Image processing device and electronic endoscope system

Abstract

Image processing device (20) comprising: a noise reduction processing unit (24a) that sets a pixel of interest and a neighboring pixel among a plurality of pixels constituting an endoscope observation image, and performs noise reduction processing on the pixel of interest according to a magnitude ratio between a difference between the pixel values ​​of the pixel of interest and the neighboring pixel and a reference threshold value; and a reference threshold changing processing unit (24b) which changes the reference threshold for the noise reduction processing based on a relationship between at least two color components contained in the pixel of interest, wherein the reference threshold changing processing unit (24b) changes the reference threshold for the noise reduction processing based on a ratio between a component G (green) and a component B (blue) contained in the pixel of interest, and wherein the reference threshold changing processing unit (24b) divides a plane GB having two orthogonal axes indicating the G component and the B component into a plurality of regions using an oblique straight line passing through an origin, and changes the reference threshold for noise reduction processing according to the region among the plurality of regions of the plane GB to which the ratio of the G component to the B component included in the pixel of interest belongs.

Description

Technical area

[0001] The present invention relates to an image processing apparatus and an electronic endoscope system. Technical background

[0002] Patent Document 1 discloses an electronic endoscope system that acquires normal observation images taken using normal light and narrow-band observation images taken using narrow-band light having a narrower bandwidth than that of normal light as endoscope observation images, and displays the endoscope observation images.

[0003] Patent Document 2 discloses an image processing apparatus and method for performing image noise reduction processing using a technology called ε-filtering. The ε-filtering technology is used to perform noise reduction processing on a pixel of interest by setting a pixel of interest and neighboring pixels among a plurality of pixels constituting an image, performing averaging using only a portion where the difference between the pixel values ​​of the pixel of interest and each neighboring pixel is less than or equal to a reference threshold, and omitting a high-frequency component where the difference between the pixel values ​​of the pixel of interest and each neighboring pixel is greater than a reference threshold. State-of-the-art documentsPatent specifications Patent Document 1: JP 2015-223249A Patent Document 2: JP 2008-124976A

[0004] US 2006 / 0245002 A1 describes an image processing device for reducing quantization noise in photographic image data. By applying a weighted average, a group of surrounding pixels around a target pixel is used to calculate a corrected pixel value. The weight assigned to each surrounding pixel depends on the difference between the target pixel and this pixel.

[0005] JP 2003-179 779 A describes a method for noise reduction. In this method, a central pixel and the pixels surrounding this central pixel are detected from a video signal. The number of surrounding pixels is determined based on a region selection signal. The difference between the central pixel and the surrounding pixels is calculated. The surrounding pixels are differentiated into those with a correlation to the central pixel and those without. Summary of the inventionProblem to be solved by the invention

[0006] However, during observation using an endoscope, cases often occur where the image acquisition environment changes significantly, such as switching from observation under normal light using normal light to observation under narrow-band light using narrow-band light, and spraying chemicals such as indigo carmine and iodine. If a conventional ε-filter with a threshold that can be set in advance is applied as is to endoscope image processing, there is a risk that sufficient noise reduction processing may not be performed due to changes in the image acquisition environment, resulting in a risk of deterioration in the quality of the endoscope observation images.

[0007] The present invention has been made based on the knowledge of the above problems, and one of its objects is to provide an image processing apparatus and an electronic endoscope system by which high-quality endoscope observation images can be acquired by performing sufficient noise reduction processing even when an image pickup environment changes. Means to solve the problem

[0008] The invention solves this problem by an image processing device according to claim 1 and by an electronic endoscope system according to claim 4.

[0009] An image processing apparatus according to one aspect of the present invention comprises: a noise reduction processing unit that sets a pixel of interest and a neighboring pixel among a plurality of pixels constituting an endoscope observation image, and performs noise reduction processing on the pixel of interest according to a magnitude ratio between a difference between the pixel values ​​of the pixel of interest and the neighboring pixel and a reference threshold; and a reference threshold changing processing unit that changes the reference threshold for noise reduction processing based on a ratio between at least two color components included in the pixel of interest.

