CN114240941B - Endoscope image noise reduction method, device, electronic apparatus, and storage medium - Google Patents
Endoscope image noise reduction method, device, electronic apparatus, and storage medium Download PDFInfo
- Publication number
- CN114240941B CN114240941B CN202210174607.6A CN202210174607A CN114240941B CN 114240941 B CN114240941 B CN 114240941B CN 202210174607 A CN202210174607 A CN 202210174607A CN 114240941 B CN114240941 B CN 114240941B
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- noise reduction
- target
- laplacian
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- 230000009467 reduction Effects 0.000 title claims abstract description 75
- 230000009466 transformation Effects 0.000 claims abstract description 47
- 238000004364 calculation method Methods 0.000 claims abstract description 35
- 238000001514 detection method Methods 0.000 claims description 20
- 238000004590 computer program Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 13
- 238000003708 edge detection Methods 0.000 claims description 12
- 230000004927 fusion Effects 0.000 claims description 7
- 230000001965 increasing effect Effects 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000012545 processing Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 9
- 230000005540 biological transmission Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 210000000056 organ Anatomy 0.000 description 3
- 230000000717 retained effect Effects 0.000 description 3
- 230000009131 signaling function Effects 0.000 description 3
- 230000000903 blocking effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 210000000214 mouth Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 210000002784 stomach Anatomy 0.000 description 1
- 238000011426 transformation method Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/10—Image enhancement or restoration by non-spatial domain filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
-
- G06T5/70—
-
- G06T5/80—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10068—Endoscopic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20016—Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20064—Wavelet transform [DWT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20192—Edge enhancement; Edge preservation
Abstract
The present application relates to an endoscopic image noise reduction method, an apparatus, an electronic device, and a storage medium, wherein the endoscopic image noise reduction method includes: performing multi-scale geometric transformation on a target area in a Laplacian image of an endoscope image to obtain a low-frequency coefficient of the target area; performing enhancement calculation on the low-frequency coefficient according to the enhancement coefficient to obtain a corrected low-frequency coefficient; carrying out inverse transformation of multi-scale geometric transformation on the pixel value of the target pixel according to the corrected low-frequency coefficient to obtain a noise-reduced target pixel; and obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image and the corresponding Gaussian image to obtain a de-noised endoscope image. By the method and the device, the problem that edge information in the endoscope image is lost by adopting a noise reduction method of a common camera image in the related technology is solved, information of an edge area of the endoscope image is reserved, and image loss is reduced.
Description
Technical Field
The present application relates to the field of endoscope technologies, and in particular, to an endoscope image noise reduction method, apparatus, electronic device, and storage medium.
Background
The endoscope is a detection instrument integrating the traditional technologies of optics, ergonomics, precision machinery, modern electronics, mathematics, software and the like. Specifically, the endoscope comprises an image sensor, an optical lens, a light source illumination, a mechanical device and the like, can enter the stomach through the oral cavity or enter the body through other natural orifices, and can see the lesion which can not be displayed by X-ray. Images acquired through an endoscope generally need to be subjected to noise reduction processing and then analyzed.
In the related art, the endoscope image noise reduction method generally follows the noise reduction algorithm of the common camera image, such as a noise reduction method in a spatial domain, a noise reduction method in a frequency domain, a dual-domain noise reduction method combining a spatial domain and a frequency domain, and a deep learning method. However, the endoscopic image is an enlarged image of an organ, edges and details in the image occupy a plurality of pixels, the characteristics of the endoscopic image are not considered in the above noise reduction method, and the edges of the endoscopic image are lost after noise reduction.
At present, no effective solution is provided for the problem that edge information in an endoscope image is lost by adopting a noise reduction method of a common camera image in the related art.
Disclosure of Invention
The embodiment of the application provides an endoscope image noise reduction method, equipment, an electronic device and a storage medium, which are used for at least solving the problem that edge information in an endoscope image is lost by adopting a noise reduction method of a common camera image in the related art.
In a first aspect, an embodiment of the present application provides an endoscope image denoising method, where an endoscope image includes a laplacian image and a gaussian image, and denoising pixels in the laplacian image includes:
performing multi-scale geometric transformation on a target region in the Laplace image to obtain a low-frequency coefficient and a high-frequency coefficient of the target region, wherein the target region is determined according to the position of a target pixel;
performing enhancement calculation on the low-frequency coefficient according to an enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target area;
carrying out inverse transformation of the multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient to obtain a noise-reduced target pixel;
and obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image and the corresponding Gaussian image to obtain a de-noised endoscope image.
In some of these embodiments, before performing the enhancement calculation on the low frequency coefficients according to enhancement coefficients, the method includes:
judging whether the target area is a continuous edge area or not;
and under the condition that the target area is a continuous edge area, performing enhancement calculation on the low-frequency coefficient in the target area according to an enhancement coefficient.
In some embodiments, the determining, based on the endoscopic image, an image pyramid that is n layers, where the image pyramid includes a laplacian pyramid and a gaussian pyramid, the laplacian pyramid is composed of laplacian images, the gaussian pyramid is composed of gaussian images, and in a case where i is greater than or equal to 0 and less than or equal to n-1, for an ith-layer laplacian image, the determining whether the target region is a continuous edge region includes:
determining a detection region corresponding to the target region in a gradient image of an i + 1-th layer Gaussian image of the image pyramid;
and for each pixel in the detection area, determining a target area corresponding to the detection area as the continuous edge area under the condition that the pixel value meets the condition that the number of pixels in the preset pixel value range is greater than or equal to a preset number threshold.
In some embodiments, the determining the enhancement coefficient according to the first pixel value mean of the pixels in the preset pixel value range in the target region comprises:
when the first pixel value mean value is smaller than or equal to a first preset gradient threshold value, the enhancement coefficient is a first fixed value;
when the first pixel value mean value is larger than the first preset gradient threshold value and smaller than a second preset gradient threshold value, the enhancement coefficient is increased along with the first pixel value mean value, wherein the size of the second preset gradient threshold value is determined according to a second pixel value mean value in the preset pixel value range in the gradient image;
and when the first pixel value mean value is greater than or equal to the second preset gradient threshold value, the enhancement coefficient is a second fixed value, and the second fixed value is greater than the first fixed value.
In some embodiments, the method for determining the gradient image of the i +1 th layer gaussian image includes:
performing edge detection on the i +1 layer Gaussian image to obtain an initial gradient image;
and selecting a preset pixel value or an initial pixel value in the initial gradient image as a pixel value in the gradient image of the final i +1 th layer Gaussian image according to the first preset gradient threshold value.
