Endoscopic image imaging method and endoscope

Abstract

The present invention relates to the technical field of medical endoscopic imaging. Disclosed are an endoscopic image imaging method and an endoscope. The imaging method comprises: when the optical power of a first narrow-band light source changes, acquiring an original acquired image of a target object acquired by an imaging unit; decomposing from an original first color channel a standard first color channel of the imaging unit corresponding to a standard optical power; fusing the standard first color channel, an original second color channel and an original third color channel to obtain a corrected image; and separately extracting brightness information representing brightness and details in the original acquired image and color information in a standard image, and fusing the brightness information and the color information to obtain an endoscopic image. Thus, the adjusted endoscopic image retains the detail enhancement brought by the variation of the proportion of the first narrow-band light source while maintaining the color representation of imaging under preset standard light.

Description

Endoscopic imaging methods and endoscopes Technical Field

[0001] This invention relates to the field of medical endoscopic imaging technology, specifically to endoscopic image imaging methods and endoscopes. Background Technology

[0002] With the booming development of endoscopic applications, doctors' demand for high-quality endoscopic imaging is also constantly increasing. In endoscopic application scenarios, especially in the examination, diagnosis and treatment of digestive endoscopy (gastrointestinal endoscopy), doctors usually use the white light imaging mode that conforms to human vision for routine operations. When it is necessary to further explore the lesion, they will switch to the narrow band light imaging mode.

[0003] Narrow-band light imaging is an optical imaging technique that uses one or more narrow-band light sources as illumination to enhance the visibility of mucosal blood vessels, lesions, inflammation, and other sites. Common narrow-band light imaging modes include narrow-band imaging (NBI), blue laser imaging (BLI), and linked color imaging (LCI). Because narrow-band light imaging only covers a portion of the visible spectrum or alters the proportion of white light spectrum, the colors of the images produced will differ significantly from those produced by white light imaging.

[0004] Because different parts of the human body cavity have different absorption and reflection characteristics for blue and blue-violet light, adjusting the proportion of blue and blue-violet light in specific areas can effectively improve the contrast between mucous membranes, blood vessels, and tissues. For example, the oral mucosa epithelium, mainly composed of stratified squamous epithelium, the gastric mucosa epithelium, mainly composed of columnar epithelial cells, and the duodenum, which receives large amounts of bilirubin from the liver and is excreted through the bile ducts, all exhibit significantly different absorption and reflection characteristics for blue and blue-violet light. Appropriately increasing or decreasing the proportion of blue and blue-violet light will significantly improve the contrast and detail information of the corresponding areas.

[0005] However, for white light illumination, the power ratio of each LED light source at a given white light brightness level is fixed. Changing the ratio of blue light or blue-violet light will significantly affect light quality parameters such as color rendering index, color temperature, and chromaticity coordinates, which will reduce the color reproduction capability of the camera system. At the same time, since there are many levels for each LED light source, there will be even more combinations of levels for multiple light sources. Changes in each combination may lead to changes in the color rendering of the light source. It is impossible to match corresponding image parameters such as white balance and CCM for each combination.

[0006] Therefore, how to maintain the color expression of white light imaging while preserving the detail enhancement brought about by varying the proportion of narrowband light sources has become an urgent problem to be solved. Summary of the Invention

[0007] In view of this, the present invention provides an endoscopic image imaging method, apparatus, computer device and storage medium and endoscope to solve the technical problem of how to maintain the color expression of white light imaging while retaining the detail enhancement brought about by the varying narrow band light source ratio.

[0008] According to a first aspect, the present invention provides an endoscopic image imaging method, wherein the endoscope includes an illumination unit, an imaging unit, and an image processing unit, wherein the illumination unit includes at least one first narrowband light source corresponding to at least one first band of a blue channel in an RGB image acquired by the imaging unit, and multiple second light sources corresponding to multiple second bands of a red channel and a green channel, wherein the first narrowband light source and the second light sources can synthesize a preset standard light at corresponding standard optical power, and under the illumination of the preset standard light, the imaging unit can acquire a standard image; the endoscopic image imaging method is applicable to the image processing unit, and the endoscopic image imaging method includes: in the first narrowband... When the light power of the light source changes, the original image of the target object acquired by the imaging unit at the current light power is obtained; wherein, the original image is composed of an original first color channel, an original second color channel, and an original third color channel; the standard first color channel of the imaging unit corresponding to the standard light power is decomposed from the original first color channel; the standard first color channel, the original second color channel, and the original third color channel are fused to obtain a corrected image; the brightness information representing brightness and detail in the original image and the color information in the standard image are extracted respectively, and the brightness information and the color information are fused to obtain an endoscopic imaging image.

[0009] In one embodiment, the step of decomposing the standard first color channel of the imaging unit corresponding to the standard optical power in the original first color channel includes: decomposing at least one original first color sub-channel of the imaging unit corresponding to the at least one first narrowband light source at the current optical power in the original first color channel, wherein each original first color sub-channel corresponds to one first narrowband light source; decomposing the standard first color sub-channel of the imaging unit corresponding to the first narrowband light source at the standard optical power in the original first color sub-channel; and fusing the standard first color sub-channels to obtain the standard first color channel.

[0010] In one embodiment, the first band includes a narrowband with a center wavelength of 415±10nm and / or a narrowband with a center wavelength of 460±10nm. Correspondingly, the first narrowband light source includes a narrowband blue-violet light source with a center wavelength of 415±10nm and / or a narrowband blue light source with a center wavelength of 460±10nm. The original first color channel includes an original blue channel. The original first color sub-channel includes an original sub-blue channel. The standard first color sub-channel includes a standard sub-blue channel.

[0011] In one embodiment, the step of decomposing at least one original sub-first color channel of the imaging unit corresponding to the at least one first narrowband light source at the current optical power from the original first color channel includes: decomposing the original blue channel into original sub-blue channels of the imaging unit corresponding to the narrowband blue-violet light source and the narrowband blue light source at the current optical power using a spectral reconstruction method; or, inputting the original blue channel into a pre-trained original sub-blue channel decomposition model to obtain the original sub-blue channels of the imaging unit corresponding to the narrowband blue-violet light source and the narrowband blue light source at the current optical power, wherein the original sub-blue channel decomposition model is based on D... train1 (i,j), D GT1 (i) and D GT1 The dataset consisting of (j) learns to incorporate D during model training. train1 (i,j) is mapped to D GT1 (i) or D GT1 The mapping relationship obtained at (j) is described in D; train1 (i,j) is obtained by simultaneously enabling only the narrowband blue-violet light source and the narrowband blue light source, and acquiring the first preset blue channel image data corresponding to each power level combination of the narrowband blue-violet light source and the narrowband blue light source; where, D train1 (i,j) represents the preset blue channel image data corresponding to the i-th narrowband blue-violet light source power level and the j-th narrowband blue light source power level; the D GT1 (i) Obtained by individually enabling the narrowband blue-violet light source and collecting second preset blue channel data at all adjustable power levels of the narrowband blue-violet light source; the D GT1 (j) The third preset blue channel data is obtained by enabling the narrowband blue light source alone and collecting data at all adjustable power levels of the narrowband blue light source.

