frequency band to which the AC component indicated as “1” in the upper left belongs to the frequency band H1 that is the lowest frequency band, and the frequency band to which the AC component indicated as “14” in the lower right belongs. This is a frequency band H14 that is the highest frequency band. The DCT feature value generation unit 52 calculates the power values F1 to F14 of the AC components in the frequency bands H1 to H14 for each pixel block, and generates a DCT feature value based on the power values F1 to F14.

  The region discriminating unit 53 discriminates the type of the region imaged in each pixel block based on the DCT feature amount of each pixel block. The region discriminating unit 53 discriminates the type of a region such as a bubble region, a lesion region, a bleeding region, a mucous membrane region, a mucus region, or a content region such as feces among the in-vivo regions imaged by each pixel block.

  Each pixel block has a different color and texture depending on whether the imaged area is, for example, a lesion area, a bleeding area, a mucosal area, a mucus area, or a contents area. In addition, the image part in which the bubble area caused by the contents and moisture in the body is imaged contains more edge information than other areas due to the reflection of illumination light at the time of imaging, and the bubble area such as the mucous membrane area The brightness distribution and the frequency distribution are different from the image portion in which other areas other than the image area are captured. Since the DCT feature amount is generated based on the AC component of the DCT coefficient including luminance information and color difference information, the region discriminating unit 53 can discriminate the region imaged in each pixel block using the DCT feature amount. .

  FIG. 5 is a flowchart illustrating an image processing procedure performed by the image processing apparatus 1. As shown in FIG. 5, the control unit 40 first reads compressed data of the in-vivo image stored in the in-vivo image storage unit 21 (step S101), and causes the DCT coefficient acquisition unit 51 to decode the compressed data to each pixel. The DCT coefficient of the block is acquired (step S102). Thereafter, the control unit 40 controls the DCT feature value generation unit 52 to generate a DCT feature value for each pixel block (step S103), and controls the area determination unit 53 to capture an image on each pixel block. A process of determining the type of area is performed (step S104). Thereafter, the control unit 40 outputs the result of the area determination process to the output unit 30 (step S105). Through the processing in steps S <b> 101 to S <b> 105, the control unit 40 determines the area captured in each image portion of the in-vivo image, and presents the result of the area determination process to the in-vivo image observer.

  FIG. 6 is a flowchart showing a procedure of processing performed by the DCT feature quantity generation unit 52 in the processing of step S103. As illustrated in FIG. 6, the DCT feature quantity generation unit 52 acquires the value of the AC component of the DCT coefficient for each pixel block under the control of the control unit 40 (step S201). Thereafter, the DCT feature quantity generation unit 52 calculates the number of AC components having a value of 0 for each frequency band as the power value of each frequency band, and sets each set of power values F1 to F14 in all frequency bands H1 to H14. The DCT feature amount of the pixel block is set (step S202). In this case, the smaller the power value, the greater the intensity of the AC component, and the larger the power value, the smaller the intensity of the AC component. Through the processing in steps S201 to S202, the DCT feature quantity calculation unit 52 generates a DCT feature quantity for each pixel block using the AC component of the DCT coefficient.

  Note that the DCT feature quantity generation unit 52 may generate the DCT feature quantity by acquiring the AC component values of all the DCT coefficients of the Y value, U value, and V value, and the Y value, U value, or V value. The DCT feature value may be generated by acquiring the value of the AC component of any one of the DCT coefficients. For example, when a bubble region is determined, since the bubble region is characterized by a luminance component, the DCT feature value generation unit 52 may generate a DCT feature value using only the AC component of the DCT coefficient of the Y value. Further, the DCT feature quantity generation unit 52 may use a set of power values in a predetermined frequency band as the DCT feature quantity instead of the set of power values F1 to F14 in all frequency bands H1 to H14.

