An endoscope image enhancement method and system

Abstract

The application relates to an endoscope image enhancement method and system, and relates to the technical field of image recognition, and comprises the following steps: acquiring a real-time image, instrument motion data and a prior standard organ model with an attached enhancement parameter plan set; fusing the above multiple source information to estimate the current spatial pose of the endoscope; matching the enhancement parameter plan list based on the current pose, and adaptively fusing the list according to the pose fluctuation rate to obtain a current enhancement parameter configuration; and performing enhancement processing on the real-time image according to the configuration to obtain an enhanced image. The application solves the technical problem that a traditional endoscope imaging enhancement mode is fixed, it is difficult to adaptively match the operation depth and the view angle change, and the key part cannot be observed clearly.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to an endoscopic image enhancement method and system. Background Technology

[0002] Endoscopy is an indispensable visualization tool in minimally invasive diagnosis and treatment. Through front-end image sensors, it acquires real-time images of the internal structures of body cavities, providing crucial visual support for lesion observation, tissue identification, and surgical navigation. In endoscopic procedures in fields such as obstetrics and gynecology, the complex anatomical structures of body cavities and the varying imaging requirements for lighting, contrast, and color saturation in different tissue regions directly impact the operator's accuracy in identifying tissue morphology, blood vessel distribution, and lesion boundaries. Currently, image enhancement in endoscopy typically relies on several factory-preset fixed imaging modes, such as standard mode, high-brightness mode, and narrowband imaging mode, which the operator manually switches between during the procedure according to observational needs. Some devices also support manual adjustment of parameters such as brightness, contrast, and sharpness.

[0003] However, this manual adjustment method has significant shortcomings. On the one hand, the operating environment within the body cavity changes rapidly, and manual adjustment distracts the operator's attention, increasing the complexity and time cost of the operation. On the other hand, fixed enhancement modes or experience-based manual adjustments cannot dynamically adapt to the spatial movement and tissue changes of the endoscope within the body cavity in real time. When the lens advances, rotates, or retracts within narrow cavities, the tissue morphology, lighting conditions, and observation requirements continuously change, and the preset parameters cannot proactively predict and match the new tissue area to be entered, resulting in a deviation between the image enhancement effect and the current surgical field state, affecting the continuity of observation and the smoothness of the operation. Summary of the Invention

[0004] This application provides an endoscopic image enhancement method and system, which improves the technical problem in the prior art where image enhancement parameters rely on manual adjustment or preset modes, and cannot dynamically adapt to the movement and position changes of the instrument in real time, resulting in unstable enhancement effects and affecting the smoothness of operation.

[0005] This application discloses the following technical solution:

[0006] In a first aspect, this application provides an endoscopic image enhancement method, the method comprising:

[0007] Acquire real-time images from the endoscope, instrument motion monitoring data, and a priori standard organ model, wherein the priori standard organ model is associated with a set of enhancement parameter plans;

[0008] Based on the real-time images, the instrument motion monitoring data, and the prior standard organ model, the current spatial pose of the endoscope is estimated through multi-source information fusion.

[0009] Based on the current spatial pose and the prior standard organ model, the enhancement parameter plan list corresponding to the enhancement parameter plan set is matched and determined, and the enhancement parameter plan list is adaptively fused based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration;

[0010] Based on the current enhancement parameter configuration, parametric enhancement processing is performed on the real-time image to obtain the enhanced image.

[0011] Secondly, this application provides an endoscopic image enhancement system, the system comprising:

[0012] The data acquisition module is used to acquire real-time images of the endoscope, instrument motion monitoring data, and a priori standard organ models, wherein the priori standard organ models are associated with a set of enhancement parameter plans.

[0013] The pose estimation module is used to estimate the current spatial pose of the endoscope by fusing multi-source information based on the real-time image, the instrument motion monitoring data and the prior standard organ model.

[0014] The parameter matching module is used to match and determine the list of enhancement parameter plans corresponding to the enhancement parameter plan set according to the current spatial pose and the prior standard organ model, and to adaptively fuse the list of enhancement parameter plans based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration.

[0015] The image enhancement module is used to perform parametric enhancement processing on the real-time image according to the current enhancement parameter configuration to obtain the enhanced image.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] The technical solution of this application provides an endoscopic image enhancement method. First, by acquiring real-time images of the endoscope, instrument motion monitoring data, and a priori standard organ model with an associated set of enhancement parameters, it achieves synchronous integration of multi-source information such as intracavitary surgical field images, instrument operation status, and individualized anatomical prior knowledge. This provides a unified data foundation for subsequent spatial positioning and parameter decision-making, and solves the problem that traditional imaging modes rely only on single-frame images and lack prior organ structure and operational background information.

[0018] Furthermore, by performing feature matching based on real-time images, establishing a position transformation matrix by combining individual organ dimensions, and performing multi-source fusion with instrument motion monitoring data to estimate the current spatial pose of the endoscope, accurate perception of the instrument's spatial position within the body cavity and quantitative expression of its motion state are achieved. This solves the problem that traditional enhancement methods cannot perceive the instrument's position and motion trend, leading to a disconnect between the enhancement strategy and the current observation task.

[0019] Furthermore, by matching the current spatial pose with the prior standard organ model, the current tissue region is determined, and then the appropriate plan is selected from the set of enhancement parameter plans. Based on the fluctuation rate of the spatial pose, adaptive fusion is performed to obtain the current enhancement parameter configuration. This realizes the dynamic tracking and forward-looking control of the enhancement strategy on the movement rhythm and perspective changes of the device, and solves the problem that traditional fixed parameters or manual adjustment methods cannot match the changes in tissue region and operational dynamics in real time.

[0020] Finally, by performing parameterized enhancement processing on the real-time image based on the current enhancement parameter configuration, the enhanced image is obtained. This realizes the closed-loop execution of the enhancement strategy from spatial perception to parameter output, providing operators with clear, uniform, and coherent high-quality images in complex body cavity environments.

[0021] In summary, the technical solution of this application realizes the adaptive prospective control of multi-source fusion perception and enhancement parameters of endoscopic spatial pose. Through prior organ model construction and tissue region annotation, joint localization of visual and motion information, conservative discrimination and pre-plan selection fusion, and closed-loop feedback correction of enhancement effect, it effectively improves the technical problem of fixed imaging enhancement mode and difficulty in dynamically adapting to changes in operation scenario in the prior art, which leads to unstable image quality. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A schematic flowchart of an endoscopic image enhancement method provided in an embodiment of this application;

[0024] Figure 2 A flowchart illustrating the adaptive parameter strategy selection method for an endoscopic image enhancement method provided in this application embodiment;

[0025] Figure 3 This is a schematic diagram of the structure of an endoscopic image enhancement system provided in an embodiment of this application.