[0012] In one embodiment, the step of decomposing the original first color sub-channels to extract the standard first color sub-channels of the imaging unit corresponding to the first narrowband light source at standard optical power includes: decomposing the original sub-blue channels to extract the standard sub-blue channels of the imaging unit corresponding to the narrowband blue-violet light source and the narrowband blue light source at standard optical power using the calculation theory of color constancy perception; or, inputting each of the original sub-blue channels into a standard sub-blue channel decomposition model to obtain the standard blue channels corresponding to the original sub-blue channels; wherein, the standard blue channel decomposition model is based on D... train2 (i,j), D GT2 (i) The dataset was trained using the dataset; the D train2 (i,j) is obtained by acquiring blue channel data at each white light brightness level while keeping other light source levels unchanged, only changing the original sub-blue channel corresponding to the narrowband blue-violet light source or the narrowband blue light source, traversing all adjustable levels of the white light brightness level and the narrowband blue-violet light source or the narrowband blue light source; representing the preset standard sub-blue channel image corresponding to the i-th white light brightness level and the j-th level of the narrowband blue-violet light source or the narrowband blue light source; the D GT2 (i) is obtained by collecting standard sub-blue channel data where the optical power remains unchanged for each white light brightness level, representing the preset sub-standard channel image corresponding to the i-th white light brightness level; during the training of the standard sub-blue channel decomposition model, it learns to decompose the D corresponding to the i-th white light brightness level. train2 (i,j) is mapped to D GT2 The mapping relationship is obtained when (i).

[0013] In one embodiment, fusing the standard first color sub-channels to obtain the standard first color channel includes: obtaining preset sub-fusion weights for each of the standard sub-blue channels during fusion; wherein each preset sub-fusion weight is determined by obtaining a synthesized blue channel image at a target power corresponding to each of the standard sub-blue channels and a decomposed blue channel image when the narrowband blue light source and the narrowband blue-violet light source are individually enabled based on the target power corresponding to each of the standard sub-blue channels, and by fusing the decomposed blue channel images into a blue channel image; and fusing based on the preset sub-fusion weights corresponding to each of the standard sub-blue channels to obtain the standard blue channel.

[0014] In one embodiment, fusing the standard first color sub-channels to obtain the standard first color channel further includes: inputting each of the standard sub-blue channels into a standard sub-blue channel fusion model to obtain the standard blue channel; wherein, the standard blue channel fusion model is based on D... train1 (i), D train2(i) and D GT3 (i) The dataset was trained using the dataset; the D train (i) is obtained by individually enabling the narrowband blue-violet light source, traversing the adjustable levels corresponding to the white light brightness levels where the narrowband blue-violet light source is located, and collecting the corresponding blue channel image data at each level; represents the first standard sub-blue channel image corresponding to the i-th narrowband blue-violet light source brightness level; the D train2 (i) is obtained by individually enabling the narrowband blue light source, traversing the adjustable levels corresponding to the white light brightness levels where the narrowband blue light source is located, and collecting the corresponding blue channel image data at each level; represents the second standard sub-blue channel image corresponding to the i-th narrowband blue light source brightness level; D GT3 (i) is obtained by traversing all adjustable levels of the narrowband blue-violet light source and the narrowband blue light source according to the preset white light brightness level, and collecting the corresponding blue channel data under the preset white light brightness level; represents the synthesized standard blue channel image corresponding to the brightness level of the i-th group of narrowband blue-violet light sources and the brightness level of the narrowband blue light source; during the training of the standard sub-blue channel fusion model, the mapping relationship is learned when mapping the first standard sub-blue channel image corresponding to the brightness level of the i-th narrowband blue-violet light source and the second standard sub-blue channel image corresponding to the brightness level of the i-th narrowband blue-violet light source to the synthesized standard blue channel image corresponding to the brightness level of the i-th group of narrowband blue-violet light sources and the brightness level of the narrowband blue light source.

[0015] In one embodiment, the step of extracting luminance information representing brightness and detail from the original acquired image and color information from the standard image, and fusing the luminance information and the color information to obtain an endoscopic imaging image, includes: mapping the original acquired image and the corrected image to independent color spaces to obtain a first mapped image corresponding to the original acquired image and a second mapped image corresponding to the color corrected image; wherein, the independent color space includes at least one of YUV space, Lab space, and HSV space; extracting a first channel containing luminance information from the first mapped image; extracting a second channel containing color information and a third channel containing saturation information from the second mapped image; and fusing the first channel, the second channel, and the third channel to obtain the endoscopic imaging image.

[0016] In one embodiment, before acquiring the original image of the target object captured by the imaging unit at the current optical power, the method further includes: acquiring an initial imaging image of the target object and an optical power coefficient table corresponding to the target object at all brightness levels of a preset standard light; wherein, the initial imaging image is composed of an initial blue channel, an initial red channel, and an initial green channel; the optical power coefficient table is used to represent the power coefficient and power value corresponding to each narrowband light source constituting the initial blue channel, initial red channel, and initial green channel when the target object is imaged at all brightness levels of the preset standard light; based on the initial... The red channel and the initial green channel determine the red light power coefficient and red light power value corresponding to red light sources, and the green light power coefficient and green light power value corresponding to green light sources. In the light power coefficient table, based on the red light power coefficient, the red light power value, the green light power coefficient, and the green light power value, the standard blue light power value of the blue light source corresponding to the current white light level is found. The actual blue light power value in the initial imaging image is extracted. When the actual blue light power value does not match the standard blue light power value, it is confirmed that the current light power has changed, and the initial imaging image is used as the original acquired image.

[0017] According to a second aspect, embodiments of this application provide an endoscope, including an illumination unit, an imaging unit, and an image processing unit. The illumination unit includes at least one first narrowband light source corresponding to at least one first band in an RGB image acquired by the imaging unit, and multiple second light sources corresponding to multiple second bands in a plurality of second bands in an RGB image acquired by the imaging unit. The first narrowband light source and the second light sources, at corresponding standard optical power, can synthesize a preset standard light. Under the illumination of the preset standard light, the imaging unit can acquire a standard image. The image processing unit includes a memory for a computer program and a processor for implementing the steps of the endoscopic image imaging method as described in any of the first aspects above when executing the computer program stored in the memory.