[0010] An electronic endoscope system according to one aspect of the present invention includes: an electronic endoscope that acquires an endoscope observation image; and an image processing device that performs image processing on the endoscope observation image.The image processing device comprises: a noise reduction processing unit that sets a pixel of interest and a neighboring pixel among a plurality of pixels constituting the endoscope observation image, and performs noise reduction processing on the pixel of interest according to a magnitude ratio between a difference between the pixel values ​​of the pixel of interest and the neighboring pixel and a reference threshold; and a reference threshold changing processing unit that changes the reference threshold for noise reduction processing based on a ratio between at least two color components included in the pixel of interest.

[0011] The reference threshold changing processing unit changes the reference threshold for the noise reduction processing based on a ratio between a component G (green) and a component B (blue) included in the pixel of interest.

[0012] The reference threshold changing processing unit divides a plane GB having two orthogonal axes indicating the G component and the B component into a plurality of regions using an oblique straight line passing through an origin, and changes the reference threshold for noise reduction processing according to the region among the plurality of regions of the plane GB to which the ratio of the G component to the B component included in the pixel of interest belongs.

[0013] The noise reduction processing unit and the reference threshold changing processing unit may perform the noise reduction processing on each of the plurality of pixels while changing the reference threshold.

[0014] The noise reduction processing unit and the reference threshold changing processing unit may perform the noise reduction processing on an entirety of the plurality of pixels or on each of a plurality of pixel groups divided from the plurality of pixels while changing the reference threshold. Advantageous results of the invention

[0015] With the present invention, it is possible to provide an image processing apparatus and an electronic endoscope system that can acquire high-quality endoscope observation images by performing sufficient noise reduction processing even when an image acquisition environment changes. Short description of the drawings Fig. 1 is a block diagram showing a configuration of an electronic endoscope system according to an embodiment of the present invention. Fig. 2 is a conceptual diagram showing noise reduction processing performed by an ε-filter processing unit. Fig. 3 is a conceptual diagram showing processing for changing the reference threshold value executed by a processing unit for changing the reference threshold value. Fig. Figure 4 is a diagram showing pixels constituting a normal viewing image mapped onto a plane GB. Fig. Figure 5 is a diagram showing changes in the imaged pixels when transitioning from a normal viewing condition to a narrowband viewing condition or a chemical sprayed condition (indigo carmine sprayed condition). Fig. 6 is a flowchart showing image processing according to the embodiment of the present invention. Description of embodiments

[0016] In the following, with reference to the Fig. 1 to 6, an electronic endoscope system 1 according to an embodiment of the present invention is described. As in Fig. 1, the electronic endoscope system 1 includes an electronic endoscope 10, a processor (an image processing device and a light source device) 20, and a monitor 30.

[0017] Although in Fig. 1, the electronic endoscope 10 includes a grip and control portion to be grasped by an operator, an insertion portion that is flexible and extends from the grip and control portion, a universal nozzle extending from the grip and control portion in a direction away from the insertion portion, and a connecting member provided at the front end of the universal nozzle.

[0018] Optical fibers 11 are incorporated into the electronic endoscope 10. The optical fibers 11 extend through the insertion section, the handle and operation section, and the universal nozzle of the electronic endoscope 10 into the interior of the connector. The connector of the electronic endoscope 10 is connected to a connector of the processor 20, thus optically connecting the electronic endoscope 10 and the processor 20. Exposure light (ordinary light or narrow-band light described below) from a filtered light source 21 incorporated into the processor 20 is guided inside the optical fibers 11 and emitted outward from an exposure lens 12 provided at the front end of the insertion section of the electronic endoscope 10 according to a predetermined light distribution pattern.