[0026] In the attached diagram, Figure 3 The components represented by each number are described as follows: data acquisition module 11, pose estimation module 12, parameter matching module 13, and image enhancement module 14. Detailed Implementation

[0027] This application provides an endoscopic image enhancement method and system to solve the technical problem that the existing endoscopic imaging enhancement lacks the ability to anticipate the operation scenario, resulting in the enhancement strategy being disconnected from the current observation task and making it difficult to continuously provide high-quality image support in complex body cavity environments.

[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that the numerical values ​​in the embodiments are for illustrative purposes only and do not constitute a limitation on this application.

[0029] Example 1, as shown in the appendix Figure 1 As shown, this application provides an endoscopic image enhancement method, the method comprising the following steps:

[0030] S100: Acquire real-time images of the endoscope, instrument motion monitoring data, and a priori standard organ model, wherein the priori standard organ model is associated with an enhancement parameter pre-set set;

[0031] In this embodiment, the imaging enhancement of the endoscope within the body cavity requires simultaneous access to real-time image information, instrument operation status, and individualized anatomical knowledge. This step, as the data acquisition and preparation stage of the overall method, necessitates the simultaneous acquisition of real-time endoscopic images, instrument motion monitoring data, and individualized prior standard organ models, providing a unified multi-source data foundation for subsequent spatial localization and parameter decision-making.

[0032] Step S100 in the method provided in this application embodiment includes:

[0033] Based on the physical characteristics profile of the target object, a set of approximate samples matching the features of the physical characteristics profile is selected from a pre-stored historical examination sample database. Statistical analysis is performed on the historical examination data in the approximate sample set to obtain the geometric structure statistics of organ cavities and the distribution statistics of each tissue region. Based on the geometric structure statistics and the tissue region distribution statistics, a three-dimensional organ model containing the geometric shape and spatial dimensions of the inner surface of the organ cavity is constructed through three-dimensional reconstruction, and each tissue region is labeled. According to the approximate sample set, approximate samples with the highest pre-set percentile of enhancement effect scores are extracted, and corresponding enhancement parameter plans are configured for each tissue region in the three-dimensional organ model to form the prior standard organ model. The prior standard organ model is then associated and stored with the physical characteristics profile. Detailed explanation follows:

[0034] In this embodiment, the prior standard organ model refers to an individualized reference model constructed through statistical analysis and 3D reconstruction based on the target object's physical characteristic portrait and approximate historical samples. This model includes the geometric structure of organ cavities, tissue region distribution annotations, and enhancement parameter plans for each region. The physical characteristic portrait refers to a multi-dimensional set of parameters describing the individual characteristics of the target object, including basic physiological indicators such as age and organ size. The approximate sample set refers to a set of historical case data selected from a historical examination sample database that has a high similarity to the current target object's physical characteristic portrait, used for statistical inference of organ morphology and tissue distribution. The enhancement parameter plan refers to a predefined combination of image enhancement parameters for a specific tissue region, including values ​​for parameters such as brightness, contrast, and sharpening intensity. The 3D organ model refers to a 3D digital model reconstructed through geometric structural statistical information and tissue region distribution statistical information, containing the geometric shape and spatial dimensions of the inner surface of cavities.

[0035] In this step, firstly, in order to obtain historical reference data that matches the individual characteristics of the target object, it is necessary to screen a set of approximate samples from the historical examination sample library based on the physical characteristics profile. The physical characteristics profile may include key parameters such as age and organ size.

[0036] Furthermore, in order to construct a personalized organ reference model, it is necessary to perform statistical analysis on the historical examination data in the approximate sample set, extract the geometric structure statistics of organ cavities and the distribution statistics of each tissue region, generate a three-dimensional organ model through three-dimensional reconstruction and label each tissue region.

[0037] Furthermore, in order to associate the enhanced experience with the anatomical location, it is necessary to extract the samples with the highest pre-defined percentile scores of the enhancement effect from the approximate samples, and configure the corresponding enhancement parameter plans for each tissue region in the three-dimensional organ model to form a priori standard organ model.

[0038] Finally, the prior standard organ model is associated with the physical characteristic portrait and stored so that the corresponding prior standard organ model can be directly retrieved based on the physical characteristic portrait of the target object in the future.

[0039] Step S100 in the method provided in this application embodiment further includes:

[0040] Real-time images are acquired at a predetermined frame rate using an image sensor at the endoscope's tip. These images are then pre-processed for noise reduction and normalization before being stored in the tissue. Instrument motion monitoring data, including linear displacement, bending angle, and rotation angle data, is collected using a motion sensor. Based on the target object's vital sign profile, a corresponding prior standard organ model is matched and retrieved. A detailed explanation follows:

[0041] In this embodiment, real-time images refer to the raw image frames continuously acquired by the endoscope's front-end image sensor within the body cavity at a predetermined frame rate, which are then used for subsequent spatial pose estimation after noise reduction and normalization preprocessing. Instrument motion monitoring data refers to instrument motion state parameters acquired by motion sensors deployed within the endoscope's operating handle or instrument channel, including linear displacement data reflecting the distance the instrument advances or retreats, bending angle data reflecting the degree of bending of the instrument's front end, and rotation angle data reflecting the rotation angle of the instrument around its own axis. Noise reduction and normalization preprocessing refer to applying Gaussian filtering or median filtering to the raw images to remove sensor noise and mapping pixel grayscale values ​​to a uniform range to enhance the stability of subsequent processing.

[0042] In this step, firstly, in order to obtain real-time images of the body cavity, real-time images need to be acquired at a predetermined frame rate through the image sensor at the front end of the endoscope, and then subjected to noise reduction preprocessing and normalization processing.

[0043] Furthermore, in order to obtain the operator's control status of the endoscope, it is necessary to collect instrument motion monitoring data such as linear displacement, bending angle and rotation angle through instrument motion sensors.

[0044] Finally, in order to obtain individualized anatomical reference benchmarks, it is necessary to match and retrieve the corresponding prior standard organ models based on the physical characteristics profile of the target subject.

[0045] In summary, this step, by combining the physical characteristics of the target subject to construct an individualized prior standard organ model and simultaneously collecting real-time images and instrument motion monitoring data, achieves the organic integration of preoperative anatomical priors and intraoperative dynamic information, providing an individualized and structured reference benchmark for subsequent pose estimation and enhanced parameter matching.