[0018] The endoscopic image imaging method provided by this invention acquires the original image of the target object captured by the imaging unit when the light power of the first narrowband light source changes; decomposes the standard first color channel of the imaging unit corresponding to the standard light power in the original first color channel; fuses the standard first color channel, the original second color channel, and the original third color channel to obtain a corrected image; extracts the brightness information representing brightness and detail from the original acquired image and the color information from the standard image, and fuses the brightness information and the color information to obtain an endoscopic imaging image. This invention utilizes the different reflection characteristics of the first narrowband light source at different parts of the human body cavity, and by selectively increasing or decreasing the proportion of the first narrowband light source at different parts, thereby causing changes in the light power of the original first color channel, it retains the enhanced detail information brought about by the change in the light power of the first narrowband light source. Furthermore, it adjusts the hue of the original acquired image of the target object captured by the imaging unit to adapt to a preset standard light, so that the adjusted endoscopic imaging image retains the detail enhancement brought about by the change in the proportion of the first narrowband light source, while maintaining the color expression of the preset standard light imaging. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 is a schematic flowchart of an endoscopic image imaging method according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] According to an embodiment of the present invention, an embodiment of an endoscopic image imaging method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0023] This embodiment provides an endoscopic image imaging method applicable to an endoscope system. In one embodiment, the endoscope system includes an illumination unit, an imaging unit, and an image processing unit. The illumination unit includes at least one first narrowband light source corresponding to at least one first band in the blue channel of the RGB image acquired by the imaging unit, and multiple second light sources corresponding to multiple second narrowband bands in the red and green channels. The first narrowband light source and the second light sources can synthesize a preset standard light at their corresponding standard optical power. Under the illumination of the preset standard light, the imaging unit can acquire a standard image.

[0024] In one embodiment, the illumination unit provides an illumination module for the endoscope system, and the illumination in the target application scenario is provided only by the illumination unit; the imaging unit is used to receive the initial imaging image obtained by the reflection of the outgoing light provided by the illumination module after it illuminates the target object; wherein, the initial imaging image consists of a blue channel, a red channel and a green channel.

[0025] In one embodiment, the illumination unit is composed of multiple light sources providing different narrowband colors, including at least one first narrowband light source corresponding to at least one first band of the blue channel and multiple second light sources corresponding to multiple second narrowband bands of the red and green channels.

[0026] In this embodiment, the first band includes a narrowband with a center wavelength of 415±10nm and / or a narrowband with a center wavelength of 460±10nm. Correspondingly, the first narrowband light source includes a narrowband blue-violet light source with a center wavelength of 415±10nm and / or a narrowband blue light source with a center wavelength of 460±10nm. It should be noted that the first band can also be other narrowband bands, which can be determined according to the specific application scenario. In the following embodiment, the first band is illustrated using narrowband bands with a center wavelength of 415±10nm and narrowband bands with a center wavelength of 460±10nm as examples.

[0027] In one embodiment, the second band may include a broadband green light source with a center wavelength of 540±40nm, a broadband amber light source with a center wavelength of 600±30nm, and a narrowband red light source with a center wavelength of 630±10nm.

[0028] In this embodiment, the illumination module uses 5 LED light sources. The emitted light output by the illumination module is synthesized from a narrowband blue light source, a narrowband blue-violet light source, a broadband green light source, a broadband amber light source, and a narrowband red light source. The center wavelength range of the narrowband blue-violet light source is 415±10nm, the center wavelength range of the narrowband blue light source is 460±10nm, the center wavelength of the broadband green light source is 540±40nm, the center wavelength of the broadband amber light source is 600±30nm, and the center wavelength of the narrowband red light source is 630±10nm. The blue channel of the initial imaging image received by the imaging unit is synthesized from the narrowband blue light source and the narrowband blue-violet light source. Taking this as an example, the technical solution of the present invention is described.

[0029] For example, FIG1 is a flowchart of an endoscopic image imaging method according to an embodiment of the present invention. As shown in FIG1, the process includes the following steps:

[0030] Step S101: When the optical power of the first narrowband light source changes, acquire the original image of the target object captured by the imaging unit at the current optical power. As mentioned above, since different parts of the human body cavity have different absorption and reflection characteristics of blue light and blue-violet light, the proportion of blue light and blue-violet light is increased or decreased for different parts. When the proportion of blue light and blue-violet light changes, the optical power of the original blue channel will change. At this time, it is necessary to adjust the original image of the target object captured by the imaging unit to adapt to the white light tone while preserving the enhanced detail information brought about by the change in optical power. This ensures that the adjusted endoscopic imaging image retains the detail enhancement brought about by the change in the narrowband light source ratio, while maintaining the color expression of white light imaging.

[0031] The original first color channel is illustrated using the original blue channel as an example, the original second color channel using the original red channel as an example, and the original third color channel using the original green channel as an example. In other embodiments, the first, second, and third colors can be replaced to suit different target object absorption and reflection characteristics.

[0032] For example, for a lighting module composed of 5 LEDs, the optimal color rendering index (CRI) and other key indicators of the light source are usually obtained by modulating the optimal ratio of each light source. During the modulation process, the light power parameters of each LED in the 5 LEDs at a certain white light brightness level are determined, and the light power parameter table of each LED corresponding to each white light level of the lighting module can be determined according to the modulation process.

[0033] For an initial image received by an imaging unit, when the proportion of blue light or blue-violet light changes, its actual optical power changes relative to the preset optical power of the narrowband blue light source and the narrowband blue-violet light source at the current white light brightness level. However, the actual optical power of other light sources providing different colors, excluding the narrowband blue light source and the narrowband blue-violet light source, remains unchanged relative to the preset optical power of the current white light brightness level. Therefore, in one embodiment, the white light brightness level of the current white light illumination can be obtained by matching the first actual optical power of other light sources providing different colors, excluding the narrowband blue light source and the narrowband blue-violet light source, with the first preset optical power. Then, the preset optical power of the narrowband blue light source and the narrowband blue-violet light source corresponding to the current white light brightness level can be looked up in the optical power parameter table using the white light brightness level. When the actual optical power of the narrowband blue light source and the narrowband blue-violet light source does not match the preset optical power, it is confirmed that a change in the optical power of the blue channel has occurred due to the targeted increase or decrease of the proportion of blue light and blue-violet light in different parts.

[0034] In one embodiment, the optical power parameter table may include level information, the optical power value of each LED at each level, and the optical power coefficient.

[0035] Step S102: Decompose the standard first color channel of the imaging unit corresponding to the standard optical power from the original first color channel.

[0036] When the narrowband blue light source and the narrowband blue-violet light source change, there are differences in color between narrowband imaging and white light imaging. In order to recover the color information of white light imaging through the original blue channel, in this embodiment, the standard blue channel of the target object relative to the standard light power is determined based on the original blue channel.

[0037] Since the original blue channel is synthesized from a narrowband blue light source and a narrowband blue-violet light source, in this embodiment, the standard blue channel of the target object relative to the standard optical power can be determined based on the changes of the narrowband blue light source and the narrowband blue-violet light source, respectively.

[0038] In one embodiment, the original blue channel can be decomposed to obtain the original sub-blue channels corresponding to the narrowband blue light source and the narrowband blue-violet light source that constitute the original blue channel. The original sub-blue channels are then restored to obtain the standard sub-blue channel, and the standard sub-blue channels are further fused to obtain the standard blue channel.

[0039] One possible implementation is to decompose the original blue channel using deep learning or spectral reconstruction.