[0019] The filtered light source 21 includes a high-intensity lamp (e.g., a xenon lamp, a halogen lamp, a mercury lamp, or a metal halide lamp) that emits white light comprising wavelength bands corresponding to R (red), G (green), and B (blue), respectively. The filtered light source 21 also includes a filter unit disposed in the light path of the white light emitted by the high-intensity lamp. The filter unit includes a rotary filter turret provided with a white filter that transmits white light from the high-intensity lamp, thereby generating normal light, and a narrow-band filter that narrows the wavelength band of the white light emitted by the high-intensity lamp, thereby generating narrow-band light.The narrow-band filter has a spectral transmittance with a narrow width of half the maximum for each of the R, G, and B wavelength bands. The rotary filter turret of the filter unit is driven to rotate, and as a result, the white light from the high-intensity lamp alternately passes through the white color filter and the narrow-band filter. The filtered light source 21 alternately emits normal light and narrow-band light with a bandwidth narrower than that of normal light. The wavelength band of narrow-band light can be set as desired, as long as it is narrower than the wavelength band of normal light. For example, narrow-band light may have a wavelength band that corresponds to the spectral properties of hemoglobin.The filter-equipped light source 21 disclosed, for example, in the above Patent Document 2 is well known, and therefore, further detailed description thereof will be omitted.

[0020] An imaging unit 13 is provided at the front end of the insertion section of the electronic endoscope 10. The imaging unit 13 is composed of multiple components including an objective lens 13a and a CCD 13b that captures an image of a subject that has passed through the objective lens 13a, and is integrated into one piece using a resin material such as epoxy resin. The CCD 13b alternately captures a normal observation image signal and a narrow-band observation image signal, each based on normal light and narrow-band light, which are alternately emitted from the filter-equipped light source 21 via the optical fibers 11 and the exposure lens 12. The normal observation image signal and the narrow-band observation image signal are sent to the processor 20 via a signal transmission cable 14.

[0021] The processor 20 includes a control unit 22, a viewing image input unit 23 (an image input processing unit), an image processing unit (a calculation unit) 24, an image memory 25, and a display processing unit 26. The control unit 22 completely controls all components of the processor 20.

[0022] The observation image input unit 23 performs input processing on the normal observation image signal and the narrow-band observation image signal transmitted via the signal transmission cable 14 of the electronic endoscope 10, so that the signals are input thereto as a normal observation image and a narrow-band observation image. The normal observation image and the narrow-band observation image input to the observation image input unit 23 are "endoscope observation images."

[0023] The image processing unit 24 performs image processing on the endoscope observation images input to the observation image input unit 23. The image processing unit 24 includes an ε-filter processing unit 24a (a noise reduction processing unit) and a reference threshold changing processing unit 24b (a threshold calculation unit).

[0024] The ε-filter processing unit 24a sets a pixel of interest and neighboring pixels among a plurality of pixels constituting an endoscope observation image, and performs noise reduction processing on the pixel of interest according to the magnitude ratio between the difference between the pixel values ​​of the pixel of interest and each neighboring pixel and a reference threshold value.

[0025] Fig. 2 is a conceptual diagram showing the noise reduction processing performed by the ε-filter processing unit 24a. In the example shown in the diagram, the plurality of pixels constituting an endoscope observation image are 3 × 3 = 9 pixels (P11, P12, P13, P21, P22, P23, P31, P32, and P33), where a pixel P22 located at the center of the plurality of pixels is defined as the pixel of interest, and the eight pixels P11, P12, P13, P21, P23, P31, P32, and P33 adjacent to the pixel of interest P22 are defined as the neighboring pixels. Note that the relationship between the pixel of interest and the neighboring pixels is not limited to the above example and can be set as desired.For example, only one or more pixels among the pixels adjacent to the pixel of interest may be defined as the neighboring pixels, or pixels that are not adjacent to the pixel of interest but surround it (e.g., omitting one pixel) may be defined as the neighboring pixels.