[0046] S200: Based on the real-time image, the instrument motion monitoring data, and the prior standard organ model, the current spatial pose of the endoscope is estimated through multi-source information fusion;

[0047] In this embodiment, after obtaining real-time images, instrument motion monitoring data, and a priori standard organ models, it is necessary to determine the current spatial position and orientation of the endoscope within the body cavity. Simply relying on image feature matching is prone to failure in areas with weak texture, while relying solely on instrument motion data suffers from cumulative errors, and general positioning lacks adaptation to the anatomical scale of individual organs. This step achieves reliable spatial pose estimation by weighted fusion of visual feature localization and instrument motion estimation, and by uniformly mapping to anatomical space using an individual organ size transformation matrix.

[0048] Step S200 in the method provided in this application embodiment includes:

[0049] The examination image data corresponding to the real-time image is acquired, and the individual organ size parameters of the target object are extracted based on the examination image data; an individual position transformation matrix is ​​established based on the proportional relationship between the individual organ size parameters and the geometric structure statistical information in the prior standard organ model; frame feature matching and tracking are performed based on the real-time image, and the perspective projection relationship is solved in combination with the individual position transformation matrix to determine the first estimated position of the endoscope in the model space of the prior standard organ model.

[0050] Based on the instrument motion monitoring data and the preset instrument motion model, combined with the individual position transformation matrix, the second estimated position of the endoscope is determined; the first estimated position and the second estimated position are weighted and fused to obtain the current spatial pose estimation result, and the matching confidence index and pose estimation error index are output simultaneously; the current spatial pose estimation result is constrained and verified, and the current spatial pose is output after passing the constraint verification, wherein the constraint verification includes at least geometric constraint verification, tissue region constraint verification, and color feature constraint verification. Detailed explanation follows:

[0051] In this embodiment, the individual position transformation matrix refers to a coordinate transformation matrix established based on the proportional relationship between the size of the target object's individual organ and the geometric structure statistical information in the prior standard organ model. This matrix is ​​used to map the endoscope's position in the actual organ to a unified model space. The first estimated position refers to the spatial position of the endoscope in the model space calculated by combining the perspective projection relationship with the individual position transformation matrix after performing inter-frame feature matching and tracking based on real-time images. The second estimated position refers to the position estimate obtained by recursively deducing the endoscope's displacement and attitude changes through a preset instrument motion model based on instrument motion monitoring data, and then mapping it to the model space using the individual position transformation matrix.

[0052] The matching confidence index quantifies the reliability of the first estimated position and can be calculated based on the proportion of inliers in feature matching. The pose estimation error index measures the uncertainty of the current pose estimation result and can be calculated based on reprojection error. Constraint verification refers to the process of using anatomical boundary information from a priori standard organ model to screen the reasonableness of the fusion estimation results. This includes checking whether the estimated pose is within the geometric range of the organ cavity, whether the tissue type of the region is consistent with the annotation, and whether the color features of the region match the expected tissue color.

[0053] In this step, firstly, in order to establish a correspondence between the actual organ size of an individual and the model, it is necessary to extract the individual organ size parameters from the test image data, calculate the proportional relationship between them and the geometric statistical information of the prior model, and establish an individual position transformation matrix.

[0054] Furthermore, to locate the endoscope visually, inter-frame feature matching and tracking based on real-time images are required. This, combined with the individual position transformation matrix, solves the perspective projection relationship to determine the initial estimated position. Visual positioning accuracy is high when the image texture is rich, but it may fail in areas with weak texture due to insufficient feature points.

[0055] Furthermore, to supplement positioning information from a motion perspective, a second estimated position needs to be calculated based on the device's motion monitoring data and a pre-set device motion model, combined with an individual position transformation matrix. Motion estimation has high short-term accuracy, but cumulative drift occurs over long-term operation.

[0056] Furthermore, to integrate the advantages of the two positioning sources, a weighted fusion of the first and second estimated positions is performed using Kalman filtering. Kalman filtering uses the machine motion model as the state prediction equation and the visual positioning result as the observation update equation. In the prediction step, it calculates the prior estimate based on the pose of the previous frame and the current motion data. In the update step, it combines the visual positioning result and the observation noise covariance to calculate the Kalman gain, correcting the prior estimate to obtain the posterior estimate.

[0057] When the matching confidence index is high, the observation noise covariance is small, the Kalman gain increases, and the fusion result is more biased towards visual localization. When the matching confidence index is low, the observation noise covariance is large, the Kalman gain decreases, and the fusion result relies more on motion model prediction, thus adaptively balancing the contributions of the two localization sources. After fusion, the matching confidence index and pose estimation error index are output simultaneously.

[0058] Finally, the estimation results are constrained and validated. Geometric constraints are used to check whether the endoscope position is within the geometric boundaries of the organ cavity; tissue region constraints are used to check whether the tissue type of the current region is consistent with the prior model annotation; and color feature constraints are used to check whether the color distribution of the current image region matches the expected tissue color. If validation is successful, the output is the current spatial pose; if validation fails, the estimation can be adjusted appropriately and marked as a low-confidence state.

[0059] For example, consider the pose estimation of three consecutive frames during an endoscopic procedure. Let the fused pose of the (k-1)th frame be the coordinates (12.5, 8.3, 45.2) in model space, in mm. The instrument motion monitoring data of the kth frame shows a linear displacement of +2.1 mm, a bending angle change of +3.2°, and a rotation angle change of -1.5°. After recursion using the instrument motion model and mapping with the individual position transformation matrix, the second estimated position is obtained as (14.3, 8.9, 46.5).

[0060] Simultaneously, the real-time image of the k-th frame, after feature matching and tracking, and solving the perspective projection relationship, yields the first estimated position as (14.8, 9.4, 47.1), with a feature matching inlier ratio of 0.82, i.e., a matching confidence level of 0.82. Kalman filtering uses the second estimated position as the state prediction and the first estimated position as the observation update. Taking the process noise covariance Q = 0.5I and the observation noise covariance R = (1 / 0.82) × I ≈ 1.22I, the fused pose is obtained after calculating the Kalman gain K, with a reprojection error of 1.3mm. After passing geometric constraint verification and tissue region constraint verification, the current spatial pose of this frame is output as (14.6, 9.2, 46.8), with a matching confidence level of 0.82 and a pose estimation error of 1.3mm.

[0061] In summary, this step utilizes a weighted fusion of visual feature localization and instrument motion estimation. It eliminates cumulative motion errors by using the absolute position reference provided by vision, compensates for visual failures in areas with weak texture by using short-term smooth trajectories provided by motion, and maps the localization to anatomical space using an individualized transformation matrix, giving the localization results anatomical semantics. The simultaneously output confidence and error indices provide a reliable reference for subsequent enhancement parameter selection, thus solving the problem of insufficient reliability of single-source localization in complex body cavity environments.