In some embodiments, the fusing the noise-reduced laplacian image with a corresponding gaussian image to obtain a noise-reduced endoscope image includes:
fusing the i +1 th layer Gaussian image and the i layer Laplace image to obtain an i layer Gaussian image;
and when the value of i is 0, taking the 0 th layer Gaussian image as the endoscope image after noise reduction.
In some embodiments, after obtaining the noise-reduced target pixel, the method includes:
determining the weight of each pixel in a second preset range according to the distance between each pixel in the second preset range and the target pixel;
and correcting the pixel value of the target pixel according to the pixel value of each pixel in the second preset range and the weight.
In some of these embodiments, the determining of the target region from the location of the target pixel comprises:
and determining the target area according to a first preset range by taking the target pixel as a center.
In a second aspect, an embodiment of the present application provides an endoscopic image noise reduction device, where an endoscopic image includes a laplacian image and a gaussian image, the device includes a multi-scale geometric transformation module, an enhancement calculation module, a noise reduction calculation module, and a fusion module:
the multi-scale geometric transformation module is used for performing multi-scale geometric transformation on a target area in the laplacian image to obtain a low-frequency coefficient and a high-frequency coefficient of the target area, wherein the target area is determined according to the position of a target pixel;
the enhancement calculation module is used for performing enhancement calculation on the low-frequency coefficient according to an enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target area;
the noise reduction calculation module is used for performing inverse transformation of the multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient to obtain a noise-reduced target pixel;
the fusion module is used for obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image and a corresponding Gaussian image to obtain a de-noised endoscope image.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the endoscopic image noise reduction method according to the first aspect.
In a fourth aspect, the present application provides a storage medium, on which a computer program is stored, which when executed by a processor, implements the steps of the endoscopic image noise reduction method as described in the first aspect above.
Compared with the related art, the endoscope image noise reduction method provided by the embodiment of the application obtains the low-frequency coefficient and the high-frequency coefficient of the target area by performing multi-scale geometric transformation on the target area in the laplacian image of the endoscope image, wherein the target area is determined according to the position of the target pixel; performing enhancement calculation on the low-frequency coefficient according to the enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target area; carrying out inverse transformation of multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient to obtain the noise-reduced target pixel; the method comprises the steps of obtaining a de-noised Laplacian image according to a de-noised target pixel, fusing the de-noised Laplacian image with a corresponding Gaussian image to obtain a de-noised endoscope image, solving the problem that edge information in the endoscope image is lost by adopting a common camera image de-noising method in the related art, reserving information of an edge area of the endoscope image, and reducing image loss.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more concise and understandable description of the application, and features, objects, and advantages of the application.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a block diagram of a hardware configuration of a terminal of the endoscopic image noise reduction method of the present embodiment;
FIG. 2 is a flowchart of an endoscopic image noise reduction method of the present embodiment;
FIG. 3 is a flowchart of a method for determining a continuous edge region according to an embodiment of the present application;
FIG. 4 is a graph illustrating enhancement coefficients as a function of a mean value of first pixel values according to an embodiment of the present application;
FIG. 5 is a flow chart of a method of determining a gradient image according to an embodiment of the present application;
FIG. 6 is a schematic diagram of an endoscopic image noise reduction method according to the present preferred embodiment;
fig. 7 is a block diagram of the configuration of the endoscopic image noise reduction apparatus of the present embodiment.
Detailed Description
For a clearer understanding of the objects, aspects and advantages of the present application, reference is made to the following description and accompanying drawings.
Unless defined otherwise, technical or scientific terms used herein shall have the same general meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The use of the terms "a" and "an" and "the" and similar referents in the context of this application do not denote a limitation of quantity, either in the singular or the plural. The terms "comprises," "comprising," "has," "having," and any variations thereof, as referred to in this application, are intended to cover non-exclusive inclusions; for example, a process, method, and system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to the listed steps or modules, but may include other steps or modules (elements) not listed or inherent to such process, method, article, or apparatus. Reference throughout this application to "connected," "coupled," and the like is not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. Reference to "a plurality" in this application means two or more. "and/or" describes an association relationship of associated objects, meaning that three relationships may exist, for example, "A and/or B" may mean: a exists alone, A and B exist simultaneously, and B exists alone. In general, the character "/" indicates a relationship in which the objects associated before and after are an "or". The terms "first," "second," "third," and the like in this application are used for distinguishing between similar items and not necessarily for describing a particular sequential or chronological order.
The method embodiments provided in the present embodiment may be executed in a terminal, a computer, or a similar computing device. For example, the method is executed on a terminal, and fig. 1 is a block diagram of a hardware configuration of the terminal according to the endoscope image noise reduction method of the present embodiment. As shown in fig. 1, the terminal may include one or more processors 102 (only one shown in fig. 1) and a memory 104 for storing data, wherein the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA. The terminal may also include a transmission device 106 for communication functions and an input-output device 108. It will be understood by those of ordinary skill in the art that the structure shown in fig. 1 is merely an illustration and is not intended to limit the structure of the terminal described above. For example, the terminal may also include more or fewer components than shown in FIG. 1, or have a different configuration than shown in FIG. 1.
The memory 104 can be used for storing computer programs, for example, software programs and modules of application software, such as a computer program corresponding to the endoscopic image noise reduction method in the present embodiment, and the processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, so as to implement the above-mentioned method. The memory 104 may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory located remotely from the processor 102, which may be connected to the terminal over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission device 106 is used to receive or transmit data via a network. The network described above includes a wireless network provided by a communication provider of the terminal. In one example, the transmission device 106 includes a Network adapter (NIC) that can be connected to other Network devices through a base station to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used to communicate with the internet in a wireless manner.
In the present embodiment, an endoscopic image noise reduction method is provided, and fig. 2 is a flowchart of the endoscopic image noise reduction method of the present embodiment, as shown in fig. 2, the flowchart includes the following steps:
step S210, performing multi-scale geometric transformation on a target region in the Laplace image to obtain a low-frequency coefficient and a high-frequency coefficient of the target region, wherein the target region is determined according to the position of a target pixel.
In the present embodiment, before denoising the endoscopic image, it is necessary to acquire a laplacian image and a gaussian image of the endoscopic image, where the laplacian image and the gaussian image are images of the endoscopic image in different frequency ranges.