[0040] As a possible implementation, the original sub-blue channel can be restored through deep learning or image processing based on Retinex theory to obtain the standard sub-blue channel.

[0041] As a possible implementation, the standard sub-blue channel can be fused using deep learning or proportional fusion to obtain the standard blue channel.

[0042] Step S103: Fuse the standard first color channel, the original second color channel, and the original third color channel to obtain the corrected image.

[0043] In this embodiment, narrowband imaging is achieved by adjusting the narrowband blue light source and narrowband blue-violet light source corresponding to the blue channel. The light power and ratio of other light sources corresponding to the original red channel and the original green channel usually do not change during narrowband imaging. That is, the light power and ratio of other light sources corresponding to the original red channel and the original green channel are maintained during white light imaging. For the standard blue channel obtained from the original blue channel, it is compared with the original red channel and the original green channel to obtain the color correction image of the target object at the predetermined brightness level of white light with standard light power. The color correction image contains color information that can characterize the target object at the predetermined brightness level of white light during white light imaging. The color information is used to characterize the white light hue of the target object during white light imaging.

[0044] Step S104: Extract the brightness information representing brightness and detail from the original acquired image and the color information from the standard image, and fuse the brightness information and the color information to obtain the endoscopic imaging image.

[0045] In order to ensure that the final endoscopic imaging image includes both the detail enhancement brought by narrowband imaging and the color correction brought by white light imaging, in this embodiment, the brightness information in the original acquired image and the color information in the color-corrected image are fused to obtain the endoscopic imaging image.

[0046] In one embodiment, the original acquired image and the color-corrected image can be converted to independent color spaces, and the brightness information in the brightness channel and the color information in the color channel can be further extracted. Finally, the brightness information and color information are fused to obtain the endoscopic imaging image.

[0047] The endoscopic image imaging method provided by this invention acquires the original image of the target object captured by the imaging unit when the light power of the original blue channel changes. The original image consists of an original blue channel, an original red channel, and an original green channel. Based on the original blue channel, a standard blue channel for the target object relative to a standard light power is determined. The standard blue channel, the original red channel, and the original green channel are fused to obtain a color-corrected image of the target object at the standard light power. The brightness information in the original image and the color information in the color-corrected image are fused to obtain an endoscopic imaging image. This invention, when the light power of the original blue channel changes, restores the color-corrected image of the endoscope at a predetermined brightness level under white light based on the original blue channel. Furthermore, it fuses the original image containing enhanced details and the color-corrected image containing color to obtain an endoscopic imaging image, restoring the white light tone while retaining enhanced details.

[0048] The technical solutions described in the above embodiments are based on narrowband imaging when changing narrowband blue light sources and narrowband blue-violet light sources. It should be understood that for solutions such as changing the blue light source or blue-violet light source alone, or changing other single light sources or all other light sources contained in the illumination module alone or together to achieve other types of imaging solutions, the original channels can be restored to obtain their corresponding standard channels by changing the blue channel, red channel and / or green channel caused by the change of the light source. Furthermore, the information contained in the standard channel is fused with the information contained in the original channel to restore the image corresponding to white light imaging. Finally, by fusing the original image containing the additional beneficial effects brought about by the change of the light source with the color correction image that can restore the endoscope at a predetermined brightness level of white light, the additional beneficial effects brought about by the change of the light source are preserved while restoring the white light tone.

[0049] As an exemplary embodiment, when the optical power of the original blue channel changes, determining the standard blue channel of the target object relative to the standard optical power based on the original blue channel includes: decomposing the original blue channel to obtain original sub-blue channels constituting the original blue channel; wherein the original sub-blue channels are respectively provided by a narrowband blue-violet light source with a center wavelength of 415±10nm and a narrowband blue light source with a center wavelength of 460±10nm; decomposing each of the original sub-blue channels to obtain a standard sub-blue channel corresponding to each of the original sub-blue channels; and fusing each of the standard sub-blue channels to obtain the standard blue channel.

[0050] Since the original blue channel is synthesized from a narrowband blue light source and a narrowband blue-violet light source, in this embodiment, the original blue channel is decomposed to obtain the original sub-blue channels that constitute the original blue channel. Each of the original sub-blue channels is further decomposed to obtain the standard sub-blue channels corresponding to each of the original sub-blue channels. Finally, the standard blue channel is obtained by fusing the standard sub-blue channels.

[0051] As described above, the original blue channel can be decomposed using deep learning. Based on this, as an exemplary embodiment, the decomposition of the original blue channel to obtain the original sub-blue channels constituting the original blue channel includes: inputting the original blue channel into an original sub-blue channel decomposition model to obtain the original sub-blue channels constituting the original blue channel; wherein, the original sub-blue channel decomposition model is based on D... train1 (i,j), D GT1 (i) and D GT1 The dataset consisting of (j) learns to incorporate D during model training. train1 (i,j) is mapped to D GT1 (i) or D GT1 The mapping relationship obtained at (j) is described in D; train1 (i,j) is obtained by simultaneously enabling only the narrowband blue-violet light source and the narrowband blue light source, and acquiring the first preset blue channel image data corresponding to each combination of the narrowband blue-violet light source and the narrowband blue light source; where, D train1 (i,j) represents the preset blue channel image data corresponding to the ith narrowband blue-violet light source brightness level and the jth narrowband blue light source brightness level; the D GT1 (i) Obtained by individually enabling the narrowband blue-violet light source and acquiring the second preset blue channel data at all adjustable levels of the narrowband blue-violet light source; the D GT1 (j) The third preset blue channel data is obtained by enabling the narrowband blue light source individually and collecting data from all adjustable levels of the narrowband blue light source.

[0052] In this embodiment, the original sub-blue channel decomposition model is used to decompose the original blue channel to obtain the original sub-blue channels that constitute the original blue channel. The original sub-blue channel decomposition model can be a deep learning model, specifically a neural network (NN) model; the deep learning model consists of D... train1 (i,j), D GT1 (i) and D GT1 The dataset (j) is used for training.

[0053] For example, the technical solution of this embodiment is illustrated by taking the training process to obtain the first original sub-blue channel corresponding to the narrowband blue-violet light source through original blue channel decomposition as an example:

[0054] In the target scene, only narrowband blue-violet light source and narrowband blue light source are enabled. The adjustable levels of the two light sources are traversed, and the corresponding blue channel image data under each level combination is collected as the training data, denoted as D. train1 (i,j) represents the composite blue channel image corresponding to the ith narrowband blue-violet light source brightness level and the jth narrowband blue light source brightness level.

[0055] Enable a narrowband blue-violet light source individually, iterate through the adjustable levels of the light source, and collect the corresponding blue channel image data at each level as the Ground Truth of the training data, denoted as D. GT (i) represents the blue channel image corresponding to the brightness level of the i-th narrowband blue-violet light source.

[0056] Furthermore, D train1 (i,j) and D GT1 (i) As image pairs, they are organized into a training dataset; using deep learning, the training data is input into a pre-built neural network for model training. During the model training process, the model learns to apply each D... train (i,j) is mapped to D GT1 (i) until the model converges.