[0062] S300: Based on the current spatial pose and the prior standard organ model, match and determine the list of enhancement parameter plans corresponding to the enhancement parameter plan set, and adaptively fuse the list of enhancement parameter plans based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration;

[0063] In this embodiment, after obtaining the current spatial pose, its confidence level, and error index, it is necessary to match the most suitable enhancement parameters for the current tissue region from the pre-set set based on the pose. When the pose estimation confidence is low or the instrument movement is fast, relying solely on the current pose may cause a lag in tissue region judgment, resulting in a delay in enhancement parameter switching. This step achieves rapid optimization with high confidence through conservatism discrimination, and predicts the region based on fluctuation rate and performs pre-set fusion based on low confidence, enabling the enhancement parameters to adapt to the tissue region to be entered in advance, following the instrument movement trend. The adaptive parameter strategy selection process in this step is as follows: Figure 2 As shown.

[0064] Step S300 in the method provided in this application embodiment includes:

[0065] The conservatism level is determined by a weighted sum of the matching confidence index and pose estimation error index associated with the current spatial pose. This conservatism level is then compared to a preset conservatism threshold for conservatism processing. When the conservatism level meets the threshold, the current tissue region of the endoscope is determined based on the current spatial pose and the prior standard organ model. The enhancement parameter plan set is traversed, and the top N enhancement parameter plans corresponding to the current tissue region are matched and obtained, where N is a positive integer greater than or equal to 2. The plan with the best historical enhancement effect score is extracted from the top N enhancement parameter plans and used as the current enhancement parameter configuration. The remaining N-1 enhancement parameter plans are cached in the candidate plan library. Detailed explanation follows:

[0066] In this embodiment, the conservatism score refers to the reliability score calculated by weighting the matching confidence index and the pose estimation error index based on the current spatial pose. The score ranges from 0 to 1, with a higher value indicating a more reliable pose estimation. The conservatism threshold is the boundary value used to distinguish between conservative and look-ahead processing strategies. When the conservatism score is higher than the threshold, a conservative strategy is used for direct selection; when the conservatism score is lower than the threshold, a look-ahead strategy is used to predict the fluctuation rate. The top N enhancement parameter plans refer to the N parameter configuration schemes with the highest historical enhancement effect scores corresponding to the current tissue region, matched from the enhancement parameter plan set. The alternative plan library is a temporary storage area used to cache alternative plans not selected in the current configuration. When subsequent enhancement quality feedback detects a decline in effect, an alternative plan can be quickly selected from this library.

[0067] In this step, firstly, to assess the reliability of the current pose estimation, the conservatism needs to be calculated based on a weighted sum of the matching confidence index and the pose estimation error index. Conservatism = α × matching confidence + β × (1 / 2) * ... The pose estimation error is defined as (α and β are preset weighting coefficients, and the error upper limit is the preset maximum tolerance error value). The closer the conservatism is to 1, the more reliable the pose is.

[0068] Furthermore, when the conservatism meets the threshold condition, it indicates that the current pose estimation has high reliability. The current tissue region where the endoscope is located can be directly determined in the prior standard organ model based on the current spatial pose. The top N historical best-scoring plans corresponding to this region are matched from the enhancement parameter plan set. The best plan is extracted as the current enhancement parameter configuration, and the remaining plans are cached in the candidate plan library for subsequent correction.

[0069] Step S300 in the method provided in this application embodiment further includes:

[0070] When the conservatism does not meet the threshold condition, the preceding spatial pose sequence within a preset time window is obtained; the current fluctuation rate is calculated by combining the preceding spatial pose sequence with the current spatial pose; based on the fluctuation rate, the predicted tissue region corresponding to the next time step is predicted, and the consistency between the predicted tissue region and the current tissue region is determined; if consistent, the top N enhancement parameter schemes corresponding to the current tissue region are matched from the enhancement parameter scheme set based on the current spatial pose, and the optimal one is extracted as the current enhancement parameter configuration; if inconsistent, scheme matching is performed on the enhancement parameter scheme set based on the predicted tissue region and the current tissue region respectively, and the scheme matching results are adaptively fused to obtain the current enhancement parameter configuration. Detailed explanation follows:

[0071] In this embodiment, the fluctuation rate refers to the rate of change of the endoscope's spatial position within a preset time window. It can be calculated based on the root mean square of the displacement or the magnitude of the velocity vector between adjacent frames in the preceding spatial pose sequence, reflecting the speed of the instrument's movement within the body cavity. The preceding spatial pose sequence refers to the sequence of spatial poses estimated and stored within the preset time window before the current frame, used to analyze motion trends. The predicted tissue region refers to the tissue region that the endoscope may reach in the next time step, extrapolated from the current pose and fluctuation rate. This is obtained by multiplying the fluctuation rate by the predicted step size to obtain the position increment, superimposing it onto the current pose, and then searching for the corresponding region in the prior model.

[0072] In this step, firstly, if the conservatism does not meet the threshold condition, it indicates that there is significant uncertainty in the current pose estimation, and a forward-looking judgment based on the motion trend is required. The preceding spatial pose sequence within a preset time window is obtained, and the fluctuation rate is calculated in conjunction with the current spatial pose to quantify the current speed of the instrument's movement.

[0073] Furthermore, the predicted tissue region for the next time step is determined based on the fluctuation rate, and its consistency is compared with the current tissue region. If they are consistent, it indicates that the instrument is moving smoothly or stationary, and a pre-set plan can be directly matched based on the current pose and the best one selected. If they are inconsistent, it indicates that the instrument is moving across regions, and pre-set plans need to be matched separately for the two regions and adaptively fused to ensure a smooth transition of enhancement parameters.

[0074] Step S300 in the method provided in this application embodiment further includes:

[0075] Based on the predicted organization region, a trend enhancement parameter plan adapted to the next time step is obtained by matching from the enhancement parameter plan set; based on the current organization region, a current enhancement parameter plan adapted to the current time step is obtained by matching from the enhancement parameter plan set; according to the conservatism, the trend enhancement parameter plan and the current enhancement parameter plan are adaptively weighted and fused to generate the enhancement parameter configuration for the current frame. A detailed explanation follows:

[0076] In this embodiment, the trend enhancement parameter plan refers to the enhancement parameter configuration matched from the enhancement parameter plan set and adapted to the predicted tissue region, representing the parameter direction that may be needed in the next time step. Adaptive fusion refers to a strategy of weighted fusion of the current plan and the trend plan based on conservatism. The lower the conservatism, the greater the uncertainty of the current pose, and the higher the weight of the trend plan, so that the parameter configuration moves closer to the predicted region in advance.