[0026] The ε-filter processing unit 24a calculates the difference between the pixel of interest P22 and each of the neighboring pixels P11, P12, P13, P21, P23, P31, P32, and P33 and determines whether the difference is greater than or equal to a reference threshold. Fig. In the example shown in Figure 2, the pixel of interest P22 and the neighboring pixels P23, P32, and P33, for which the above-described difference is greater than or equal to the reference threshold, are displayed against a gray background, and the neighboring pixels P11, P12, P13, P21, and P31, for which the above-described difference is smaller than the reference threshold, are displayed against a white background. The ε-filter processing unit 24a calculates the average value for the pixel of interest P22 and the neighboring pixels P11, P12, P13, P21, and P31, for which the above-described difference is smaller than the reference threshold, to determine a new pixel of interest (a pixel of interest after correction) P22' with reduced noise.The adjacent pixels P23, P32, and P33 where the above-described difference is greater than or equal to the reference threshold may be, for example, edge portions or the like, and using these adjacent pixels for averaging may potentially cause the endoscope observation image to appear blurry. Therefore, by discarding the adjacent pixels P23, P32, and P33 where the above-described difference is greater than or equal to the reference threshold, it is possible to perform highly precise noise reduction processing. A specific formula for determining P22' is as follows: P22'=(P11+P12+P13+P21+P31+P22) / 6

[0027] The reference threshold changing processing unit 24b changes a reference threshold used by the ε-filter processing unit 24a for performing the noise reduction processing, based on the ratio between a component G and a component B among the component R (red), the component G (green), and the component B (blue) included in the pixel of interest (in the above-described example, P22).

[0028] Fig. 3 is a conceptual diagram showing reference threshold changing processing performed by the reference threshold changing processing unit 24b. As shown in the drawing, the reference threshold changing processing unit 24b divides a plane GB having two orthogonal axes indicating the G component and the B component into four regions using three oblique lines passing through the origin, and changes the reference threshold used by the ε-filter processing unit 24a to perform noise reduction processing according to the region to which the ratio of the G component to the B component included in the pixel of interest belongs among the four regions of the plane GB.

[0029] In this example, in the case where component B takes a constant value, the degrees with the smallest slope (the ratio of component G is the largest) are defined as slope degrees 1, the degrees with the second smallest slope (the ratio of component G is the second largest) are defined as slope degrees 2, and the degrees with the largest slope (the ratio of component G is the smallest) are defined as slope degrees 3.

[0030] Based on the experience of various types of sampling or the like, the inventors of the present invention found that the viewing environment (image pickup environment) corresponding to the area below oblique grade 1 is in an “iodine-sprayed viewing state,” the viewing environment corresponding to the area between oblique grade 1 and oblique grade 2 is in a “narrow-band light viewing state,” the viewing environment corresponding to the area between oblique grade 2 and oblique grade 3 is in a “normal light viewing state,” and the viewing environment corresponding to the area above oblique grade 3 is in an “indigo carmine-sprayed viewing state.”

[0031] For example, the reference threshold changing processing unit 24b may set the reference threshold to be the largest when the viewing environment is in the "iodine-sprayed viewing state," the reference threshold to be the smallest when the viewing environment is in the "narrow-band light viewing state," the reference threshold to be the second largest when the viewing environment is in the "normal light viewing state," and the reference threshold to be the third largest when the viewing environment is in the "indigo carmine-sprayed viewing state." The reason for this is that in the narrow-band light viewing state and the indigo carmine-sprayed viewing state, minute features are emphasized, and it is preferable that the reference threshold is not too large (i.e.,that minute features are caused to appear blurry if the reference threshold is too high), whereas in the iodine-sprayed viewing condition, no minute features are picked up and the image appears darker (contains a greater amount of noise) than in the normal light viewing condition, and that it is preferable that the reference threshold is set relatively high.

[0032] It is pointed out that in Fig. 3 the number of oblique degrees passing through the origin and the number of regions divided by the oblique degrees can be set as desired, and various conceptual modifications can be applied. For example, the plane GB can be divided into two regions using one oblique degree, and the plane GB can be divided into three regions using two oblique degrees. This means that the plane GB can be divided into n+1 regions using n oblique degrees (where n is a positive integer). Likewise, the viewing environment to be recognized (the image acquisition environment) is not limited to the Fig. 3 (the iodine-sprayed condition, the narrow-band light condition, the normal viewing condition, and the indigo carmine-sprayed condition), and various modifications of the concept are applicable.