[0057] It should be understood that this can be achieved by using D GT1 (i) Replace with D GT1 The implementation of (j) involves training the original sub-blue channel decomposition model to obtain the second original sub-blue channel corresponding to the narrowband blue light source through the original blue channel decomposition, which will not be elaborated here.

[0058] As described above, the original blue channel can also be decomposed through spectral reconstruction. Based on this, as an exemplary embodiment, the decomposition of the original blue channel to obtain the original sub-blue channels constituting the original blue channel further includes: obtaining the single-component estimation matrix corresponding to each of the original sub-blue channels; wherein the single-component estimation matrix is ​​determined by at least one of the R matrix method, principal component analysis method, Wiener estimation method, and comparative measurement method; and multiplying the original acquired image by each of the single-component estimation matrices to obtain each of the original sub-blue channels.

[0059] In this embodiment, for the original acquired image, let Q1 be a single-component estimation matrix related to the system spectral response characteristics of the narrowband blue-violet light source, Q1=[q 11 ,q 21 ,q 31Q2 is a single-component estimation matrix related to the spectral response characteristics of the narrowband blue light source. Q1 and Q2 are determined by at least one of the following methods: R matrix method, principal component analysis method, Wiener estimation method, and comparative measurement method. The original blue channel can then be decomposed using the following formula to obtain the original sub-blue channels that constitute the original blue channel:

[0060] In the formula, B cur1 B represents the first original sub-blue channel corresponding to the narrowband blue-violet light source. cur2 Q1 represents the second original sub-blue channel corresponding to the narrowband blue light source, Q2 represents the single-component estimation matrix related to the system spectral response characteristics of the narrowband blue-violet light source, R represents the original red channel, G represents the original green channel, and B1 represents the original blue channel.

[0061] Since both deep learning models and spectral reconstruction-based methods have their own errors when performing blue channel decomposition, in one embodiment, the original blue channel can be decomposed simultaneously using both deep learning-based and spectral reconstruction-based methods. Specifically, after obtaining the first original blue channel decomposition result based on the deep learning method and the second original blue channel decomposition result based on the spectral reconstruction method, the first error of the original sub-blue channel decomposition model and the second error based on the spectral reconstruction method are obtained respectively. Furthermore, the first and second original blue channel decomposition results are fused together to obtain the corrected original channel decomposition result, thereby achieving the decomposition of the original blue channel. As a possible implementation, the respective fusion weights are inversely proportional to their errors.

[0062] As described above, the original sub-blue channel can be restored using deep learning methods to obtain the standard sub-blue channel. Based on this, as an exemplary embodiment, each of the original sub-blue channels is decomposed to obtain the corresponding standard sub-blue channel, including: inputting each of the original sub-blue channels into a standard sub-blue channel decomposition model to obtain the standard sub-blue channel corresponding to the original sub-blue channel; wherein, the standard sub-blue channel decomposition model is based on D... train2 (i,j), D GT2 (i) The dataset was trained using the dataset; the D train2(i,j) is obtained by acquiring blue channel data at each white light brightness level while keeping other light source levels unchanged, only changing the original sub-blue channel corresponding to the narrowband blue-violet light source or the narrowband blue light source, traversing all adjustable levels of the white light brightness level and the narrowband blue-violet light source or the narrowband blue light source; representing the preset standard sub-blue channel image corresponding to the i-th white light brightness level and the j-th level of the narrowband blue-violet light source or the narrowband blue light source; the D GT2 (i) is obtained by collecting standard sub-blue channel data where the optical power remains unchanged for each white light brightness level, representing the preset sub-standard channel image corresponding to the i-th white light brightness level; during the training of the standard sub-blue channel decomposition model, it learns to decompose the D corresponding to the i-th white light brightness level. train2 (i,j) is mapped to D GT2 The mapping relationship is obtained when (i).

[0063] In this embodiment, the standard sub-blue channel decomposition model is used to decompose the original sub-blue channel to obtain the standard sub-blue channel corresponding to the original sub-blue channel. The standard sub-blue channel decomposition model can be a deep learning model, specifically a neural network (NN) model; the deep learning model consists of D... train2 (i,j), D GT2 (i) is the dataset used for training.

[0064] For example, the technical solution of this embodiment is illustrated by taking the training of a standard sub-blue channel decomposition model to obtain the first standard sub-blue channel through the decomposition of the first original blue sub-channel corresponding to a narrow blue-violet light source as an example:

[0065] In the target scene, images corresponding to each white light brightness level are acquired, and the blue channel image data is extracted as the Ground Truth for training data, denoted as D. GT2 (i) represents the standard blue channel target image corresponding to the i-th white light brightness level.

[0066] Keeping other light source intensities constant, only changing the intensity of the narrowband blue-violet light source corresponding to the current brightness level, we iterate through the adjustable levels of this light source to obtain a training dataset for this light source at the current brightness level, denoted as D. train2 (i,j) represents the blue channel image corresponding to the j-th level of the narrowband blue-violet light source corresponding to the i-th white light brightness level.

[0067] D train2 (i,j) and D GT2(i) As image pairs, these are organized into a training dataset. Using deep learning, the training data is input into a pre-built neural network for model training. During model training, the model learns to apply each D... train2 (i,j) is mapped to D GT2 (i) until the model converges to obtain the standard sub-blue channel decomposition model.

[0068] It should be understood that the standard sub-blue channel decomposition model can be obtained by keeping the intensity of other light sources constant and only changing the intensity of the narrow-band blue light corresponding to the current brightness level, traversing the adjustable levels of the light source, obtaining a set of training datasets for the light source at the current brightness level, and further training the model. The standard sub-blue channel decomposition model is used to obtain the second standard sub-blue channel by decomposing the second original blue sub-channel corresponding to the narrow blue light source.

[0069] As mentioned above, the original sub-blue channels can also be restored through image processing based on Retinex theory. Therefore, as an exemplary embodiment, the decomposition of each of the original sub-blue channels to obtain the standard sub-blue channels corresponding to each original sub-blue channel includes:

[0070] The original acquired image is decomposed into an original illumination image and an original reflectance image;

[0071] Obtain the nonlinear mapping function corresponding to each of the original sub-blue channels; wherein, the nonlinear mapping function is used to represent the correspondence between the illumination component of the light source and the light source intensity corresponding to the original sub-blue channel;

[0072] The standard sub-blue channel is calculated using the following formula: B dst =B cur *[f -1 (f(L cur )*W dst / W cur )] / L cur

[0073] Among them, B dst Let L represent the original blue channel, f represent the nonlinear mapping function of the original sub-blue channel, and L represent the original blue channel. cur W represents the original reflection image. dst1 This indicates the standard optical power of the light source corresponding to the original sub-blue channel at the current white light brightness level.