[0077] In this step, to achieve smooth parameter evolution during cross-regional movement, it is necessary to match the trend enhancement parameter plan for the predicted organizational region and the current enhancement parameter plan for the current organizational region from the plan set. The fusion weights are determined based on the conservatism level; the lower the conservatism, the higher the weight of the trend plan. The two sets of parameters are then weighted and fused to obtain the enhancement parameter configuration for the current frame. The fusion formula is as follows: ,in This is the parameter vector for the current contingency plan. This is the parameter vector for the trend prediction plan.

[0078] For example, taking the pose estimation of three consecutive frames in a single endoscopic procedure as an example, the current spatial pose of the k-th frame is (14.6, 9.2, 46.8), the matching confidence is 0.82, and the pose estimation error is 1.3 mm. Let the weighting coefficients α = 0.6, β = 0.4, the upper limit of error be 5.0 mm, and the conservatism = 0.6 × 0.82 + 0.4 × (1 - 1.3 / 5.0) = 0.6 × 0.82 + 0.4 × 0.74 = 0.492 + 0.296 = 0.788. The preset conservatism threshold is 0.75; since 0.788 > 0.75, the condition is met. Based on the current spatial pose, the endoscope is located in the fundus region of the uterus by searching within the 3D organ model. Traverse the plans associated with the fundus region in the enhanced parameter plan set, match and obtain the top 3 plans with the highest historical scores, extract the parameter configuration of the highest scorer as brightness gain 1.15, contrast enhancement 1.30, and sharpening intensity 0.40, and use it as the current enhanced parameter configuration, and cache the other 2 plans to the candidate plan library.

[0079] In summary, this step uses a conservatism discrimination mechanism to quickly match the optimal plan when the pose is reliable, and to perform plan fusion based on the fluctuation rate prediction region when the pose is uncertain. This allows the enhancement parameters to respond to position changes in a timely manner, while avoiding frequent abrupt changes when uncertainty is high, thus achieving a balance between the stability and foresight of the enhancement strategy.

[0080] S400: Based on the current enhancement parameter configuration, perform parametric enhancement processing on the real-time image to obtain the enhanced image; and perform quality assessment and feedback correction on the enhanced image.

[0081] In this embodiment of the application, after the real-time image enhancement processing is completed according to the enhancement parameter configuration, the enhancement result needs to be evaluated for quality. When the effect does not meet expectations, correction is triggered in time. At the same time, a proxy model is built using accumulated historical data to iteratively optimize the enhancement configuration in order to continuously improve image quality.

[0082] Step S400 in the method provided in this application embodiment includes:

[0083] Based on the current enhancement parameter configuration, historical data is collected to construct a regression analysis sample set including associated historical spatial poses, historical enhancement parameter configurations, and historical enhancement effect scores. According to the regression analysis sample set, an enhancement parameter surrogate model based on Gaussian process regression is established and trained. The input space of the enhancement parameter surrogate model is the spatial pose of the endoscope, and the output space is the enhancement parameter configuration vector. The historical enhancement effect scores are used to characterize the quality weights of the samples. The current spatial pose is input into the enhancement parameter surrogate model for iterative solution to obtain the predicted enhancement parameter configuration. The predicted enhancement parameter configuration is compared with the current enhancement parameter configuration. When the difference exceeds a preset vector threshold, the current enhancement parameter configuration is updated with the predicted enhancement parameter configuration. Detailed explanation follows:

[0084] In this embodiment, the regression analysis sample set refers to a training data set constructed using spatial poses recorded in historical operations as input, augmentation parameter configurations adopted at corresponding times as output, and augmentation effect scores evaluated afterward as sample weights. Gaussian process regression is a nonparametric regression method based on a Bayesian framework, capable of providing predicted values ​​and their uncertainties, suitable for modeling nonlinear pose-parameter mapping relationships under small sample conditions. The augmentation parameter surrogate model refers to a prediction model trained using the regression analysis sample set, capable of outputting parameter configurations that are expected to achieve higher augmentation effects based on the input spatial pose. The predicted augmentation parameter configuration refers to the optimal parameter vector predicted by the surrogate model for the current spatial pose. The vector threshold refers to a preset vector distance threshold used to determine whether the difference between the predicted configuration and the current configuration is significant.

[0085] In this step, firstly, to accumulate the data needed to train the surrogate model, after each use of a certain enhancement parameter configuration, the corresponding spatial pose, parameter configuration, and subsequent enhancement effect score are recorded, gradually building a regression analysis sample set. Once the sample size reaches the preset number, training of the surrogate model can begin.

[0086] Furthermore, in order to establish a nonlinear mapping relationship between pose and parameters, Gaussian process regression is used to fit the sample set. During training, the historical enhancement effect score is used as the sample weight, so that the sample points with better performance contribute more to the model, and the generated surrogate model can be tilted towards the high-score region.

[0087] Furthermore, to explore better configurations based on the preliminary selection, the current spatial pose needs to be input into the trained surrogate model, and the prediction enhancement parameter configuration needs to be obtained through optimization. Gaussian process regression not only outputs the prediction mean but also provides the prediction variance, which can be used to balance exploration and utilization.

[0088] Finally, the predicted enhancement parameter configuration is compared with the current configuration obtained by the optimization of the pre-plan. If the Euclidean distance or the difference in each dimension between the two exceeds the preset vector threshold, it means that the surrogate model believes there is a significantly better configuration. At this time, the current enhancement parameter configuration is updated with the predicted configuration; otherwise, the current configuration remains unchanged.

[0089] For example, the pre-training process of the Gaussian process regression surrogate model is as follows: During system operation, after each endoscopic examination, the spatial pose corresponding to each frame, the enhancement parameter configuration used, and the enhancement effect score obtained by the operator's evaluation or automatic assessment are summarized. After accumulating 200 sets of valid samples, the Gaussian process regression surrogate model is constructed and trained using the three-dimensional coordinates of the spatial pose as input, the three-dimensional vector of the enhancement parameter configuration as output, and the normalized enhancement effect score as the sample weight. A radial basis function kernel is used as the kernel function, and the optimizer is L-BFGS, with a maximum of 500 iterations. After the model is trained, it is continuously updated online, with an incremental update performed every 50 new samples to maintain adaptability to changes in operator preferences.