  FIG. 7 is a flowchart showing the procedure of the area determination process performed by the area determination unit 53 in step S104. As illustrated in FIG. 7, the region determination unit 53 acquires the DCT feature amount of each pixel block of the entire image under the control of the control unit 40 (step S301). After that, the region discriminating unit 53 compares the power value in a predetermined frequency band with the threshold value among the DCT feature values, and discriminates the type of the region imaged in each pixel block based on the comparison result (step S302). ). Note that the threshold value and the frequency band of the power value to be compared with the threshold value are stored in advance in the threshold value storage unit 22 for each area to be determined. Through the processing in steps S301 to S302, the region determining unit 53 determines the type of region captured in each pixel block based on the DCT feature amount.

  FIG. 8 schematically shows the ratio of the AC component having a value of 0 for each frequency band H1 to H14 with respect to the DCT coefficient of the Y value of the pixel block in which the bubble region and other regions other than the bubble region are imaged. FIG. The other areas are a lesion area, a bleeding area, a mucosa area, a contents area, a mucus area, and the like.

  As shown in FIG. 8, in both the bubble region and other regions, the ratio of the AC component having a value of 0 increases as the frequency band is increased. However, since the bubble region includes a lot of edge components, the ratio of the AC component having a value of 0 in the frequency bands H3 to H8 is lower than that in the other regions. For this reason, in the frequency bands H3 to H8, the power value of the bubble region is smaller than the power values of the other regions. When determining the bubble region, the region determination unit 53 compares a power value in a predetermined frequency band among the power values F3 to F8 with a threshold value for determining the bubble region, and the power value is a threshold value. If smaller, the area captured by the pixel block corresponding to the DCT feature value is determined as the bubble area.

  Since the distribution of the power value differs depending on the area captured in each pixel block, not limited to the bubble area, the area determination unit 53 captures an image in each pixel block using a threshold set according to the distribution of the power value. Processing to determine the type of area is performed. Note that the frequency band of the power values that the area determination unit 53 compares with the threshold is also set in advance according to the distribution of the power values.

  The threshold value may be one or plural for each area to be determined. For example, when determining a bubble region, the region determination unit 53 compares the power value F5 with a threshold value and determines that the region imaged in the pixel block is a bubble region when the power value F5 is smaller than the threshold value. Alternatively, the power values F4 and F7 may be compared with the corresponding threshold values, and if both power values are smaller than the respective threshold values, it may be determined that the region imaged in the pixel block is a bubble region.

  Since the image processing apparatus 1 according to the first embodiment performs region discrimination processing based on the feature amount generated based on the power value of the AC component of the DCT coefficient, it is based on two-dimensional Fourier transform, co-occurrence matrix, and the like. Compared to a conventional image processing apparatus that performs an area determination process using arithmetic processing, the area determination process can be performed at high speed.

  Further, the image processing apparatus 1 generates the feature amount of each image portion based on the DCT coefficient calculated when the decompression process of the compressed image compressed by the DCT encoding method is performed. Is compressed by the DCT encoding method, it is possible to shorten the calculation processing time for generating the feature amount and to perform the discrimination processing at a higher speed than the conventional image processing apparatus.

  Further, since the image processing apparatus 1 performs the region determination process based on the DCT feature amount generated using the AC component of the DCT coefficient, the region determination process can be performed based on the frequency information and the color difference information. Further, it is possible to determine a region where the color distribution is wide and the color tends to change with high accuracy. For example, if the area discrimination process is performed using only the color information as the feature value because the color distribution of the bubble area is wide and overlaps with the color distribution of other areas other than the bubble area, the bubble area can be identified with high accuracy. There were cases where it was difficult to do. However, the bubble region includes a lot of edge components, has a texture from medium frequency to high frequency, and has a large AC component value from medium frequency band to high frequency band when compared with other regions, the image processing device 1 The bubble region discriminating process can be performed with high accuracy by evaluating the AC component in the medium frequency band to the high frequency band.