[0074] For example, the technical solution of this embodiment is described by taking the training of obtaining the first standard sub-blue channel by decomposing the first original blue sub-channel corresponding to a narrow blue-violet light source as an example:

[0075] Based on Retinex theory, the image I(x,y) is decomposed into an illumination image and a reflectance image, which can be expressed as I(x,y)=R(x,y)*L(x,y)

[0076] In the formula, I(x,y) represents the image signal observed or received by the camera, L(x,y) represents the illumination component of the ambient light, and R(x,y) represents the reflection component of the target object carrying image detail information.

[0077] Let B cur1 R is the first original blue sub-channel corresponding to the narrowband blue-violet light source. cur1 The reflection diagram corresponding to a narrowband blue-violet light source, L cur1 R represents the illumination component of a narrow-band blue-violet light source. Since the reflection component is an inherent property of the object and is independent of the lighting and equipment, R... cur1 =R dst1 Among them, R dst1 This is the first standard sub-blue channel corresponding to a narrowband blue-violet light source.

[0078] For example, let W cur1 The current optical power of the narrowband blue-violet light source, W dst1 Let L be the standard optical power corresponding to the narrowband blue-violet light source at the current white light brightness level. The illumination component is positively correlated with the light source intensity. Let f() be the nonlinear mapping function corresponding to the narrowband blue-violet light source. dst1 The relationship between the luminance components mapped to device-independent spaces such as XYZ space and their corresponding values ​​can be expressed as: f(L dst1 ) / W dst1 ≈f(L cur1 ) / W cur1

[0079] Then B dst1 =B cur1 *[f -1 (f(L cur1 )*W dst1 / W cur1 )] / L cur1

[0080] Among them, B dst1 Let L represent the original blue channel, f represent the nonlinear mapping function of the original sub-blue channel, and L represent the original blue channel. cur W represents the original reflection image. dst1 This indicates the standard optical power of the light source corresponding to the original sub-blue channel at the current white light brightness level.

[0081] It should be understood that the second standard sub-blue channel can be further obtained by decomposing the second original blue sub-channel corresponding to the narrow blue light source using the above method, which will not be elaborated here.

[0082] Since both deep learning models and Retinex-based image processing methods have their own errors when restoring the original sub-blue channel, in one embodiment, the original blue channel can be decomposed simultaneously using both deep learning-based and Retinex-based image processing methods. Specifically, after obtaining the first standard blue channel decomposition result based on the deep learning method and the second standard blue channel decomposition result based on the Retinex-based image processing method, the first error of the standard sub-blue channel decomposition model and the second error of the Retinex-based image processing method are obtained respectively. The first and second standard blue channel decomposition results are then fused together to obtain the corrected standard channel decomposition result. In a possible implementation, the fusion weights are inversely proportional to their respective errors.

[0083] As an exemplary embodiment, fusing the brightness information in the original acquired image and the color information in the color-corrected image to obtain an endoscopic imaging image includes: mapping the original acquired image and the color-corrected image to independent color spaces to obtain a first mapped image corresponding to the original acquired image and a second mapped image corresponding to the color-corrected image, respectively; wherein, the independent color space includes at least one of YUV space, Lab space, and HSV space; extracting a first channel containing brightness information from the first mapped image; extracting a second channel containing color information and a third channel containing saturation information from the second mapped image; and fusing the first channel, the second channel, and the third channel to obtain the endoscopic imaging image.

[0084] In this embodiment, by extracting the first channel containing brightness information from the first mapped image and the second channel containing color information and the third channel containing saturation information from the second mapped image, an endoscopic imaging image can be obtained by fusing the original acquired image containing enhanced details and the color-corrected image containing color. This can preserve enhanced details while restoring the white light tone.

[0085] In one embodiment, after extracting the second channel containing color information and the third channel containing saturation information from the second mapped image, before fusing the first channel, the second channel and the third channel, the method further includes: processing the third channel containing saturation information using a stretching algorithm to obtain an enhanced third channel with improved contrast; and fusing the first channel, the second channel and the enhanced third channel to obtain a first contrast-enhanced endoscopic imaging image.

[0086] One possible implementation is to use the first channel as a guide graph to perform guided filtering on the third channel to obtain an enhanced third channel.

[0087] In one embodiment, before fusing the first channel, the second channel, and the third channel, the method further includes: using a stretching algorithm to obtain an enhanced first channel, an enhanced second channel, and an enhanced third channel with improved contrast from the first channel, the second channel, and the third channel; and fusing the enhanced first channel, the enhanced second channel, and the enhanced third channel to obtain a second contrast-enhanced endoscopic imaging image.

[0088] As described above, the standard sub-blue channels can be fused using a proportional fusion method to obtain a standard blue channel. Based on this, as an exemplary embodiment, fusing each of the standard sub-blue channels to obtain the standard blue channel includes: obtaining a preset sub-fusion weight for each of the standard sub-blue channels during fusion; wherein, each preset sub-fusion weight is determined by obtaining a synthesized blue channel image at the target power corresponding to each of the standard sub-blue channels and a decomposed blue channel image when the narrowband blue light source and the narrowband blue-violet light source are individually enabled based on the target power corresponding to each of the standard sub-blue channels, and by fusing the decomposed blue channel images into a blue channel image; and fusing based on the preset sub-fusion weights corresponding to each of the standard sub-blue channels to obtain the standard blue channel.

[0089] Specifically, since the power values ​​of each light source corresponding to each white light brightness level are fixed, in the target scene, a composite blue channel image B can be obtained when narrowband blue-violet light sources and narrowband blue light sources are simultaneously enabled at the target power. merge And the blue channel image B when the target power of the narrowband blue-violet light source and the narrowband blue light source are individually enabled. t1 and B t2 It can be done through B merge B t1 and B t2 The fusion ratio coefficients are fitted to obtain the first preset sub-fusion weight and the second standard sub-blue channel during fusion, where β represents the second preset sub-fusion weight of the second standard blue channel during fusion.

[0090] Furthermore, the standard sub-blue channel can be fused using the following formula: B2 = α * B dst1 +β*B dst2

[0091] In the formula, B2 represents the standard blue channel, B dst1 B represents the first standard sub-blue channel, α represents the first preset sub-fusion weight of the first standard sub-blue channel during fusion, and B represents the first standard sub-blue channel.dst2 β represents the second standard sub-blue channel, and β represents the second preset sub-fusion weight of the second standard sub-blue channel during fusion.

[0092] As described above, the standard sub-blue channels can be fused using deep learning to obtain a standard blue channel. Based on this, as an exemplary embodiment, the fusion of the standard sub-blue channels to obtain the standard blue channel further includes:

[0093] Each of the aforementioned standard sub-blue channels is input into the standard sub-blue channel fusion model to obtain the standard blue channel; wherein, the standard sub-blue channel fusion model is based on D... train1 (i), D train2 (i) and D GT3 (i) The dataset was trained using this dataset;

[0094] The D train (i) is obtained by individually enabling the narrow-band blue-violet light source, traversing the adjustable levels corresponding to the white light brightness level where the narrow-band blue-violet light source is located, and collecting the corresponding blue channel image data at each level; represents the first standard sub-blue channel image corresponding to the brightness level of the i-th narrow-band blue-violet light source.