[0090] Step S400 in the method provided in this application embodiment further includes:

[0091] An enhancement quality assessment is performed on the enhanced image to obtain an enhancement effect score. This score is compared to a preset quality benchmark. If the enhancement effect score is lower than the benchmark, an alternative enhancement parameter plan is selected from a candidate plan library, and a correction signal is generated. Based on the correction signal, the current spatial pose is corrected, and the current enhancement parameter configuration is corrected. Based on the corrected current spatial pose, the fusion estimation step is optimized through feedback. Detailed explanation follows:

[0092] In this embodiment, the enhancement effect score refers to a quantitative evaluation index of the enhanced image quality, which can be calculated by comprehensively considering multiple dimensions such as image contrast, signal-to-noise ratio, and edge sharpness. The quality benchmark refers to a lower limit threshold set based on the statistical distribution of enhancement effect scores during historical normal operation; a score below this threshold indicates a significant decrease in the current enhancement effect. The alternative plan library refers to the alternative enhancement parameter plans cached in the aforementioned steps, serving as a replacement source when the current plan's effect is unsatisfactory. The correction signal refers to the control signal that triggers pose correction and parameter correction, carrying information about the degree and source of the effect degradation.

[0093] In this step, firstly, to evaluate the actual performance of the current enhancement, a quality assessment needs to be performed on the enhanced image to obtain an enhancement effect score. When the score is lower than the preset quality benchmark, it indicates that the current parameter configuration is no longer suitable for the current scene, and an alternative enhancement parameter plan needs to be selected from the alternative plan library, and a correction signal needs to be generated.

[0094] Furthermore, based on the correction signal, the current spatial pose is corrected. For example, possible position drift can be corrected by adjusting the offset of the pose estimation or rerunning the pose estimation. At the same time, the current enhancement parameter configuration is corrected with alternative plans.

[0095] Finally, the corrected current spatial pose is fed back to the fusion estimation step. Parameters such as the process noise covariance or observation noise covariance of the Kalman filter can be adjusted to make subsequent pose estimation more sensitive to motion changes or smoother, thus achieving closed-loop joint adjustment from enhancement effect to localization strategy.

[0096] For example, taking the pose estimation of three consecutive frames in an endoscopic operation as an example, after enhancing the image in the k-th frame using the current enhancement parameter configuration, the enhancement effect score obtained by quality evaluation is 0.72, while the preset quality benchmark is 0.80. Since the score is lower than the benchmark, the correction process is triggered. The enhancement parameter preset ranked second in historical score is selected from the alternative preset library, with parameters of brightness gain 1.20, contrast enhancement 1.25, and sharpening intensity 0.45. After generating the correction signal, the current spatial pose (14.6, 9.2, 46.8) is adjusted to (14.4, 9.0, 46.6) through position correction, and the current enhancement parameter configuration is updated to the parameters of the alternative preset. At the same time, the corrected pose is fed back to the Kalman filter module, and the process noise covariance Q is appropriately increased to improve the tracking response speed of subsequent frames to motion changes. Through correction, the enhancement effect score of the (k+1)-th frame is restored to 0.85, and the image quality returns to a normal level.

[0097] In summary, this step uses a Gaussian process regression surrogate model to perform secondary optimization of the enhancement parameters, and utilizes a quality improvement closed-loop feedback to link and correct the parameters and pose, thereby achieving continuous iterative improvement of the enhancement strategy and enabling the system to maintain high-quality image output when facing complex and ever-changing endoscopic operating environments.

[0098] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0099] This application proposes an endoscopic image enhancement method. First, it acquires real-time images of the endoscope, instrument motion monitoring data, and a priori standard organ model associated with a set of enhancement parameter plans. Then, through multi-source information fusion, combining visual feature matching and instrument motion data, it estimates the current spatial pose of the endoscope and outputs a matching confidence index and a pose estimation error index. Next, based on the current spatial pose, it matches the enhancement parameter plan set, and adaptively fuses the plans based on conservatism discrimination and fluctuation rate to obtain the current enhancement parameter configuration. Finally, it performs parameterized enhancement on the real-time image according to the current enhancement parameter configuration, and performs quality assessment and feedback correction on the enhanced image. This method achieves dynamic enhancement and closed-loop optimization of endoscopic imaging through spatial pose perception and adaptive look-ahead control of enhancement parameters.

[0100] Example 2, as shown in the appendix Figure 3 As shown, based on the inventive concept of the endoscopic image enhancement method provided in Embodiment 1, this application also provides an endoscopic image enhancement system, specifically including:

[0101] The data acquisition module 11 is used to acquire real-time images of the endoscope, instrument motion monitoring data and a priori standard organ model, wherein the priori standard organ model is associated with an enhancement parameter pre-set set;

[0102] The pose estimation module 12 is used to estimate the current spatial pose of the endoscope by multi-source information fusion based on the real-time image, the instrument motion monitoring data and the prior standard organ model.

[0103] The parameter matching module 13 is used to match and determine the list of enhancement parameter plans corresponding to the set of enhancement parameter plans according to the current spatial pose and the prior standard organ model, and to adaptively fuse the list of enhancement parameter plans based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration.

[0104] The image enhancement module 14 is used to perform parametric enhancement processing on the real-time image according to the current enhancement parameter configuration to obtain the enhanced image.

[0105] In one embodiment, the data acquisition module 11 is further configured to: based on the physical characteristic portrait of the target object, filter a set of approximate samples that match the features of the physical characteristic portrait from a pre-stored historical examination sample library; perform statistical analysis on the historical examination data in the approximate sample set to obtain the geometric structure statistics of the organ cavity and the distribution statistics of each tissue region; based on the geometric structure statistics and the tissue region distribution statistics, construct a three-dimensional organ model containing the geometric shape and spatial size information of the inner surface of the organ cavity through three-dimensional reconstruction, and label each tissue region; extract approximate samples with the highest preset percentile of the enhancement effect score according to the approximate sample set, and configure corresponding enhancement parameter plans for each tissue region in the three-dimensional organ model to form the prior standard organ model; and associate and store the prior standard organ model with the physical characteristic portrait.

[0106] Furthermore, the data acquisition module 11 is also used to: acquire the real-time image at a predetermined frame rate through the image sensor at the front end of the endoscope, and perform noise reduction preprocessing and normalization processing on the real-time image before storing it in tissue; acquire the instrument motion monitoring data through the instrument motion sensor, the instrument motion monitoring data including linear displacement data, bending angle data and rotation angle data; and match and retrieve the corresponding prior standard organ model according to the vital sign profile of the target object.