[0033] In this way, the ε-filter processing unit 24a and the reference threshold changing processing unit 24b of the image processing unit 24 perform noise reduction processing on each of the plurality of pixels constituting the endoscope observation image while changing the reference threshold.

[0034] Fig. Figure 4 is a diagram showing the pixels constituting a normal viewing image mapped onto the GB plane. Fig. Figure 5 is a diagram showing changes in the imaged pixels when transitioning from the normal viewing state to the narrowband viewing state or to the chemical sprayed state (the indigo carmine sprayed state).

[0035] Since the R component is dominant in the endoscope observation image, the value of the G component and the B component change in essentially the same ratio, and the image on the GB plane essentially has the shape of a gradient. The slope of these gradients (B / G) can be interpreted as a feature value and is used to identify the observation environment (image acquisition environment).

[0036] As in Fig. 4, the ratio between the component G and the component B in the case of a normal viewing image is approximately 1:1, and the inclination of the degrees approaches 1. As shown in Fig. 5, on the other hand, when a transition occurs from the normal viewing state to the narrow-band viewing state or the chemical-sprayed state (the indigo carmine-sprayed state), the G component and the B component become unbalanced, and the inclination of the degrees changes to become smaller or larger than 1. In such a context, it is possible to achieve the optimal noise reduction effect by subtracting the current viewing environment (the image pickup environment) from the value of the inclination of the degrees and setting one of the reference thresholds prepared for each of the viewing environments (see Fig. 3).

[0037] It is possible to calculate the above-described inclination (B / G) of the degrees separately for each pixel. Alternatively, it is also possible to calculate the average of the inclinations in the entire image, determine the reference threshold corresponding to this average, consider the average as the inclination to be applied to the entire image (the representative inclination of the image), and apply the reference threshold uniformly to the entire image. This means that the method for adjusting the oblique degrees according to Fig. 3 can be defined as desired and that various modifications of the concept are applicable.

[0038] The image memory 25 stores an endoscope observation image that has undergone image processing performed by the image processing unit 24. The display processing unit 26 displays the endoscope observation image stored in the image memory 25 on the monitor 30.

[0039] In the following, with reference to the Fig. 6, the image processing performed by the electronic endoscope system 1 and the processor 20 according to the present embodiment is described.

[0040] In a step S1, the image processing unit 24 acquires an endoscope observation image of the current frame.

[0041] In a step S2, the reference threshold changing processing unit 24b of the image processing unit 24 calculates the ratio (B / G) between the component G and the component B included in the pixel of interest of the endoscope observation image acquired in step S1.

[0042] In step S3, the reference threshold changing processing unit 24b of the image processing unit 24 derives the current image capturing environment by comparing the ratio (B / G) between component G and component B calculated in step S2 with an environment recognition criterion. The reference threshold changing processing unit 24b of the image processing unit 24 infers, for example, based on the region among the regions of the plane GB to which the image defined by the ratio (B / G) between component G and component B calculated in step S2 belongs, which state the image capturing environment is in at the current time among the iodine-sprayed state, the narrow-band light state, the normal viewing state, and the indigo carmine-sprayed state described above.

[0043] In a step S4, the reference threshold changing processing unit 24b of the image processing unit 24 acquires (sets) the reference threshold corresponding to the image pickup environment at the current time point derived in step S3.

[0044] In a step S5, the ε-filter processing unit 24a of the image processing unit 24 performs noise reduction processing on the pixel of interest in the endoscope observation image using the reference threshold value determined in step S4.

[0045] In step S6, the image processing unit 24 determines whether the calculations in which each pixel of the endoscope observation image is treated as the pixel of interest are completed. If it is determined that the calculations in which each pixel of the endoscope observation image is treated as the pixel of interest are completed (step S6: Yes), step S7 is executed next. If it is determined that the calculations in which each pixel of the endoscope observation image is treated as the pixel of interest are not completed (step S6: No), the pixel of interest in the endoscope observation image is shifted, and the processing loop from step S2 to step S5 is repeated.