[0095] The D train2 (i) is obtained by individually enabling the narrowband blue light source, traversing the adjustable levels corresponding to the white light brightness level where the narrowband blue light source is located, and collecting the corresponding blue channel image data at each level; representing the second standard sub-blue channel image corresponding to the brightness level of the i-th narrowband blue light source.

[0096] D GT3 (i) is obtained by traversing all adjustable levels of the narrowband blue-violet light source and the narrowband blue light source according to the preset white light brightness level, and collecting the corresponding blue channel data under the preset white light brightness level; represents the composite standard blue channel image corresponding to the brightness level of the i-th group of narrowband blue-violet light source and the brightness level of the narrowband blue light source.

[0097] During the training of the standard sub-blue channel fusion model, the mapping relationship is learned when mapping the first standard sub-blue channel image corresponding to the i-th narrowband blue-violet light source brightness level and the second standard sub-blue channel image corresponding to the i-th narrowband blue light source brightness level to the synthesized standard blue channel image corresponding to the i-th group of narrowband blue-violet light source brightness levels and narrowband blue light source brightness levels.

[0098] To determine whether the current endoscopic system has performed narrowband imaging, as an exemplary embodiment, before acquiring the original image of the target object acquired by the imaging unit, the method further includes: acquiring an initial imaging image of the target object and a table of optical power coefficients corresponding to the target object at each white light brightness level; wherein, the initial imaging image consists of an initial blue channel, an initial red channel, and an initial green channel; the optical power coefficient table is used to represent the power coefficient and power value corresponding to each light source constituting the initial blue channel, initial red channel, and initial green channel when the target object is imaged at each white light brightness level; based on the initial... The red channel and the initial green channel determine the red light power coefficient and red light power value corresponding to red light sources, and the green light power coefficient and green light power value corresponding to green light sources. In the light power coefficient table, based on the red light power coefficient, the red light power value, the green light power coefficient, and the green light power value, the standard blue light power value of the blue light source corresponding to the current white light level is found. The actual blue light power value in the initial imaging image is extracted. When the actual blue light power value does not match the standard blue light power value, it is confirmed that the light power of the original blue channel has changed, and the initial imaging image is used as the original acquired image.

[0099] When performing narrow-band imaging in an endoscopic system, only the power values ​​and power coefficients of the narrow-band blue-violet light source and the narrow-band blue light source change at the current white light level, while the power values ​​and power coefficients of other light sources remain unchanged. Therefore, in this embodiment, the red light power coefficient and red light power value corresponding to the red light source and the green light power coefficient and green light power value corresponding to the green light source are determined based on the initial red channel and the initial green channel. Then, the current white light level can be determined based on the power coefficients and power values ​​corresponding to the initial red channel and the initial green channel.

[0100] Furthermore, based on the white light setting, find the standard blue light power value of the blue light source corresponding to the current white light setting in the light power coefficient table.

[0101] When the actual blue light power value in the initial imaging image does not match the standard light power value, it is confirmed that the light power of the original blue channel has changed. At this time, the initial acquired image needs to be processed. Based on this, the initial acquired image is used as the original acquired image. The original acquired image is then processed by the endoscopic image imaging method in the above embodiment to obtain the endoscopic imaging image, so that the processed image restores the white light tone while retaining enhanced details.

[0102] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0103] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods in the above embodiments.

[0104] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0105] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the solution provided in this embodiment, depending on actual needs.

[0106] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0107] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0108] The above are merely preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. An endoscopic image imaging method, characterized in that, The endoscope includes an illumination unit, an imaging unit, and an image processing unit. The illumination unit includes at least one first narrowband light source corresponding to at least one first band in the blue channel of the RGB image acquired by the imaging unit, and multiple second light sources corresponding to multiple second bands in the red and green channels. The first narrowband light source and the second light sources can synthesize a preset standard light at their corresponding standard light power. Under the illumination of the preset standard light, the imaging unit can acquire a standard image. The endoscopic image imaging method is applicable to the image processing unit, and the endoscopic image imaging method includes: When the optical power of the first narrowband light source changes, the original image of the target object acquired by the imaging unit at the current optical power is obtained; wherein, the original image is composed of the original first color channel, the original second color channel and the original third color channel; The standard first color channel of the imaging unit corresponding to the standard optical power is decomposed from the original first color channel; The standard first color channel, the original second color channel, and the original third color channel are fused to obtain the corrected image; Brightness information representing brightness and detail from the original acquired image and color information from the standard image are extracted respectively, and the brightness information and color information are fused to obtain an endoscopic imaging image.

2. The endoscopic image imaging method as described in claim 1, characterized in that, The step of decomposing the original first color channel into the standard first color channel corresponding to the imaging unit at the standard optical power includes: In the original first color channel, at least one original first color sub-channel of the imaging unit corresponding to the at least one first narrowband light source under the current optical power is decomposed, wherein each original first color sub-channel corresponds to one first narrowband light source; The standard first color sub-channel of the imaging unit corresponding to the standard optical power of the first narrowband light source is decomposed from the original first color sub-channel; The standard first color sub-channel is obtained by fusing the standard first color sub-channel.

3. The endoscopic image imaging method as described in claim 2, characterized in that, The first band includes a narrowband with a center wavelength of 415±10nm and / or a narrowband with a center wavelength of 460±10nm. Correspondingly, the first narrowband light source includes a narrowband blue-violet light source with a center wavelength of 415±10nm and / or a narrowband blue light source with a center wavelength of 460±10nm. The original first color channel includes the original blue channel; the original first color sub-channel includes the original sub-blue channel; the standard first color sub-channel includes the standard sub-blue channel.

4. The endoscopic image imaging method as described in claim 3, characterized in that, The step of decomposing at least one original sub-first color channel from the original first color channel to the imaging unit corresponding to the at least one first narrowband light source at the current optical power includes: The original blue channel is decomposed into the original sub-blue channels of the imaging unit corresponding to the narrowband blue-violet light source and the narrowband blue light source at the current optical power using the spectral reconstruction method. or, The original blue channel is input into a pre-trained original sub-blue channel decomposition model to obtain the original sub-blue channels of the imaging units corresponding to the narrowband blue-violet light source and the narrowband blue light source at the current optical power. The original sub-blue channel decomposition model is based on D train1 (i,j), D GT1 (i) and D GT1 The dataset consisting of (j) learns to incorporate D during model training. train1 (i,j) is mapped to D GT1 (i) or D GT1 The mapping relationship is obtained when (j) is used; The D train1 (i,j) is obtained by simultaneously enabling only the narrowband blue-violet light source and the narrowband blue light source, and acquiring the first preset blue channel image data corresponding to each power level combination of the narrowband blue-violet light source and the narrowband blue light source; where, D train1 (i,j) represents the preset blue channel image data corresponding to the i-th narrowband blue-violet light source power level and the j-th narrowband blue light source power level; The D GT1 (i) The data of the second preset blue channel is obtained by enabling the narrow-band blue-violet light source individually and collecting data of the second preset blue channel at all adjustable power levels of the narrow-band blue-violet light source. The D GT1 (j) The third preset blue channel data is obtained by enabling the narrowband blue light source alone and collecting data at all adjustable power levels of the narrowband blue light source.