  Further, since the image processing apparatus 1 performs region determination using the power value, the image processing device 1 can perform the region determination with higher accuracy than when performing the region determination processing using the AC component value of the coding coefficient as it is as the feature amount of the pixel block. it can. Since the AC component includes information on the directionality of the texture, if it is difficult to determine a predetermined region due to the difference in texture directionality, if the region determination is performed using the AC component value as it is as the feature amount of the pixel block There is. However, since the image processing apparatus 1 performs region determination using a set of power values not including texture direction information as a feature amount, the region determination can be performed with high accuracy.

  In the first embodiment, the DCT feature value generation unit 52 may use a set of the power values F1 to F14 and the DC component value as the DCT feature value. In this case, for example, when determining the bubble region, the region determination unit 53 compares a preset threshold value with the DC component of the Y value, and if the Y value DC component is equal to or greater than the threshold value, the region block 53 is imaged on the pixel block. It is determined that there is a possibility that the region is a bubble region, and a process of comparing the power value with the threshold value is performed. In addition, when determining the region having the color characteristic, the region determining unit 53 compares the DC component of the U value or the V value with the threshold value, and the region captured in the pixel block is the region where the determination is desired. After it is determined that there is a possibility, a process of comparing the power value and the threshold value is performed. In this case, since the image processing apparatus 1 performs the region determination process using the DC component in addition to the power value, the accuracy of the region determination is improved. Note that after the region determination process based on the power value, the region determination process may be performed based on the DC component.

(Modification 1)
In the first embodiment, the number of AC components having a value of 0 among the AC components belonging to each frequency band of the DCT coefficient is calculated as the power value of one frequency component included in the frequency information. In Example 1, the number of AC components whose values are within a predetermined range is calculated as the power value.

  The image processing apparatus according to the first modification has the same configuration as the image processing apparatus 1 according to the first embodiment. FIG. 9 is a flowchart illustrating a procedure of processing performed by the DCT feature quantity generation unit 52 in step S103 in the case of the first modification. As illustrated in FIG. 9, the DCT feature quantity generation unit 52 acquires the value of the AC component of the DCT coefficient for each pixel block under the control of the control unit 40 (step S211). Thereafter, the DCT feature quantity generation unit 52 calculates the number of AC components whose values fall within a predetermined range for each frequency band as the power value of each frequency band, and the power values F1 to F14 of all frequency bands H1 to H14. Is set as the DCT feature amount of each pixel block (step S212). Through the processing in steps S211 to S212, the DCT feature value generation unit 52 generates the DCT feature value of each pixel block using the AC component of the DCT coefficient.

  Note that the predetermined range is a range of values including the 0 value and close to the 0 value, for example, a range in which the value of the AC component is −10 or more and 10 or less. Further, the range may be set to other values by experiment. The ratio of the AC component whose value falls within a predetermined range is the same as the distribution of the ratio of the AC component whose value is 0 shown in FIG. 8, for example, the AC component whose value is 0 is included. In a frequency band with a large ratio, a ratio including an AC component that is a value near zero is also large. For this reason, the image processing apparatus according to the first modification can perform the area determination process in the same manner as the image processing apparatus 1 according to the first embodiment.

(Modification 2)
In Embodiment 1 and Modification 1, the number of AC components satisfying a predetermined condition among the AC components of the DCT coefficient is calculated as the power value of one frequency component included in the frequency information. 2 calculates an average value of absolute values of AC components belonging to each frequency band as a power value.

  The image processing apparatus according to the second modification has the same configuration as that of the image processing apparatus 1 according to the first embodiment. FIG. 10 is a flowchart showing a procedure of processing performed by the DCT feature quantity generation unit 52 in step S103 in the case of the second modification. As illustrated in FIG. 10, the DCT feature quantity generation unit 52 acquires the value of the AC component of the DCT coefficient for each pixel block under the control of the control unit 40 (step S221). Thereafter, the DCT feature quantity generation unit 52 calculates the average value of the AC component values for each frequency band as the power value of each frequency band, and sets the power values F1 to F14 of all frequency bands H1 to H14. Is the DCT feature value of each pixel block (step S222). In this case, the higher the power value, the higher the AC component strength, and the smaller the power value, the lower the AC component strength. Through the processing in steps S221 to S222, the DCT feature value generation unit 52 generates the DCT feature value of each pixel block using the AC component of the DCT coefficient.