In the noise reduction process, firstly, performing multi-scale geometric transformation on the laplacian image to obtain a low-frequency coefficient and a high-frequency coefficient of a target region, wherein the multi-scale geometric transformation is a filtering method or a frequency domain transformation method capable of decomposing high frequency and low frequency, and the method comprises the following steps: non-downsampling laplacian transforms, wavelet transforms, Contourlet transforms, curvelet transforms, and the like. In the embodiment, the multi-scale geometric transformation is preferably wavelet transformation, the wavelet transformation is partial analysis of time or space frequency, multi-scale refinement is gradually carried out on a signal function of an image through telescopic translation operation, and finally high-frequency time subdivision and low-frequency subdivision are achieved. Specifically, in this embodiment, the wavelet transform is based on a series of wavelets with different scales to decompose the signal function of the image, and the coefficients of the signal function under the wavelets with different scales are obtained after the wavelet transform. The coefficient comprises a high-frequency coefficient and a low-frequency coefficient, the high-frequency coefficient is used for analyzing high-frequency information with obvious change, and the low-frequency coefficient is used for researching low-frequency information with smooth transformation.
Since the noise reduction of the target pixel needs to be performed according to the degree of change of other pixels around the target pixel, when performing the noise reduction of the target pixel, it is necessary to determine a target region from the target pixel, and preferably, the target pixel may be determined according to a first preset range, for example, the target pixel may be p (x, y), the rectangle with the size of m × m may be the target region, and the circle with the radius r and the center of p (x, y) may be the target region.
In other embodiments, p (x, y) may be taken as a point on the boundary of the target region in the case where the target region is rectangular, or as a focus of an ellipse in the case where the target region is elliptical.
And step S220, performing enhancement calculation on the low-frequency coefficient according to the enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to the first pixel value mean value of pixels in a preset pixel value range in the target area.
In order to retain image information with insignificant changes in the endoscopic image, in this embodiment, enhancement calculation needs to be performed on the low-frequency coefficient, when determining the enhancement coefficient, a preset pixel value range, for example, a pixel with a pixel value not being 0, or a pixel with a pixel value greater than a certain pixel value, needs to be determined first, and then a mean value of pixels in the preset pixel value range is calculated and recorded as a first mean value of pixel values, for example, in a target area, a mean value of pixels with a pixel value not being 0 is calculated. The larger the first pixel value mean value is, the larger the enhancement coefficient is correspondingly. This is because the first pixel mean value statistics are the mean values of pixels that are not 0 after edge detection, that is, the intensity information of the target region edge. The larger the intensity of the edge is, the more obvious the current gradient information is, so that a larger weight needs to be set and the higher the enhancement degree is.
And enhancing the calculated low-frequency coefficient to obtain the corrected low-frequency coefficient. In other embodiments, the high frequency coefficients may also be subjected to soft thresholding, where the hard thresholding may well retain local features such as signal edges, and the soft thresholding is relatively smooth.
Step S230, performing inverse transformation of multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient, to obtain the noise-reduced target pixel.
The inverse transformation is also called inverse transformation, and is point transformation for restoring transformed pixel points in an image to original pixel points. In the case of a multi-scale geometric transform, which is an inverse wavelet transform, in particular, an inverse wavelet transform is an inverse process of the wavelet transform, after which an image needs to be reconstructed by the inverse wavelet transform. Therefore, in this embodiment, after obtaining the corrected low-frequency coefficient, reconstruction may be performed according to the high-frequency coefficient and the corrected low-frequency coefficient, so as to obtain the target pixel after noise reduction.
And step S240, obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image and the corresponding Gaussian image to obtain a de-noised endoscope image.
Taking each pixel in the laplacian image as a target pixel, performing noise reduction through the steps S210 to S230 to obtain a noise-reduced laplacian image.
Since the laplacian image and the gaussian image are images of the endoscope image in different frequency ranges, after the laplacian image is denoised, the laplacian image and the gaussian image need to be fused to obtain the denoised endoscope image.
In general, the endoscopic image is an enlarged organ, and the edge and the detail in the endoscopic image occupy a plurality of pixels, so through the above steps S210 to S240, when the laplacian image denoising process is performed, the embodiment performs the enhancement calculation on the low-frequency coefficient in the laplacian image based on the enhancement coefficient, and the low-frequency information in the laplacian image of the endoscopic image after the process is completed is retained or enhanced. Therefore, the method in the embodiment solves the problem that edge information in the endoscope image is lost by adopting a noise reduction method of a common camera image in the related art, retains the information of the edge area of the endoscope image through the enhancement processing of the low-frequency coefficient, and reduces the image loss.
The low-frequency information to be enhanced in the present application refers to low-frequency information in the laplacian image, and belongs to high-frequency information for the entire endoscopic image.
Specifically, when a patient is detected by an endoscope, the acquired endoscope image magnifies actual human tissue many times, and therefore, after the endoscope image is divided into high and low frequencies, high-frequency information useful for the entire endoscope image can be regarded as low frequencies in the extracted high-frequency portion, i.e., laplacian image, and high-frequency information useful for the entire endoscope image is information of the edge region. Compared with an endoscope image, the edge of a common image is thin, and corresponding high-frequency information and noise are not distinguished obviously, so that all high frequencies in the image can be uniformly filtered by the existing noise reduction method, and useful high-frequency information is ignored.
Preferably, in a target region of the laplacian image, before performing enhancement calculation on the low-frequency coefficient according to the enhancement coefficient, it is necessary to determine whether the target region is a continuous edge region, and perform enhancement calculation on the low-frequency coefficient in the target region according to the enhancement coefficient under the condition that the target region is the continuous edge region, where the edge region is a region where the gray level in the endoscopic image changes suddenly, and the continuous edge region is a region where the gray level changes continuously and suddenly. In this embodiment, a continuous edge area in an image is determined, and then a low-frequency coefficient of a target area that is the continuous edge area is enhanced, so as to retain low-frequency information of the continuous edge area and reduce the influence of discontinuous noise. Because the endoscope image has more continuous edges and details and the continuous edge area exists in the flat area, when the noise reduction processing of the laplacian layer is performed, if the target area belongs to low frequency and has continuous edges, the low frequency coefficient of the part is enhanced, and the edge area of the endoscope image can be reserved or enhanced after the processing is completed.