[0107] In one embodiment, the pose estimation module 12 is further configured to: acquire the examination image data corresponding to the real-time image, and extract the individual organ size parameters of the target object based on the examination image data; establish an individual position transformation matrix based on the proportional relationship between the individual organ size parameters and the geometric structure statistical information in the prior standard organ model; perform frame feature matching and tracking based on the real-time image, and solve the perspective projection relationship in combination with the individual position transformation matrix to determine the first estimated position of the endoscope in the model space of the prior standard organ model; determine the second estimated position of the endoscope based on the instrument motion monitoring data and the preset instrument motion model, combined with the individual position transformation matrix; perform weighted fusion of the first estimated position and the second estimated position to obtain the current spatial pose estimation result, and simultaneously output the matching confidence index and the pose estimation error index; perform constraint verification on the current spatial pose estimation result, and output the current spatial pose after passing the verification, wherein the constraint verification includes at least geometric constraint verification, tissue region constraint verification and color feature constraint verification.

[0108] In one embodiment, the parameter matching module 13 is further configured to: determine the conservatism based on the weighted sum of the matching confidence index and the pose estimation error index associated with the current spatial pose, and compare the conservatism with a preset conservatism threshold for conservatism processing and discrimination; when the conservatism meets the passing condition of the conservatism threshold, determine the current tissue region of the endoscope based on the current spatial pose and the prior standard organ model; traverse the enhancement parameter plan set, match and obtain the first N enhancement parameter plans corresponding to the current tissue region, where N is a positive integer greater than or equal to 2; extract the one with the best historical enhancement effect score from the first N enhancement parameter plans as the current enhancement parameter configuration, and cache the remaining N-1 enhancement parameter plans in the candidate plan library.

[0109] Furthermore, the parameter matching module 13 is also used to: when the conservatism does not meet the pass condition of the conservatism threshold, obtain the preceding spatial pose sequence within a preset time window; combine the preceding spatial pose sequence with the current spatial pose to calculate the current fluctuation rate; predict the predicted organization region corresponding to the next time step based on the fluctuation rate, and determine the consistency between the predicted organization region and the current organization region; if consistent, match the first N enhancement parameter schemes corresponding to the current organization region from the enhancement parameter scheme set based on the current spatial pose, and extract the optimal one as the current enhancement parameter configuration; if inconsistent, perform scheme matching based on the predicted organization region and the current organization region in the enhancement parameter scheme set respectively, and adaptively fuse the scheme matching results to obtain the current enhancement parameter configuration.

[0110] Furthermore, the parameter matching module 13 is also used to: based on the predicted organization region, match and obtain a trend enhancement parameter plan adapted to the next time step from the enhancement parameter plan set; based on the current organization region, match and obtain a current enhancement parameter plan adapted to the current time step from the enhancement parameter plan set; and according to the conservatism, perform adaptive weighted fusion of the trend enhancement parameter plan and the current enhancement parameter plan to generate the enhancement parameter configuration for the current frame.

[0111] In one embodiment, the image enhancement module 14 is further configured to: collect historical data based on the current enhancement parameter configuration, and construct a regression analysis sample set including associated historical spatial poses, historical enhancement parameter configurations, and historical enhancement effect scores; establish and train an enhancement parameter surrogate model based on Gaussian process regression according to the regression analysis sample set, wherein the input space of the enhancement parameter surrogate model is the spatial pose of the endoscope, the output space is the enhancement parameter configuration vector, and the historical enhancement effect scores are used to characterize the quality weights of the samples; input the current spatial pose into the enhancement parameter surrogate model for iterative solution to obtain a predicted enhancement parameter configuration; compare the predicted enhancement parameter configuration with the current enhancement parameter configuration, and when the difference exceeds a preset vector threshold, update the current enhancement parameter configuration with the predicted enhancement parameter configuration.

[0112] Furthermore, the image enhancement module 14 is also configured to: perform enhancement quality assessment on the enhanced image to obtain an enhancement effect score; compare the enhancement effect score with a preset quality benchmark, and when the enhancement effect score is lower than the quality benchmark, select an alternative enhancement parameter plan from the alternative plan library and generate a correction signal; perform position correction on the current spatial pose and parameter correction on the current enhancement parameter configuration based on the correction signal; and perform feedback optimization on the fusion estimation step based on the corrected current spatial pose.

[0113] The endoscopic image enhancement system provided in this application enables intelligent endoscopic imaging enhancement management in scenarios such as obstetric and gynecological intrauterine examinations, minimally invasive surgical visualization navigation, and digestive endoscopy diagnosis and treatment. This management process encompasses multi-source data acquisition and spatial pose fusion estimation, adaptive matching and optimal fusion of enhancement parameters, and image enhancement execution and quality feedback correction. It can be integrated into endoscopic image processing workstations, surgical image management systems, or minimally invasive surgical navigation platforms. This effectively improves the real-time performance and tissue adaptability of image enhancement in complex body cavity environments, reduces the risk of unclear observation and operational interruptions caused by fixed enhancement modes or manual adjustments, and provides structured data support including spatial pose, enhancement parameter configuration, and enhancement effect scores for intraoperative decision-making and postoperative review. For specific enhancement procedures and optimization details of this system, please refer to Example 1.

[0114] It should be noted that the chronological order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

Claims

1. A method for enhancing endoscopic images, characterized in that, include: Acquire real-time images from the endoscope, instrument motion monitoring data, and a priori standard organ model, wherein the priori standard organ model is associated with a set of enhancement parameter plans; Based on the real-time images, the instrument motion monitoring data, and the prior standard organ model, the current spatial pose of the endoscope is estimated through multi-source information fusion. Based on the current spatial pose and the prior standard organ model, the enhancement parameter plan list corresponding to the enhancement parameter plan set is determined by matching, and the enhancement parameter plan list is adaptively fused based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration; Based on the current enhancement parameter configuration, parametric enhancement processing is performed on the real-time image to obtain the enhanced image.

2. The endoscopic image enhancement method as described in claim 1, characterized in that, The acquisition of the prior standard organ model includes: Based on the physical characteristics profile of the target object, a set of approximate samples matching the characteristics profile is selected from a pre-stored historical examination sample library; Statistical analysis is performed on the historical examination data in the approximate sample set to obtain statistical information on the geometric structure of organ cavities and the distribution statistics of each tissue region; Based on the geometric structure statistics and the tissue region distribution statistics, a three-dimensional organ model containing the geometric shape and spatial size information of the inner surface of the organ cavity is constructed through three-dimensional reconstruction, and each tissue region is labeled. Based on the approximate sample set, extract the approximate samples with the highest percentile scores of enhancement effect, and configure corresponding enhancement parameter plans for each tissue region in the three-dimensional organ model to form the prior standard organ model. The prior standard organ model is associated with and stored with the physical characteristic portrait.