[0046] In step S7, the display processing unit 26 displays the endoscope observation image subjected to the image processing performed by the image processing unit 24 (subjected to the noise reduction processing performed by the ε-filter processing unit 24a and the reference threshold changing processing unit 24b) on the monitor 30.

[0047] In step S8, it is determined whether the image acquisition performed by the electronic endoscope system 1 is completed or not. If it is determined that the image acquisition performed by the electronic endoscope system 1 is completed (step S8: Yes), the processing is terminated. If it is determined that the image acquisition performed by the electronic endoscope system 1 is not completed (continuing) (step S8: No), the processing loop from step S1 to step S7 is repeated.

[0048] As described above, in the electronic endoscope system 1 and the processor 20 according to the present embodiment, the ε-filter processing unit (the noise reduction processing unit) 24a sets a pixel of interest and neighboring pixels among a plurality of pixels constituting an endoscope observation image, and performs noise reduction processing on the pixel of interest according to the magnitude ratio between the difference between the pixel values ​​of the pixel of interest and each neighboring pixel and a reference threshold value, and the reference threshold changing processing unit 24b changes the reference threshold value for the noise reduction processing based on a ratio between at least two color components included in the pixel of interest.Accordingly, it is possible to obtain a high-quality endoscope observation image by performing sufficient noise reduction processing even when the observation environment (image acquisition environment) changes.

[0049] Specifically, the electronic endoscope system 1 and the processor 20 according to the present embodiment automatically detect narrow-band light observation and chemical spray when applied, and automatically correct the reference threshold, thereby achieving the optimal noise reduction effect without requiring any user action. This means that in the present embodiment, the optimal reference threshold is dynamically adjusted (in real time) according to changes in the observation environment (image acquisition environment).

[0050] In the above embodiment, an example was described in which the reference threshold changing processing unit 24b changes the reference threshold for noise reduction processing based on the ratio between the G component and the B component included in the pixel of interest. However, the comparison value that can be used to change the reference threshold is not limited to this, and various conceptual modifications are applicable. The reference threshold changing processing unit 24b can change the reference threshold for noise reduction processing based on, for example, the ratio (G / R) between the G component and the R component included in the pixel of interest, or the ratio (B / R) between the B component and the R component included in the pixel of interest.Furthermore, the processing unit 24b for changing the reference threshold may convert RGB to YCbCr or the like in another color space and then use the relationship between them. This means that the processing unit 24b for changing the reference threshold only needs to change the reference threshold for noise reduction processing based on the relationship between at least two color components contained in the pixel of interest.

[0051] In the above embodiment, an example was described in which the ε-filter processing unit 24a and the reference threshold changing processing unit 24b of the image processing unit 24 perform noise reduction processing on each of the plurality of pixels constituting the endoscope observation image while changing the reference threshold. However, the ε-filter processing unit 24a and the reference threshold changing processing unit 24b of the image processing unit 24 may perform noise reduction processing on the entirety of the plurality of pixels constituting the endoscope observation image or on each of a plurality of pixel groups divided from the plurality of pixels while changing the reference threshold.

[0052] In the above embodiment, an example of noise reduction processing performed by the ε-filter processing unit 24a of the image processing unit 24 using ε-filter technology was described. However, the algorithm that can be used to perform noise reduction processing is not limited to the ε-filter. That is, any algorithm can be used as long as noise reduction processing is performed on the pixel of interest according to the magnitude ratio between the difference between the pixel values ​​of the pixel of interest and each neighboring pixel and a reference threshold. In addition to the ε-filter, the present invention can be applied, for example, to a digital filter such as a bilateral filter that uses the difference between the pixel values ​​of the pixel of interest and each neighboring pixel.