In some of these embodiments, the endoscopic image noise reduction method is implemented based on an image pyramid. Specifically, for a set of images of the same endoscopic image with different resolutions, the images of the images are arranged in a pyramid shape, and the images with different resolutions are arranged in a pyramid shape, wherein the smaller the hierarchy from the top of the pyramid to the bottom of the pyramid, the larger the image is, and the higher the resolution is.
In this embodiment, an image pyramid determined based on an endoscopic image is n layers, and the image pyramid includes a laplacian pyramid and a gaussian pyramid, the laplacian pyramid is a band-pass image pyramid obtained by forming an image representing a difference between consecutive layers of the gaussian pyramid, the laplacian pyramid is composed of laplacian images, the gaussian pyramid is composed of gaussian images, and in a case where i is greater than or equal to 0 and less than or equal to n-1, as for the laplacian image of the ith layer, fig. 3 is a flowchart of a method for determining a continuous edge region according to an embodiment of the present application, the method includes the following steps:
step S310, in the gradient image of the i +1 st layer Gaussian image of the image pyramid, a detection area corresponding to the target area is determined.
In this embodiment, the detection of the continuous edge region of the i-th layer laplacian image needs to be implemented based on the gradient image of the i + 1-th layer gaussian image. The gradient image refers to a change rate of a pixel in an image compared with an adjacent pixel in different directions, and can be obtained by performing edge detection on an i +1 th layer gaussian image, specifically, the edge detection can be realized by solving a first derivative or a second derivative of an image attribute, and the image attribute can be gray information or brightness information. Further, the detection region is a region corresponding to the position of the target region in the gradient image.
In other embodiments, before the edge detection, the i +1 st layer gaussian image may be denoised and upsampled.
Step S320, for each pixel in the detection area, determining that the target area corresponding to the detection area is a continuous edge area when the pixel value satisfies that the number of pixels in the preset pixel value range is greater than or equal to the preset number threshold.
Since the detection area corresponds to the target area position, it is also an area determined according to the position of the target pixel. Assuming that the target pixel is represented by p (x, y), and the detection region is an m × m region, when determining a continuous edge region according to the gradient image, a predetermined pixel value range, for example, a pixel having a pixel value different from 0 or a pixel having a pixel value greater than a certain pixel value, needs to be determined. And then calculating the number of pixels of the pixels in the preset pixel value range, and judging whether the target area is a continuous edge area or not according to the number of the pixels.
In the case where the number of pixels is greater than or equal to the preset number threshold, since the detection region is a part of the gradient image, it is indicated that there are a sufficient number of pixels with changed image properties in the detection region, and the corresponding target region can thus be identified as a continuous edge region. The preset number threshold value can be set by experience, and can also be adaptively modified according to actual scenes and requirements. In other embodiments, the preset number threshold may also be trained through a deep learning network to obtain an optimal determination result of the continuous edge region.
Through the above steps S310 and S320, in this embodiment, whether the target region in the i +1 th layer laplacian image is a continuous edge region is determined based on the gradient image of the i +1 th layer of gaussian image, so that the accuracy of determining the continuous edge region can be improved, and the influence of discontinuous noise can be further reduced.
In some embodiments, when the target region is a continuous edge region, the low-frequency coefficient obtained after the target region is subjected to multi-scale geometric transformation is subjected to enhancement calculation, otherwise, the low-frequency coefficient obtained after the multi-scale geometric transformation is used as a final low-frequency coefficient. For example, the number of pixels in the detection area that is not 0 is denoted by sumEdge, and the preset number threshold is denoted by numEdge, so that the low frequency coefficient can be calculated by the following formula 1:
In the formula 1, the first and second groups of the compound,p' denotes the low-frequency coefficient after correction,prepresenting the low frequency coefficients obtained after the multi-scale geometric transformation,ratiois the enhancement factor. Is determined firstsumEdgeWhether or not it is greater than or equal tonumEdgeIf, ifsumEdgeGreater than or equal tonumEdgeThen there are continuous edges in the target area and the low frequency coefficients can be corrected by the enhancement coefficients.
In some embodiments, in the process of changing the enhancement coefficient with the first pixel value mean value, it is necessary to set a maximum value and a minimum value for the enhancement coefficient according to the magnitude of the first pixel value mean value, and the enhancement coefficient linearly increases between the maximum value and the minimum value.
Specifically, when the first pixel value mean value is less than or equal to a first preset gradient threshold, the enhancement coefficient is a first fixed value, which is a minimum value of the enhancement coefficient, and may be set to 1, or may be set to other values according to a scene requirement; when the first pixel value mean value is larger than a first preset gradient threshold value and smaller than a second preset gradient threshold value, the enhancement coefficient is increased along with the first pixel value mean value, wherein the size of the second preset gradient threshold is determined according to the second pixel value mean value within the preset pixel value range in the gradient image, specifically, the second preset gradient threshold can be determined by the second pixel value mean value and the corresponding obtained adjusting parameter, the second pixel value average may be obtained based on the pixel value average within a preset pixel value range in the gradient image of the i +1 th layer gaussian image, for example, setting the pixel value to 0 when the pixel value is smaller than a first preset gradient threshold value for the gradient image obtained by edge detection, otherwise, keeping the pixel value in the original gradient image, accumulating and averaging pixels with pixel values different from 0 to obtain a second pixel value average value, wherein the adjusting parameters can be set in a self-defined mode; and under the condition that the mean value of the first pixel values is greater than or equal to a second preset gradient threshold value, the enhancement coefficient is a second fixed value which is greater than the first fixed value, so that unnatural gradient inversion or black and white edges caused by continuous increase of the enhancement coefficient are avoided.
The enhancement coefficient may increase linearly with the first pixel value mean value when the first pixel value mean value is greater than a first preset gradient threshold value and less than a second preset gradient threshold value. In the linear increasing process, the slope is calculated according to a point determined by the first preset gradient threshold and the first fixed value and a point determined by the second preset gradient threshold and the second fixed value. As shown in equation 2:
In the formula 2, the first and second groups,ratiorepresenting the enhancement factor, 1 is a first constant value, A is a second constant value,thrEdgeis a first preset gradient threshold value and is,avgEdge1is the average value of the first pixel values,
is the second preset gradient threshold value and is,wfor a tuning parameter greater than 1, the value of,avgEdgeis the second pixel value mean.