  The value of the AC component of the DCT coefficient can take both positive and negative values depending on the directionality of the texture. For this reason, the DCT feature quantity generation unit 52 obtains information on the intensity of the AC component in each frequency band H1 to H14 as a power value by converting the value of the AC component into an absolute value and calculating an average value. A DCT feature is generated.

  FIG. 11 shows the average value of the absolute values of the AC component values belonging to each of the frequency bands H1 to H14 with respect to the AC component of the DCT coefficient of the Y value of the image portion in which the bubble region and other regions other than the bubble region are imaged. That is, it is a figure which shows distribution of the power values F1-F14 concerning this modification 2. FIG. As shown in FIG. 11, in both the bubble region and other regions, the power value decreases as the frequency band increases. However, since the bubble region includes a lot of edge components, the power value is larger in the frequency bands H3 to H8 than the other regions. When determining the bubble region, the region determination unit 53 compares a power value in a predetermined frequency band among the power values F3 to F8 with a threshold value for determining the bubble region, and the power value is a threshold value. If larger, the region captured by the pixel block corresponding to the DCT feature value is determined as the bubble region.

  Since the power value distribution varies depending on the imaged region, not limited to the bubble region, the region determination unit 53 uses a threshold value set in advance according to the power value distribution, as in the first embodiment. A process of determining the area imaged in each pixel block is performed. For this reason, the image processing apparatus according to the modified example 2 can perform the area determination process as in the first embodiment.

(Modification 3)
In the second modification, the average value of the AC component values of the DCT coefficient is calculated as the power value of one frequency component included in the frequency information. However, in the third modification, the AC value of the AC component is used as the power value. A value obtained by weighting the average absolute value of the values for each frequency band is calculated.

  The image processing apparatus according to the third modification has a configuration similar to that of the image processing apparatus 1 according to the first embodiment. FIG. 12 is a flowchart illustrating a procedure of processing performed by the DCT feature quantity generation unit 52 in step S103 in the case of the third modification. As illustrated in FIG. 12, the DCT feature quantity generation unit 52 acquires the value of the AC component of the DCT coefficient for each pixel block under the control of the control unit 40 (step S231). Thereafter, the DCT feature quantity generation unit 52 calculates a value obtained by weighting the average value of the AC component values belonging to each frequency band according to each frequency band as the power value of each frequency band, A set of power values F1 to F14 in the bands H1 to H14 is set as a DCT feature amount of each pixel block (step S232). Since the absolute value of the AC component value tends to become smaller as the frequency band to which it belongs becomes a higher frequency band, the greater the average value of the absolute value of the AC component in the higher frequency band, the greater the weighting. Through the processes in steps S231 to S232, the DCT feature value generation unit 52 generates a DCT feature value of each pixel block using the AC component of the DCT coefficient.

  In the case of the third modification, in step S104, the region determination unit 53 compares an average value of a plurality of predetermined power values with a threshold value, and determines a region captured by each pixel block based on the comparison result. To do.