3. The endoscopic image enhancement method as described in claim 2, characterized in that, Acquire real-time images from the endoscope, instrument motion monitoring data, and prior standard organ models, including: The real-time images are acquired at a predetermined frame rate through the image sensor at the front end of the endoscope, and the real-time images are preprocessed for noise reduction and normalization before being stored in the tissue. The motion monitoring data of the instrument is collected by the instrument motion sensor, and the motion monitoring data of the instrument motion includes linear displacement data, bending angle data and rotation angle data; Based on the physical characteristics profile of the target object, the corresponding prior standard organ model is matched and retrieved.

4. The endoscopic image enhancement method as described in claim 1, characterized in that, Based on the real-time images, the instrument motion monitoring data, and the prior standard organ model, the current spatial pose of the endoscope is estimated through multi-source information fusion, including: Obtain the examination image data corresponding to the real-time image, and extract the individual organ size parameters of the target object based on the examination image data; Based on the proportional relationship between the individual organ size parameters and the geometric structure statistics in the prior standard organ model, an individual position transformation matrix is ​​established; Based on the real-time image, frame feature matching and tracking are performed, and combined with the individual position transformation matrix, the perspective projection relationship is solved to determine the first estimated position of the endoscope in the model space of the prior standard organ model. Based on the instrument motion monitoring data and the preset instrument motion model, combined with the individual position transformation matrix, the second estimated position of the endoscope is determined; The first estimated position and the second estimated position are weighted and fused to obtain the current spatial pose estimation result, and the matching confidence index and pose estimation error index are output simultaneously. The current spatial pose estimation result is constrained and verified. Once the verification is successful, the current spatial pose is output. The constraint verification includes geometric constraint verification, tissue region constraint verification, and color feature constraint verification.

5. The endoscopic image enhancement method as described in claim 1, characterized in that, Based on the current spatial pose and the prior standard organ model, a list of enhancement parameter plans corresponding to the enhancement parameter plan set is determined by matching. The list of enhancement parameter plans is then adaptively fused based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration, including: The conservatism is determined by the weighted sum of the matching confidence index and the pose estimation error index associated with the current spatial pose, and the conservatism is compared with a preset conservatism threshold for conservatism processing and discrimination. When the conservatism meets the threshold condition, the current tissue region of the endoscope is determined based on the current spatial pose and the prior standard organ model. Traverse the set of enhancement parameter plans and match the first N enhancement parameter plans corresponding to the current tissue region, where N is a positive integer greater than or equal to 2; Extract the one with the best historical enhancement effect score from the first N enhancement parameter plans and use it as the current enhancement parameter configuration, and cache the remaining N-1 enhancement parameter plans in the candidate plan library.

6. The endoscopic image enhancement method as described in claim 5, characterized in that, Based on the current spatial pose and the prior standard organ model, a list of enhancement parameter plans corresponding to the enhancement parameter plan set is determined by matching. The list of enhancement parameter plans is adaptively fused based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration. The method also includes: When the conservatism does not meet the pass condition of the conservatism threshold, the preceding spatial pose sequence within the preset time window is obtained. Calculate the current fluctuation rate by combining the preceding spatial pose sequence with the current spatial pose; Based on the fluctuation rate, predict the predicted tissue region corresponding to the next time step, and determine the consistency between the predicted tissue region and the current tissue region; If they match, then based on the current spatial pose, match the top N enhancement parameter plans corresponding to the current tissue region from the enhancement parameter plan set, and extract the best one as the current enhancement parameter configuration; If they are inconsistent, the predicted organization region and the current organization region are respectively matched in the set of enhancement parameter plans, and the plan matching results are adaptively fused to obtain the current enhancement parameter configuration.

7. The endoscopic image enhancement method as described in claim 6, characterized in that, Adaptive fusion of the pre-plan matching results to obtain the current enhancement parameter configuration, including: Based on the predicted organization region, a trend enhancement parameter plan adapted to the next time step is obtained by matching from the set of enhancement parameter plans; Based on the current organizational region, a current enhancement parameter plan that is suitable for the current situation is obtained by matching from the set of enhancement parameter plans; Based on the conservatism, the trend enhancement parameter scheme and the current enhancement parameter scheme are adaptively weighted and fused to generate the enhancement parameter configuration for the current frame.

8. The endoscopic image enhancement method as described in claim 1, characterized in that, The optimal configuration is extracted as the current enhancement parameter configuration. Following this, the following steps are also included: Based on the current enhancement parameter configuration, historical data is collected to construct a regression analysis sample set including associated historical spatial poses, historical enhancement parameter configurations, and historical enhancement effect scores; Based on the regression analysis sample set, an enhanced parameter surrogate model based on Gaussian process regression is established and trained. The input space of the enhanced parameter surrogate model is the spatial pose of the endoscope, the output space is the enhanced parameter configuration vector, and the historical enhancement effect score is used to characterize the quality weight of the sample. The current spatial pose is input into the augmentation parameter surrogate model for iterative solution to obtain the predicted augmentation parameter configuration; The predicted enhancement parameter configuration is compared with the current enhancement parameter configuration. When the difference exceeds a preset vector threshold, the current enhancement parameter configuration is updated with the predicted enhancement parameter configuration.

9. The endoscopic image enhancement method as described in claim 1, characterized in that, Also includes: An enhancement quality assessment is performed on the enhanced image to obtain an enhancement effect score; The enhancement effect score is compared with a preset quality benchmark. When the enhancement effect score is lower than the quality benchmark, an alternative enhancement parameter plan is selected from the alternative plan library, and a correction signal is generated. Based on the correction signal, the current spatial pose is corrected, and the current enhancement parameter configuration is corrected. Based on the corrected current spatial pose, the fusion estimation step is optimized through feedback.

10. An endoscopic image enhancement system, characterized in that, The system is used to perform the endoscopic image enhancement method according to any one of claims 1-9, the system comprising: The data acquisition module is used to acquire real-time images of the endoscope, instrument motion monitoring data, and a priori standard organ models, wherein the priori standard organ models are associated with a set of enhancement parameter plans. The pose estimation module is used to estimate the current spatial pose of the endoscope by fusing multi-source information based on the real-time image, the instrument motion monitoring data and the prior standard organ model. The parameter matching module is used to match and determine the list of enhancement parameter plans corresponding to the enhancement parameter plan set according to the current spatial pose and the prior standard organ model, and to adaptively fuse the list of enhancement parameter plans based on the fluctuation rate of the current spatial pose to obtain the current enhancement parameter configuration. The image enhancement module is used to perform parametric enhancement processing on the real-time image according to the current enhancement parameter configuration to obtain the enhanced image.

Images