FIG. 4 is a graph illustrating the variation of enhancement coefficients with the mean value of the first pixel values, as shown in FIG. 4, according to an embodiment of the present applicationavgEdge1Is less than or equal tothrEdgeIn the case of (a) in (b),ratiohas a value of 1 inavgEdge1Greater than or equal to
In the case of (a) the (b),ratiohas a value of A inavgEdge1Is located atthrEdgeAnd
in the case of (a) in (b),ratiofollowed byavgEdge1And (4) linear growth.
On the other hand, when the first pixel value mean is greater than the first preset gradient threshold and smaller than the second preset gradient threshold, the enhancement coefficient may be obtained through gauss transformation and increases as an S-curve with the first pixel value mean, as shown in formula 3:
equation 3
In the formula 3, the first and second groups,sigmafor a parameter variance of the gauss transform, the fat-thin of the gaussian can be determined. In particular, the amount of the solvent to be used,sigmacan be calculated according to the first fixed value 1 and the second fixed value A. In thatavgEdge1Is located atthrEdgeAnd
in the case of (a) in (b),ratiofollowed byavgEdge1The surface of the steel plate is increased in an S-shaped curve, and the smoothness is better.
In this embodiment, the enhancement coefficient may be adaptively adjusted according to edge gradient changes of different target regions, so as to improve the accuracy of denoising the endoscopic image.
In some embodiments, fig. 5 is a flowchart of a method for determining a gradient image according to an embodiment of the present application, and as shown in fig. 5, the method includes the following steps:
step S510, edge detection is carried out on the i +1 layer Gaussian image to obtain an initial gradient image.
The edge detection may be implemented by obtaining a first derivative or a second derivative of an image attribute, where the image attribute may be gray scale information or luminance information.
Step S520, according to the first preset gradient threshold, selecting a preset pixel value or an initial pixel value in the initial gradient image as a pixel value in the gradient image of the final i +1 th layer gaussian image.
After obtaining an initial gradient image, adjusting pixel values in the gradient image according to a first preset gradient threshold, specifically, in the initial gradient image, if the initial pixel value is greater than or equal to the preset gradient threshold, taking the initial pixel value as a pixel value in a gradient image of a final i +1 th layer gaussian image; and if the initial pixel value is smaller than the preset gradient threshold value, taking the preset pixel value as the pixel value in the gradient image of the final i +1 th layer Gaussian image. As shown in equation 4:
In equation 4, for the initial gradient image, the pixel value less than the first preset gradient threshold will be set to 0, otherwise the pixel value in the initial gradient image is kept unchanged.
Through the steps S510 and S520, the initial gradient image is binarized according to the first preset gradient threshold, and the continuous edge region is determined based on the binarized gradient image, so that the calculation mode can be simplified, and the calculation efficiency can be improved.
In some embodiments, when the image pyramid is n layers, the process of obtaining the noise-reduced endoscope image by fusing the noise-reduced laplacian image and the corresponding gaussian image specifically includes: fusing the i +1 th layer of Gaussian image and the i layer of Laplacian image from the bottom of the image pyramid to the top to obtain an i layer of Gaussian image; and when the value of i is 0, taking the 0 th layer Gaussian image as the endoscope image after noise reduction. For the multilayer image pyramids, the resolution of the image representation of each layer of pyramid is different, the laplacian images of each layer are subjected to noise reduction in sequence, and then the laplacian images subjected to noise reduction are fused layer by layer, so that more information in the endoscope image can be reserved.
In some embodiments, if the enhancement coefficient is too large, the noise-reduced endoscopic image may have a significant blocking effect, and therefore, after the noise-reduced target pixel is obtained, the blocking effect may be weakened by adopting a manner of weighted overlap of pixel values of pixels in a neighborhood. Therefore, after performing the multi-scale geometric transformation, not only the processing value of the target pixel is retained, but also the pixel values around the target pixel can be multiplied by a certain ratio to act on the processed target pixel, as shown in equation 5:
In the formula 5, the first and second groups,
in order to be the target pixel, the pixel is,
in order to obtain the corrected target pixel,
to
Are the pixels in the neighborhood of the pixel,
to
The neighborhood of the target pixel is determined by a second predetermined range for the weights corresponding to the pixels in the neighborhood, so that each pixel in the second predetermined range is based onDetermining the weight of each pixel in a second preset range according to the distance between the pixel and the target pixel; and correcting the pixel value of the target pixel according to the pixel value and the weight of each pixel in the second preset range to obtain the corrected target pixel. Wherein the respective weights may be determined by a gaussian distribution.
The present embodiment is described and illustrated below by means of preferred embodiments.
Fig. 6 is a schematic diagram of an endoscopic image noise reduction method according to the present preferred embodiment, as shown in fig. 6.
In this embodiment, a 2-layer image pyramid is constructed according to an endoscopic image, an original endoscopic image is recorded as g0, g0 is downsampled to obtain a first-layer gaussian image which is recorded as g1, and g0 after upsampling is subtracted from g0 to obtain a laplacian image L0. Downsampling g0 twice in sequence to obtain a second layer gaussian image, which is denoted as g 2. The upsampled g2 is subtracted from g1 to obtain a laplacian image L1.
1. And g2 'is obtained after g2 is subjected to noise reduction and upsampling, an initial gradient image is obtained by performing edge detection on g 2', and the initial gradient image is subjected to binarization processing according to thrEdge to obtain a final gradient image.
2. Performing wavelet edge-preserving denoising processing on the laplacian image L1 by using a wavelet algorithm, taking a target pixel p (x, y) as an example, the wavelet edge-preserving denoising processing comprises the following steps:
(1) taking p (x, y) as a center, judging whether the number sumEdge of pixels which are not 0 in the range of m multiplied by m is larger than numEdge or not, and if the sumEdge is larger than a threshold numEdge, judging that a continuous edge area exists in the current target area;
(2) when a continuous edge region exists in the target region, assuming that the value of m is 8, 3-layer wavelet processing is performed on a small block of 8 × 8 pixels centered around p (x, y), and a high-frequency coefficient and a low-frequency coefficient are obtained. When the value of m changes, the layer number of wavelet processing also changes, the high-frequency coefficient is processed by a hard threshold, and the low-frequency coefficient is adaptively adjusted according to the pixel mean value avgEdge1 which is not 0 in the range of m multiplied by m;
(3) after the corrected low-frequency coefficient is obtained, the denoised pixel is obtained through inverse wavelet transform, and then the denoised pixel is corrected again in a neighborhood pixel weighted superposition mode, so that the block effect of the denoised image is avoided.