Here, the power values F1 to F14 according to the third modification example are given by assuming that the average values of the AC component values in each frequency band are A1 to A14, and the weights corresponding to each frequency band are W1 to W14. It is shown in the following formula (1).
F1 = A1 × W1,..., F14 = A14 × W14 (1)

In this case, the average value F of the power values used when the region discriminating unit 53 compares the threshold value with the threshold value is expressed by the following expression (2) when the frequency band of the power value used for the comparison with the threshold value is a frequency band H3 to H8. .
F = (F3 + F4 + F5 + F6 + F7 + F8) / 6 (2)

FIG. 13 shows the distribution of the power value for each frequency band according to the third modification calculated based on the AC component of the DCT coefficient of the Y value of the image portion in which the bubble region and other regions other than the bubble region are imaged. FIG. As shown in FIG. 13, for example, when weighting is performed so that the power value of an image portion obtained by imaging other regions other than the bubble region is substantially constant regardless of the frequency band, The power value is larger in the frequency bands H3 to H8 than the power values in other regions. In this case, the average value BF _3-8 of the power values F3 to F8 in the bubble region is larger than the average value MF _3-8 of the power values F3 to F8 in the other regions. For this reason, for example, when the average value of the power values F3 to F8 is larger than a preset threshold value, the region determination unit 53 determines that the region captured in the pixel block is a bubble region.

  Since the distribution of power values differs depending on the imaged area, not limited to the bubble area, the area determination unit 53 compares the threshold value set according to the distribution of power values with the average value of predetermined power values, A process of determining the area imaged in the pixel block is performed. In the image processing apparatus according to the third modification, the degree of weighting, the frequency band of the power value used for comparison with the threshold value, and the threshold value are determined in advance and stored in the threshold value storage unit 22 according to the area to be determined. Has been.

  The image processing apparatus according to the third modification determines a region based on a result obtained by comparing a weighted average value of absolute values of AC component values, that is, a value reflecting the AC component intensity of a plurality of frequency bands with a threshold value. Therefore, the region determination process can be performed with high calculation load of the same degree as compared with the case where the region determination process is performed by comparing the power value of one frequency band and the threshold value.

  In the first embodiment and the first to third modifications, different power values are calculated to generate different DCT feature amounts. However, in the image processing apparatus according to the embodiment of the present invention, the first and A DCT feature value setting unit that selects and sets a DCT feature value used for region discrimination processing from the four types of DCT feature values described in the first to third modifications may be provided.

(Embodiment 2)
In the first embodiment, the region determination process is performed by comparing the power value and the threshold value. In the second embodiment, the power value distribution and the teacher data distribution in the feature space are as follows. The region determination process is performed by comparing the two.

  FIG. 14 is a diagram illustrating a schematic configuration of the image processing apparatus 2 according to the second embodiment. The image processing device 2 includes a storage unit 60, a control unit 70, and an image processing unit 80 instead of the storage unit 20, the control unit 40, and the image processing unit 50 included in the image processing device 1. The storage unit 60 includes a teacher data storage unit 62 instead of the threshold storage unit 22 included in the storage unit 20. The image processing unit 80 includes an area determination unit 83 instead of the area determination unit 53 provided in the image processing unit 50. The control unit 70 controls the processing and operation of each unit of the image processing apparatus 2, similarly to the control unit 40. The other configuration of the image processing apparatus 2 is the same as the other configuration of the image processing apparatus 1.

  The teacher data storage unit 62 stores a plurality of DCT feature quantities classified as classes for each region imaged in the pixel block as teacher data. The area determination unit 83 performs an area determination process for each pixel block based on the DCT feature value of each pixel block and the DCT feature value of teacher data.

  The procedure of the image processing performed by the image processing apparatus 2 is the same as the procedure of the image processing performed by the image processing apparatus 1 (FIG. 5: Steps S101 to S105), but the area determination process of Step S104 is different. FIG. 15 is a flowchart illustrating the procedure of the process performed by the area determination unit 83 in step S104. As illustrated in FIG. 15, the region determination unit 83 acquires the DCT feature value of each pixel block of the entire image under the control of the control unit 70 (step S <b> 401), and reads teacher data from the teacher data storage unit 62. (Step S402). Thereafter, the region determination unit 83 determines the region imaged in each pixel block based on the DCT feature amount of each pixel block and the DCT feature amount of the teacher data (step S403). Through the processing in steps S401 to S403, the region determination unit 83 determines a region imaged in each pixel block based on the DCT feature amount.