5. The endoscopic image imaging method as described in claim 3, characterized in that, The step of decomposing the original first color sub-channel into the standard first color sub-channel of the imaging unit corresponding to the first narrowband light source at standard optical power includes: Using the computational theory of color constancy perception, the standard sub-blue channels of the imaging unit corresponding to the narrowband blue-violet light source and the narrowband blue light source at standard optical power are respectively decomposed from the original sub-blue channel; or, Each of the original sub-blue channels is input into the standard sub-blue channel decomposition model to obtain the standard sub-blue channel corresponding to the original sub-blue channel; wherein, the standard sub-blue channel decomposition model is based on D... train2 (i,j), D GT2 (i) The dataset was trained using this dataset; The D train2 (i,j) is obtained by acquiring blue channel data at each white light brightness level while keeping other light source levels unchanged, and only changing the narrowband blue-violet light source or the narrowband blue light source corresponding to the original sub-blue channel. This is done by traversing all adjustable levels of the white light brightness level and the narrowband blue-violet light source or the narrowband blue light source, and obtaining the blue channel data at each brightness level. It represents the preset standard sub-blue channel image corresponding to the i-th white light brightness level and the j-th level of the narrowband blue-violet light source or the narrowband blue light source. The D GT2 (i) is obtained by collecting standard sub-blue channel data with no change in optical power corresponding to each white light brightness level, representing the preset sub-standard channel image corresponding to the i-th white light brightness level; During the training of the standard sub-blue channel decomposition model, the model learns to decompose the D corresponding to the i-th white light brightness level. train2 (i,j) is mapped to D GT2 The mapping relationship is obtained when (i).

6. The endoscopic image imaging method as described in claim 3, characterized in that, The process of fusing the standard first color sub-channel to obtain the standard first color channel includes: Obtain the preset sub-fusion weights of each of the standard sub-blue channels during fusion; wherein, each of the preset sub-fusion weights is determined by obtaining the synthesized blue channel image under the target power corresponding to each of the standard sub-blue channels and the decomposed blue channel image when the narrowband blue light source and the narrowband blue-violet light source are individually enabled based on the target power corresponding to each of the standard sub-blue channels, and by fusing the decomposed blue channel image into a blue channel image. The standard blue channels are obtained by fusing based on the preset sub-fusion weights corresponding to each of the standard sub-blue channels.

7. The endoscopic image imaging method as described in claim 3, characterized in that, The process of fusing the standard first color sub-channel to obtain the standard first color channel further includes: Each of the aforementioned standard sub-blue channels is input into the standard sub-blue channel fusion model to obtain the standard blue channel; wherein, the standard sub-blue channel fusion model is based on D... train1 (i), D train2 (i) and D GT3 (i) The dataset was trained using this dataset; The D train (i) is obtained by individually enabling the narrow-band blue-violet light source, traversing the adjustable levels corresponding to the white light brightness level where the narrow-band blue-violet light source is located, and collecting the corresponding blue channel image data at each level; represents the first standard sub-blue channel image corresponding to the brightness level of the i-th narrow-band blue-violet light source. The D train2 (i) is obtained by individually enabling the narrowband blue light source, traversing the adjustable levels corresponding to the white light brightness level where the narrowband blue light source is located, and collecting the corresponding blue channel image data at each level; representing the second standard sub-blue channel image corresponding to the brightness level of the i-th narrowband blue light source. D GT3 (i) is obtained by traversing all adjustable levels of the narrowband blue-violet light source and the narrowband blue light source according to the preset white light brightness level, and collecting the corresponding blue channel data under the preset white light brightness level; represents the composite standard blue channel image corresponding to the brightness level of the i-th group of narrowband blue-violet light source and the brightness level of the narrowband blue light source. During the training of the standard sub-blue channel fusion model, the mapping relationship is learned when mapping the first standard sub-blue channel image corresponding to the i-th narrowband blue-violet light source brightness level and the second standard sub-blue channel image corresponding to the i-th narrowband blue light source brightness level to the synthesized standard blue channel image corresponding to the i-th group of narrowband blue-violet light source brightness levels and narrowband blue light source brightness levels.

8. The endoscopic image imaging method as described in claim 1, characterized in that, The process involves extracting brightness information representing brightness and detail from the original acquired image and color information from the standard image, and then fusing the brightness information and color information to obtain an endoscopic imaging image, including: The original acquired image and the corrected image are mapped to independent color spaces to obtain a first mapped image corresponding to the original acquired image and a second mapped image corresponding to the color corrected image, respectively; wherein, the independent color space includes at least one of YUV space, Lab space, and HSV space; Extract the first channel containing brightness information from the first mapped image; Extract the second channel containing color information and the third channel containing saturation information from the second mapped image; The endoscopic imaging image is obtained by fusing the first channel, the second channel, and the third channel.

9. The endoscopic image imaging method as described in claim 1, characterized in that, Before acquiring the original image of the target object captured by the imaging unit at the current optical power, the method further includes: Acquire an initial imaging image of the target object and a table of optical power coefficients corresponding to the target object at all brightness levels of a preset standard light; wherein, the initial imaging image consists of an initial blue channel, an initial red channel, and an initial green channel; the table of optical power coefficients is used to represent the power coefficients and power values ​​of each narrowband light source that constitutes the initial blue channel, the initial red channel, and the initial green channel when the target object is imaged at all brightness levels of the preset standard light; Based on the initial red channel and the initial green channel, determine the red light power coefficient and red light power value corresponding to the red light source, and the green light power coefficient and green light power value corresponding to the green light source; In the optical power coefficient table, based on the red light power coefficient, the red light power value, the green light power coefficient, and the green light power value, find the standard blue light power value of the blue light source corresponding to the current white light level; Extract the actual blue light power value from the initial imaging image; When the actual blue light power value does not match the standard blue light power value, it is confirmed that the current light power has changed, and the initial imaging image is used as the original acquired image.

10. An endoscope, characterized in that, The system includes an illumination unit, an imaging unit, and an image processing unit. The illumination unit includes at least one first narrowband light source corresponding to at least one first band in the blue channel of the RGB image acquired by the imaging unit, and multiple second light sources corresponding to multiple second bands in the red and green channels. The first narrowband light source and the second light sources can synthesize a preset standard light at their respective standard light power. Under the illumination of the preset standard light, the imaging unit can acquire a standard image. The image processing unit includes a memory for computer programs; A processor configured to implement the steps of the endoscopic image imaging method as described in any one of claims 1 to 9 when executing a computer program stored in the memory.

Images