3. And obtaining a Laplace image after wavelet edge-preserving and noise reduction according to the target pixel after the block effect correction, and marking the Laplace image as L1'.
4. Adding and fusing g2 ' and L1 ', and sequentially denoising and upsampling the fused image to obtain a Gaussian image g1 ';
5. processing the L0 according to a method of 1-3 to obtain L0 ', wherein in the process of performing wavelet edge preservation and denoising on the L0, edge detection needs to be performed on the g 1' to determine a continuous edge region;
6. and g1 ' and L0 ' are fused and subjected to noise reduction to obtain a final output image g0 '.
Because the edge of the endoscope image is lost when the conventional denoising method is used for denoising, in the embodiment, wavelet denoising is applied when the high-frequency information in the laplacian image is denoised, if the target region corresponding to the low-frequency coefficient obtained by the wavelet denoising belongs to the continuous edge region, the low-frequency information in the frequency domain is enhanced, and only the high-frequency coefficient is subjected to threshold processing compared with the common wavelet denoising, the method in the embodiment can enable the edge region of the endoscope to be reserved or enhanced.
It should be noted that the steps illustrated in the above-described flow diagrams or in the flow diagrams of the figures may be performed in a computer system, such as a set of computer-executable instructions, and that, although a logical order is illustrated in the flow diagrams, in some cases, the steps illustrated or described may be performed in an order different than here.
In this embodiment, an endoscopic image noise reduction device is further provided, which is used to implement the above embodiments and preferred embodiments, and the description of the device already described is omitted. The terms "module," "unit," "subunit," and the like as used below may implement a combination of software and/or hardware for a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
Fig. 7 is a block diagram of the configuration of the endoscopic image noise reduction device of the present embodiment, and as shown in fig. 7, in the case where the endoscopic image includes a laplacian image and a gaussian image, the device includes a multi-scale geometric transformation module 71, an enhancement calculation module 72, a noise reduction calculation module 73, and a fusion module 74:
a multi-scale geometric transformation module 71, configured to perform multi-scale geometric transformation on a target region in the laplacian image to obtain a high-frequency coefficient and a low-frequency coefficient of the target region, where the target region is determined according to a position of a target pixel; specifically, the target area is determined according to a first preset range with the target pixel as a center.
The enhancement calculation module 72 is configured to perform enhancement calculation on the low-frequency coefficient according to an enhancement coefficient to obtain a corrected low-frequency coefficient, where the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target region;
the noise reduction calculation module 73 is configured to perform inverse transformation of multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient to obtain a noise-reduced target pixel;
and the fusion module 74 is configured to obtain a noise-reduced laplacian image according to the noise-reduced target pixel, and fuse the noise-reduced laplacian image with a corresponding gaussian image to obtain a noise-reduced endoscope image.
In general, the endoscopic image is an enlarged organ, and edges and details in the endoscopic image occupy a plurality of pixels, so that with the above apparatus, when performing the laplacian image denoising process, the enhancement calculation module 72 of this embodiment performs the enhancement calculation on the low-frequency coefficient in the laplacian image based on the enhancement coefficient, and after the process is completed, the low-frequency information of the laplacian image in the endoscopic image is retained or enhanced. Therefore, the method in the embodiment solves the problem that the edge information in the endoscope image is lost by adopting a noise reduction method of a common camera image in the related art, retains the information of the edge area of the endoscope image by enhancing the low-frequency coefficient, and reduces the image loss
In some embodiments, the endoscopic image noise reduction apparatus includes a continuous edge determination module configured to determine whether the target region is a continuous edge region; and under the condition that the target area is a continuous edge area, performing enhancement calculation on the low-frequency coefficient in the target area according to an enhancement coefficient.
In some embodiments, the image pyramid determined based on the endoscopic image is n layers, the image pyramid includes a laplacian pyramid and a gaussian pyramid, the laplacian pyramid is composed of laplacian images, the gaussian pyramid is composed of gaussian images, and in the case that i is greater than or equal to 0 and less than or equal to n-1, for the i-th layer laplacian image, the continuous edge determining module is further configured to determine a detection region corresponding to the target region in a gradient image of the i + 1-th layer gaussian image of the image pyramid; and for each pixel in the detection area, determining a target area corresponding to the detection area as the continuous edge area under the condition that the pixel value meets the condition that the number of pixels in the preset pixel value range is greater than or equal to a preset number threshold.
In some of these embodiments, the enhancement calculation module 72 is further configured to determine the enhancement factor as a first fixed value if the first pixel value mean is smaller than or equal to a first preset gradient threshold; when the first pixel value mean value is larger than the first preset gradient threshold value and smaller than a second preset gradient threshold value, the enhancement coefficient is increased along with the first pixel value mean value, wherein the size of the second preset gradient threshold value is determined according to a second pixel value mean value in the preset pixel value range in the gradient image; and when the first pixel value mean value is greater than or equal to the second preset gradient threshold value, the enhancement coefficient is a second fixed value, and the second fixed value is greater than the first fixed value.
In some embodiments, the enhancement calculation module 72 is further configured to perform edge detection on the i +1 th layer gaussian image to obtain an initial gradient image; and selecting a preset pixel value or an initial pixel value in the initial gradient image as a pixel value in the gradient image of the final i +1 th layer Gaussian image according to the first preset gradient threshold value.
In some embodiments, the fusion module 74 is further configured to fuse the i +1 th layer gaussian image with the i layer laplacian image to obtain an i layer gaussian image; and when the value of i is 0, taking the 0 th layer Gaussian image as the endoscope image after noise reduction.
In some embodiments, the endoscopic image noise reduction device further includes a modification module, configured to determine a weight of each pixel in a second preset range according to a distance between each pixel in the second preset range and the target pixel; and correcting the pixel value of the target pixel according to the pixel value of each pixel in the second preset range and the weight.
The above modules may be functional modules or program modules, and may be implemented by software or hardware. For a module implemented by hardware, the modules may be located in the same processor; or the modules can be respectively positioned in different processors in any combination.
There is also provided in this embodiment an electronic device comprising a memory having a computer program stored therein and a processor arranged to run the computer program to perform the steps of any of the above method embodiments.
Optionally, the electronic apparatus may further include a transmission device and an input/output device, wherein the transmission device is connected to the processor, and the input/output device is connected to the processor.