  FIG. 16 is an example of teacher data, and is a diagram showing the distribution of the DCT feature value of the bubble region and the DCT feature value of the mucous membrane region in the feature space. In FIG. 16, the feature space is two-dimensional, and the power value in the low frequency band and the power value in the high frequency band are selected as the feature axis of the feature space. The power value in the low frequency band is, for example, the power value F4, and the power value in the high frequency band is, for example, the power value F7. The power value shown in FIG. 16 is an average value of absolute values of AC components in each frequency band, and it can be seen that the foam region has a higher power value in the high frequency band than the mucosal region.

  When a value obtained by weighting the average value of the absolute values of the AC components in each frequency band according to each frequency band is used as the power value, the average value of a plurality of predetermined power values can be used as the feature axis. For example, in the case shown in FIG. 16, the average value of the power values F3 to F5 is taken as the power value in the low frequency band, the average value of the power values F6 to F8 is taken as the power value in the high frequency band, and each average value is taken as the feature axis. Also good. In FIG. 16, the feature space is two-dimensional. However, the feature space is not limited to two dimensions. For example, if each of the power values F1 to F14 is a feature axis, the feature space can be a maximum of 14 dimensions. it can. Furthermore, a DC component value may be added as a feature axis.

  Since the distribution of the DCT feature amount in the feature space is different for each of other regions such as a lesion region and a bleeding region, not limited to the bubble region and the mucous membrane region, the region discriminating unit 83 uses the teacher data to store each pixel block. A process for discriminating the type of the imaged area is performed.

  In step S403, the region discriminating unit 53 performs the kNN (k-Nearest Neighbor Method) method and the partial space based on the DCT feature value of each pixel block and the DCT feature value distribution of the teacher data on the feature space. The area imaged in each pixel block is discriminated using a known discrimination method such as the method.

  The image processing apparatus 2 according to the second embodiment stores a plurality of DCT feature amounts per region as teacher data, and the distribution of the DCT feature amount of the teacher data and the DCT feature amount of each pixel block in the feature space. Since the region captured by each pixel block is determined based on the original, even if the DCT coefficient fluctuates due to, for example, individual differences in the subject or lighting conditions at the time of image capturing, the region determination processing can be performed with high accuracy. Can do.

  In the above embodiment and modification, JPEG is used as an example of image data compression processing by DCT encoding. However, compression processing by DCT encoding is not limited to JPEG processing, and includes processing by MPEG and the like. It is. In particular, MPEG is a moving image data compression process, but since the compression process is performed by a DCT encoding process similar to JPEG, the technical idea according to the present invention can be applied.

  Further, in the above embodiment and the modification, the DCT coefficient is given as the frequency information. However, if the compression encoding method is used for calculating the frequency information for each predetermined image portion in accordance with the decoding process of the compressed image, The technical idea according to the present invention can be applied. For example, the region discriminating unit 53 is calculated by a coding method using a known discrete wavelet transform (DWT), discrete Fourier transform (DFT), Hadamard transform, Laplace transform, or the like other than DCT coding. Frequency information (for example, DCT coefficients in the case of a DCT encoding conversion method, and conversion results of the frequency conversion in the case of Hadamard conversion method), and a plurality of frequency components of each image portion. The type of the region captured in the image portion may be determined based on the feature amount generated based on the frequency information including information. Of course, it will be understood by those skilled in the art that the compression coding method for compressing each predetermined range is not limited to the above-described examples such as the discrete wavelet transform. Even in this case, as in the case of DCT encoding, the region discrimination can be performed at high speed.

  In recent years, in an image processing apparatus, since it is common to compress and store an original image by encoding and compression processing, when frequency information is calculated when decoding processing of a compressed image is performed, the present embodiment As in the first embodiment, the time required for the area determination process can be further shortened.

  The frequency information is not limited to the frequency information calculated at the time of decoding the compressed image. That is, the image processing apparatus according to the present invention may perform the region determination process after performing the process of calculating the predetermined frequency information regardless of the presence or absence of the decoding process.

  Further, as the area determination process performed by the image processing apparatus according to the above-described embodiment and the modification, the area determination process for one in-vivo image has been described. However, when there are a plurality of in-vivo images, the in-vivo images of a series of in-vivo image groups. Alternatively, the region determination process may be automatically performed sequentially.

  In the embodiment and the modification described above, the compression encoding technique for compressing the target image to be processed into a predetermined range, for example, every 8 × 8 pixel image portion, is described. It is not limited to this. For example, the compression may be performed for each 4 × 4 pixel image portion, or may be performed for each 16 × 16 pixel image portion. On the other hand, in another example, a compression encoding method may be used in which the image portion is regarded as the entire image and the entire image is compressed.

  In the above-described embodiment and modification, the DCT coefficient acquisition unit 51 performs the decoding process of the compressed image to acquire (calculate) the DCT coefficient. However, as another modification, a predetermined frequency conversion is performed. Frequency information such as DCT coefficients may be acquired from another external device (not shown) that performs processing.

  The present invention has been described based on the embodiments. This embodiment is an exemplification, and it will be understood by those skilled in the art that various modifications can be made to combinations of the respective constituent elements and processing processes, and such modifications are also within the scope of the present invention. is there.

1 is a block diagram illustrating a schematic configuration of an image processing apparatus according to a first embodiment of the present invention. It is a figure which shows the DCT coefficient of a pixel block. It is a figure explaining arrangement | positioning of the AC component of a DCT coefficient. It is a figure which shows the frequency band to which AC component belongs, and the power value of each frequency band. 3 is a flowchart illustrating an image processing procedure performed by the image processing apparatus illustrated in FIG. 1. It is a flowchart which shows the procedure of the DCT feature-value production | generation process which the DCT feature-value production | generation part shown in FIG. 1 performs. It is a flowchart which shows the procedure of the area | region discrimination | determination process which the area | region discrimination | determination part shown in FIG. 1 performs. It is a figure which shows the relationship between the frequency band to which an AC component belongs, and the ratio of the AC component in which the value becomes zero. 10 is a flowchart illustrating a procedure of DCT feature value generation processing performed by an image processing apparatus according to Modification 1; 10 is a flowchart illustrating a procedure of DCT feature value generation processing performed by an image processing apparatus according to Modification 2; It is a figure which shows the relationship between the frequency band to which AC component belongs, and the average value of the absolute value of the value of AC component of each frequency band. 15 is a flowchart illustrating a procedure of DCT feature value generation processing performed by an image processing apparatus according to Modification 3; It is a figure which shows the relationship between the value which weighted the average value of the absolute value of the value of the frequency band to which an AC component belongs, and the AC component of each frequency band. It is a block diagram which shows schematic structure of the image processing apparatus concerning Embodiment 2 of this invention. It is a flowchart which shows the procedure of the area | region discrimination | determination process which the area | region discrimination | determination part shown in FIG. 14 performs. It is a figure which shows an example of teacher data.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 2 Image processing apparatus 10 Input part 20,60 Storage part 21 In-vivo image storage part 22 Threshold storage part 30 Output part 40,70 Control part 50,80 Image processing part 51 DCT coefficient acquisition part 52 DCT feature-value production | generation part 53, 83 region discriminating unit 62 teacher data storage unit F1-F14 power value H1-H14 frequency region

Claims (19)

For an image portion of a predetermined range included in the target image, an acquisition unit that acquires frequency information representing a plurality of frequency components of the image portion;
A power value indicating the intensity of one frequency component among a plurality of frequency components represented by the frequency information is calculated for each predetermined frequency band, and a feature of an image portion associated with the frequency information based on the power value A feature amount generation unit for generating a feature amount indicating
An area discriminating unit for discriminating the type of the area captured in the image portion based on the feature amount;
An image processing apparatus comprising:   The feature quantity generation unit generates the feature quantity by calculating, as the power value, the number of the one frequency component having a value of 0 among the one frequency component belonging to each frequency band. The image processing apparatus according to claim 1.   The feature quantity generation unit generates the feature quantity by calculating, as the power value, the number of the one frequency component having a value within a predetermined range among the one frequency component belonging to each frequency band. The image processing apparatus according to claim 1, wherein:   The image processing according to claim 1, wherein the feature quantity generation unit generates the feature quantity by calculating an average value of absolute values of the one frequency component belonging to each frequency band as the power value. apparatus.   The feature amount generation unit generates the feature amount by calculating a value calculated by weighting an average value of the absolute values of the one frequency component belonging to each frequency band according to each frequency band as the power value. The image processing apparatus according to claim 1, wherein:   The said feature-value production | generation part calculates the said power value based on the luminance component which is said one frequency component, and produces | generates the said feature-value. Image processing apparatus.   The said feature-value production | generation part produces | generates the said feature-value based on the other frequency component and said power value different from the said one frequency component, It is any one of Claims 1-6 characterized by the above-mentioned. Image processing apparatus.   8. The region determination unit according to claim 1, wherein the region determination unit determines a type of a region captured in the image portion corresponding to the feature amount by comparing the feature amount with a threshold value. An image processing apparatus according to 1. A teacher data storage unit that stores teacher data that is a plurality of the feature quantities classified into different areas;
8. The region determination unit according to claim 1, wherein the region determination unit determines a type of a region captured in the image portion based on the feature amount of the teacher data and the feature amount of the image portion. The image processing apparatus according to claim 1.   The image processing apparatus according to claim 1, wherein the processing target image is a medical image.   The image processing apparatus according to claim 1, wherein the processing target image is an in-vivo image captured by a capsule endoscope.   The region discriminating unit discriminates at least one of a bubble region, a lesion region, a bleeding region, a mucous membrane region, a content region, or a mucus region as a region type captured in the image portion. The image processing apparatus according to claim 1.   The image according to any one of claims 1 to 12, wherein the acquisition unit acquires, as the frequency information, information obtained by performing a predetermined frequency conversion process for each image portion. Processing equipment.   The image processing apparatus according to claim 13, wherein the frequency conversion process is a conversion using a DCT encoding method.   The frequency information is a DCT coefficient calculated when a decoding process of a compressed image compressed by a DCT encoding method is performed as the frequency conversion process. Image processing device. The frequency information is a DCT coefficient calculated when a decoding process of a compressed image compressed by the DCT encoding method is performed,
The image processing apparatus according to claim 1, wherein the one frequency component is an AC component of the DCT coefficient. The frequency information is a DCT coefficient calculated when a decoding process of a compressed image compressed by the DCT encoding method is performed,
The image processing apparatus according to claim 7, wherein the other frequency component is a DC component of the DCT coefficient. An acquisition step of acquiring frequency information representing a plurality of frequency components of the image portion for a predetermined range of the image portion included in the target image;
A power value indicating the intensity of one frequency component among a plurality of frequency components represented by the frequency information is calculated for each predetermined frequency band, and a feature of an image portion associated with the frequency information based on the power value A feature amount generation step for generating a feature amount indicating
A determination step of determining a type of an area captured in the image portion based on the feature amount;
An image processing method comprising: An acquisition step of acquiring frequency information representing a plurality of frequency components of the image portion for a predetermined range of the image portion included in the target image;
A power value indicating the intensity of one frequency component among a plurality of frequency components represented by the frequency information is calculated for each predetermined frequency band, and a feature of an image portion associated with the frequency information based on the power value A feature amount generation step for generating a feature amount indicating
A determination step of determining a region imaged in the image portion based on the feature amount;
An image processing program for causing a computer to exhibit the above.

Images