[0053] In the above embodiment, an example was described in which noise reduction processing is performed on an RGB image that has undergone demosaicing. However, noise reduction processing may be performed on a raw image that has not undergone demosaicing. When noise reduction processing is performed on a raw image that has not undergone demosaicing, necessary information is acquired from a nearby pixel or an RGB image of the previous frame, so that the ratio (B / G) between the B component and the G component can be calculated. Industrial applicability

[0054] The image processing apparatus and the electronic endoscope system according to the present invention can be preferably used in the field of medical endoscopes, for example, as an image processing apparatus and an electronic endoscope system. Description of reference symbols 1 Electronic endoscope system 10 Electronic endoscope 11 optical fibers 12 exposure lens 13 Imaging Unit 13a Objective lens 13b CCD 14 signal transmission cables 20 Processor (image processing device, light source device) 21 Light source fitted with a filter 22 Control unit 23 Viewing image input unit (image input processing unit) 24 Image processing unit (calculation unit) 24a ε-filter processing unit (noise reduction processing unit) 24b Processing unit for changing the reference threshold (threshold calculation unit) 25 image memories 26 Display processing unit 30 monitors

Claims

[1] Image processing device (20) comprising: a noise reduction processing unit (24a) that sets a pixel of interest and a neighboring pixel among a plurality of pixels constituting an endoscope observation image, and performs noise reduction processing on the pixel of interest according to a magnitude ratio between a difference between the pixel values ​​of the pixel of interest and the neighboring pixel and a reference threshold value; and a reference threshold changing processing unit (24b) which changes the reference threshold for the noise reduction processing based on a relationship between at least two color components contained in the pixel of interest, wherein the reference threshold changing processing unit (24b) changes the reference threshold for the noise reduction processing based on a ratio between a component G (green) and a component B (blue) contained in the pixel of interest, and wherein the reference threshold changing processing unit (24b) divides a plane GB having two orthogonal axes indicating the G component and the B component into a plurality of regions using an oblique straight line passing through an origin, and changes the reference threshold for noise reduction processing according to the region among the plurality of regions of the plane GB to which the ratio of the G component to the B component included in the pixel of interest belongs. [2] The image processing apparatus (20) according to claim 1, wherein the noise reduction processing unit (24a) and the reference threshold changing processing unit (24b) execute the noise reduction processing on each of the plurality of pixels while changing the reference threshold. [3] The image processing apparatus according to claim 1, wherein the noise reduction processing unit (24a) and the reference threshold changing processing unit (24b) perform the noise reduction processing on a whole of the plurality of pixels or on each of a plurality of pixel groups divided from the plurality of pixels while changing the reference threshold. [4] Electronic endoscope system (1) comprising: an electronic endoscope (10) that captures an endoscope viewing image; and an image processing device (20) which performs image processing on the endoscope observation image, wherein the image processing device (20) comprises: a noise reduction processing unit (24a) that sets a pixel of interest and a neighboring pixel among a plurality of pixels constituting the endoscope observation image, and performs noise reduction processing on the pixel of interest according to a magnitude ratio between a difference between the pixel values ​​of the pixel of interest and the neighboring pixel and a reference threshold value; and a reference threshold changing processing unit (24b) which changes the reference threshold for the noise reduction processing based on a relationship between at least two color components contained in the pixel of interest, wherein the reference threshold changing processing unit (24b) changes the reference threshold for the noise reduction processing based on a ratio between a component G (green) and a component B (blue) contained in the pixel of interest, and wherein the reference threshold changing processing unit (24b) divides a plane GB having two orthogonal axes indicating the G component and the B component into a plurality of regions using an oblique straight line passing through an origin, and changes the reference threshold for noise reduction processing according to the region among the plurality of regions of the plane GB to which the ratio of the G component to the B component included in the pixel of interest belongs. [5] The electronic endoscope system (1) according to claim 4, wherein the noise reduction processing unit (24a) and the reference threshold changing processing unit (24b) execute the noise reduction processing on each of the plurality of pixels while changing the reference threshold. [6] The electronic endoscope system (1) according to claim 4, wherein the noise reduction processing unit (24a) and the reference threshold changing processing unit (24b) perform the noise reduction processing on a whole of the plurality of pixels or on each of a plurality of pixel groups divided from the plurality of pixels while changing the reference threshold.

Images