Optionally, in this embodiment, the processor may be configured to execute the following steps by a computer program:
s1, performing multi-scale geometric transformation on a target region in the Laplace image to obtain a low-frequency coefficient of the target region, wherein the target region is determined according to the position of a target pixel;
s2, performing enhancement calculation on the low-frequency coefficient according to an enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target area;
s3, performing inverse transformation of multi-scale geometric transformation on the pixel value of the target pixel according to the corrected low-frequency coefficient to obtain a noise-reduced target pixel;
and S4, obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image and the corresponding Gaussian image to obtain a de-noised endoscope image.
It should be noted that, for specific examples in this embodiment, reference may be made to the examples described in the foregoing embodiments and optional implementations, and details are not described again in this embodiment.
In addition, in combination with the endoscope image noise reduction method provided in the foregoing embodiment, a storage medium may also be provided in this embodiment. The storage medium has a computer program stored thereon; the computer program, when executed by a processor, implements any one of the endoscopic image noise reduction methods in the above embodiments.
It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to be limiting. All other embodiments, which can be derived by a person skilled in the art from the examples provided herein without any inventive step, shall fall within the scope of protection of the present application.
It is obvious that the drawings are only examples or embodiments of the present application, and it is obvious to those skilled in the art that the present application can be applied to other similar cases according to the drawings without creative efforts. Moreover, it should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which may vary from one implementation to another.
The term "embodiment" is used herein to mean that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is to be expressly or implicitly understood by one of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the patent protection. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims (10)
1. An endoscopic image noise reduction method, wherein an endoscopic image includes a laplacian image and a gaussian image, and wherein reducing noise of pixels in the laplacian image comprises:
performing multi-scale geometric transformation on a target region in the Laplace image to obtain a low-frequency coefficient and a high-frequency coefficient of the target region, wherein the target region is determined according to the position of a target pixel;
performing enhancement calculation on the low-frequency coefficient according to an enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target area;
carrying out inverse transformation of the multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient to obtain a noise-reduced target pixel;
obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image with a corresponding Gauss image to obtain a de-noised endoscope image;
under the condition that the image pyramid is n layers, the process of obtaining the endoscope image after noise reduction by fusing the laplacian image after noise reduction and the corresponding gaussian image specifically comprises the following steps: fusing the i +1 th layer of Gaussian image and the i layer of Laplacian image from the bottom of the image pyramid to the top to obtain an i layer of Gaussian image; and when the value of i is 0, taking the 0 th layer Gaussian image as the endoscope image after noise reduction.
2. The endoscopic image noise reduction method according to claim 1, wherein before performing enhancement calculation on the low frequency coefficient based on an enhancement coefficient, the method comprises:
judging whether the target area is a continuous edge area or not;
and under the condition that the target area is a continuous edge area, performing enhancement calculation on the low-frequency coefficient in the target area according to an enhancement coefficient.
3. The endoscopic image noise reduction method according to claim 2, wherein an image pyramid determined based on the endoscopic image is n layers, the image pyramid includes a laplacian pyramid and a gaussian pyramid, the laplacian pyramid is composed of laplacian images, the gaussian pyramid is composed of gaussian images, and in the case where i is greater than or equal to 0 and less than or equal to n-1, the determining whether the target region is a continuous edge region for an i-th layer laplacian image includes:
determining a detection region corresponding to the target region in a gradient image of the i +1 st layer Gaussian image of the image pyramid;
and for each pixel in the detection area, determining a target area corresponding to the detection area as the continuous edge area under the condition that the pixel value meets the condition that the number of pixels in the preset pixel value range is greater than or equal to a preset number threshold.
4. The endoscopic image noise reduction method according to claim 3, wherein determining the magnitude of the enhancement coefficient according to the first pixel value mean of pixels within a preset pixel value range in the target region comprises:
when the first pixel value mean value is smaller than or equal to a first preset gradient threshold value, the enhancement coefficient is a first fixed value;
when the first pixel value mean value is larger than the first preset gradient threshold value and smaller than a second preset gradient threshold value, the enhancement coefficient is increased along with the first pixel value mean value, wherein the size of the second preset gradient threshold value is determined according to a second pixel value mean value in the preset pixel value range in the gradient image;
and when the first pixel value mean value is greater than or equal to the second preset gradient threshold value, the enhancement coefficient is a second fixed value, and the second fixed value is greater than the first fixed value.
5. The endoscopic image noise reduction method according to claim 4, wherein the method of determining the gradient image of the i +1 th layer gaussian image comprises:
performing edge detection on the i +1 layer Gaussian image to obtain an initial gradient image;
and selecting a preset pixel value or an initial pixel value in the initial gradient image as a pixel value in the gradient image of the final i +1 th layer Gaussian image according to the first preset gradient threshold value.
6. The endoscopic image noise reduction method according to claim 1, wherein after obtaining the noise-reduced target pixel, the method comprises:
determining the weight of each pixel in a second preset range according to the distance between each pixel in the second preset range and the target pixel;
and correcting the pixel value of the target pixel according to the pixel value of each pixel in the second preset range and the weight.
7. The endoscopic image noise reduction method according to claim 1, wherein the target region determination based on the position of the target pixel comprises:
and determining the target area according to a first preset range by taking the target pixel as a center.
8. An endoscopic image noise reduction device, characterized in that an endoscopic image comprises a laplacian image and a gaussian image, the device comprising a multi-scale geometric transformation module, an enhancement computation module, a noise reduction computation module and a fusion module:
the multi-scale geometric transformation module is used for performing multi-scale geometric transformation on a target area in the Laplace image to obtain a low-frequency coefficient and a high-frequency coefficient of the target area, wherein the target area is determined according to the position of a target pixel;
the enhancement calculation module is used for performing enhancement calculation on the low-frequency coefficient according to an enhancement coefficient to obtain a corrected low-frequency coefficient, wherein the size of the enhancement coefficient is determined according to a first pixel value mean value of pixels in a preset pixel value range in the target area;
the noise reduction calculation module is used for performing inverse transformation of the multi-scale geometric transformation on the pixel value of the target pixel according to the high-frequency coefficient and the corrected low-frequency coefficient to obtain a noise-reduced target pixel;
the fusion module is used for obtaining a de-noised Laplacian image according to the de-noised target pixel, and fusing the de-noised Laplacian image with a corresponding Gaussian image to obtain a de-noised endoscope image;
under the condition that the image pyramid is n layers, the process of obtaining the endoscope image after noise reduction by fusing the laplacian image after noise reduction and the corresponding gaussian image specifically comprises the following steps: from the bottom of the image pyramid to the top, fusing the i +1 th layer Gaussian image with the i layer Laplacian image to obtain an i layer Gaussian image; and when the value of i is 0, taking the 0 th layer Gaussian image as the endoscope image after noise reduction.
9. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, and the processor is configured to execute the computer program to perform the endoscopic image noise reduction method according to any one of claims 1 to 7.
10. A storage medium, in which a computer program is stored, wherein the computer program is arranged to perform the steps of the endoscopic image noise reduction method according to any one of claims 1 to 7 when executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210174607.6A CN114240941B (en) | 2022-02-25 | 2022-02-25 | Endoscope image noise reduction method, device, electronic apparatus, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210174607.6A CN114240941B (en) | 2022-02-25 | 2022-02-25 | Endoscope image noise reduction method, device, electronic apparatus, and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114240941A CN114240941A (en) | 2022-03-25 |
| CN114240941B true CN114240941B (en) | 2022-05-31 |
Family
ID=80748171
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210174607.6A Active CN114240941B (en) | 2022-02-25 | 2022-02-25 | Endoscope image noise reduction method, device, electronic apparatus, and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114240941B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115345800B (en) * | 2022-10-18 | 2023-04-07 | 极限人工智能有限公司 | Self-adaptive noise reduction method and system for medical endoscope moving image |
| CN117173031A (en) * | 2023-06-01 | 2023-12-05 | 广州科易光电技术有限公司 | Image processing method and device for gas, electronic equipment and storage medium |
| CN116681788B (en) * | 2023-06-02 | 2024-04-02 | 萱闱(北京)生物科技有限公司 | Image electronic dyeing method, device, medium and computing equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2015198875A (en) * | 2014-04-10 | 2015-11-12 | Hoya株式会社 | Endoscope image processing device and endoscope system |
| CN105913075A (en) * | 2016-04-05 | 2016-08-31 | 浙江工业大学 | Endoscopic image focus identification method based on pulse coupling nerve network |
| CN110706174A (en) * | 2019-09-27 | 2020-01-17 | 集美大学 | Image enhancement method, terminal equipment and storage medium |
| CN111784605A (en) * | 2020-06-30 | 2020-10-16 | 珠海全志科技股份有限公司 | Image denoising method based on region guidance, computer device and computer readable storage medium |
| CN113793272A (en) * | 2021-08-11 | 2021-12-14 | 东软医疗系统股份有限公司 | Image noise reduction method and device, storage medium and terminal |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1938274A2 (en) * | 2005-09-12 | 2008-07-02 | D.V.P. Technologies Ltd. | Medical image processing |
| CN109993703A (en) * | 2017-12-29 | 2019-07-09 | 展讯通信（上海）有限公司 | Multi-scale image noise-reduction method and device |
| CN112734663A (en) * | 2020-12-30 | 2021-04-30 | 浙江工业大学 | Profile wave transformation medical CT image noise reduction method based on self-adaptive threshold |
-
2022
- 2022-02-25 CN CN202210174607.6A patent/CN114240941B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2015198875A (en) * | 2014-04-10 | 2015-11-12 | Hoya株式会社 | Endoscope image processing device and endoscope system |
| CN105913075A (en) * | 2016-04-05 | 2016-08-31 | 浙江工业大学 | Endoscopic image focus identification method based on pulse coupling nerve network |
| CN110706174A (en) * | 2019-09-27 | 2020-01-17 | 集美大学 | Image enhancement method, terminal equipment and storage medium |
| CN111784605A (en) * | 2020-06-30 | 2020-10-16 | 珠海全志科技股份有限公司 | Image denoising method based on region guidance, computer device and computer readable storage medium |
| CN113793272A (en) * | 2021-08-11 | 2021-12-14 | 东软医疗系统股份有限公司 | Image noise reduction method and device, storage medium and terminal |
Non-Patent Citations (2)
| Title |
|---|
| An impedance mismatching compensation based CM noise rejection method for capsule endoscope system;Ju Hyung Song et al.;《2014 Sixth International Conference on Ubiquitous and Future Networks (ICUFN)》;20140814;284-288 * |
| 基于高斯金字塔和拉普拉斯金字塔融合的图像对比度增强算法研究;马永强;《信息与电脑》;20180430;38-40 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114240941A (en) | 2022-03-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114240941B (en) | Endoscope image noise reduction method, device, electronic apparatus, and storage medium | |
| Al-Ameen et al. | An innovative technique for contrast enhancement of computed tomography images using normalized gamma-corrected contrast-limited adaptive histogram equalization | |
| Karaoğlu et al. | Removal of speckle noises from ultrasound images using five different deep learning networks | |
| Kimmel et al. | A variational framework for retinex | |
| Dai et al. | Retinal fundus image enhancement using the normalized convolution and noise removing | |
| Coupé et al. | Bayesian non local means-based speckle filtering | |
| Wu et al. | Contrast-accumulated histogram equalization for image enhancement | |
| Ben Abdallah et al. | Adaptive noise-reducing anisotropic diffusion filter | |
| US9443286B2 (en) | Gray image processing method and apparatus based on wavelet transformation | |
| Li et al. | Robust retinal image enhancement via dual-tree complex wavelet transform and morphology-based method | |
| WO2013038333A2 (en) | Method and system for enhancing image quality | |
| JP2001057677A (en) | Image processing method, system and recording medium | |
| US9401009B2 (en) | Method and apparatus for enhancing quality of 3D image | |
| KR102195047B1 (en) | Method and apparatus for enhancing quality of 3D image | |
| Wang et al. | Total generalized variation-based Retinex image decomposition | |
| Yao et al. | An improved multi-scale image enhancement method based on retinex theory | |
| CN113192067B (en) | Intelligent prediction method, device, equipment and medium based on image detection | |
| CN114155161A (en) | Image denoising method and device, electronic equipment and storage medium | |
| Paranjape | Fundamental enhancement techniques | |
| CN112334942A (en) | Image processing method and device | |
| SA et al. | Enhanced Homomorphic Unsharp Masking method for intensity inhomogeneity correction in brain MR images | |
| JP6256680B2 (en) | Image processing method, image processing apparatus, and image processing program | |
| Judson et al. | Efficient and robust non-local means denoising methods for biomedical images | |
| Vyas et al. | Image enhancement | |
| JP2004233124A (en) | Radar image processing device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |