High-frame-rate imaging method for improving vertical resolution of image

Abstract

The invention relates to a high-frame-rate imaging method for improving the vertical resolution of an image, and the method comprises the steps: collecting a dynamic deformation image of a cornea section in a cornea deformation process of a patient in a process of blowing air to the cornea of the patient; a target area of the dynamic deformation image is determined, and the target area is a minimum area including the cornea; intercepting a target area in each frame of dynamic deformation image to form a continuous frame image of the target area; performing high-resolution processing on the continuous frame images of the target area; and performing cornea biomechanical characteristic evaluation on the continuous frame images after high-resolution processing. According to the method, the target area of the dynamic deformation image in the cornea deformation process of the patient is determined, the target area is the minimum area including the cornea, and high-resolution processing is performed on the continuous frame images of the target area, so that the data precision of the cornea area in the continuous frame images is improved; and the accuracy of the cornea biomechanical property evaluation result of the continuous frame images after high-resolution processing is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a high-frame-rate imaging method for improving the vertical resolution of images. Background Art

[0002] Accurate measurement of corneal biomechanical properties is very important for ophthalmic examinations and treatments. Existing corneal biomechanical properties are obtained based on consecutive frame images. For example, Corvis ST applies an air pulse to the center of a patient's cornea while a Scheimpflug high-speed imaging technique collects consecutive frame images of the entire process of the patient's cornea deformation caused by the blowing. The consecutive frame images collected by the image acquisition device are analyzed to obtain the patient's corneal biomechanical property parameters (such as deformation amplitude, deformation speed, and curvature, etc.).

[0003] However, the vertical resolution of Corvis ST is affected by light diffraction, resulting in poor resolution of the collected images, which will affect the accuracy of the analysis of corneal biomechanical properties. Summary of the Invention

[0004] To solve the above problems, the present invention provides a high-frame-rate imaging method for improving the vertical resolution of images, and the method includes:

[0005] During the process of blowing air to a patient's cornea, collect dynamic deformation images of the corneal cross-section during the deformation process of the patient's cornea; wherein, the dynamic deformation images are grayscale images;

[0006] Determine the target region of the dynamic deformation image, wherein the target region is the smallest region including the cornea;

[0007] Intercept the target region in each frame of the dynamic deformation image to form consecutive frame images of the target region;

[0008] Perform high-resolution processing on the consecutive frame images of the target region;

[0009] Evaluate the corneal biomechanical properties of the consecutive frame images after high-resolution processing.

[0010] Optionally, determining the target region of the dynamic deformation image includes:

[0011] Obtain the first frame of the dynamic deformation image;

[0012] Obtain the labeled edge of the cornea in the first frame of the dynamic deformation image;

[0013] Identify the recognized edge of the cornea in the first frame of the dynamic deformation image;

[0014] Form a labeled pixel set M according to the labeled edge, and form a recognized pixel set D according to the recognized edge;

[0015] The set of differential pixels I = (D ∪ M) - (D ∩ M);

[0016] Determine the change threshold according to the gray values of the elements in M, D, and I;

[0017] Determine the target region of the dynamic deformation image according to the change threshold.

[0018] Optionally, determining the change threshold according to the gray values of the elements in M, D, and I includes:

[0019] Determine the first mean μ1 and the first standard deviation σ1 of the gray values of all elements in M;

[0020] Determine the second mean μ2 and the second standard deviation σ2 of the gray values of all elements in D;

[0021] If I is an empty set, determine the change threshold as

[0022] If I is not an empty set, determine the third mean μ3 and the third standard deviation σ3 of the gray values of all elements in I; Determine the change threshold according to the relationship between σ1 and σ2, and μ3 and σ3.

[0023] Optionally, determining the change threshold according to the relationship between σ1 and σ2, and μ3 and σ3 includes:

[0024] If σ1 = σ2, determine the change threshold as

[0025] If σ1 ≠ σ2, determine the change threshold as

[0026] Optionally, determining the target region of the dynamic deformation image according to the change threshold includes:

[0027] Classify the elements in M by column to form the first set of labeled pixel points D1 for each column;

[0028] Determine the minimum row number and the maximum row number of all pixel points in D1 for each column;

[0029] Determine the pixel points one row up from the minimum row number and one row down from the maximum row number of each column as the increased pixel points, and form the second set of labeled pixel points D2 for each column from the increased pixel points and D1;

[0030] Determine the changed pixel points for each column according to D1, D2, and the change threshold for each column;

[0031] Determine the minimum rectangle containing the changed pixel points of all columns as the target region.

[0032] Optionally, according to D1, D2 of each column and the change threshold, determine the changed pixel points of each column, including:

[0033] For any column, determine the fourth mean value μ4 of the gray values of all elements in D1 of the column;

[0034] Determine the fifth mean value μ5 of the gray values of all elements in D2 of the column;

[0035] If μ5 > μ4 and then determine all pixel points in D1 of any column as the changed pixel points of the column; where α1 is a preset gray threshold, and α1 ∈ (0, 1);

[0036] If μ5 ≤ μ4, or μ5 > μ4 but then determine the changed pixel points according to D1, D2 of any column and the change threshold.

[0037] Optionally, according to D1, D2 of any column and the change threshold, determine the changed pixel points, including:

[0038] Determine the change set D3 = D2 - D1;

[0039] Determine the sixth mean value μ6 and the sixth standard deviation σ6 of the gray values of all elements in D3;

[0040] If then determine all pixel points in D1 of any column as the changed pixel points of the column; where α2 is a preset change accuracy, Δ is the change threshold, μ1 is the first mean value of the gray values of all elements in M, and σ1 is the first standard deviation of the gray values of all elements in M;

[0041] If then update all pixel points in D1 of any column to all pixel points in D2 of the column, and repeatedly execute the steps of determining the pixel points in the upper row of the minimum number of rows of each column and the pixel points in the lower row of the maximum number of rows as the increased pixel points, and forming the second labeled pixel point set D2 of each column from the increased pixel points and D1 and the subsequent steps.

[0042] Optionally, determine the target area by determining the minimum rectangle containing the changed pixel points of all columns, including:

[0043] Expand the changed pixel points of each column upward and downward by α3 pixel points respectively; where α3 is a preset redundancy value, α3 is determined according to the patient attributes, α3 ∈ [0, α4], α4 is the maximum redundancy value, and α4 = min{R u , R d}-R n , R uR is the minimum number of pixels between the center point of the changed pixel points in each column and the upper edge of the first-frame dynamic deformation image. d R is the minimum number of pixels between the center point of the changed pixel points in each column and the lower edge of the first-frame dynamic deformation image. n is the maximum value of half of the total number of changed pixel points in each column;

[0044] Determine the smallest rectangle containing all the expanded changed pixel points in all columns as the target area.

[0045] Optionally, perform high-resolution processing on the consecutive frame images of the target area, including:

[0046] For any frame image in the consecutive frame images of the target area, through the following function Perform high-resolution processing;

[0047] where i is the frame image identifier in the consecutive frame images of the target area, j is the pixel point identifier in frame image i, is the value of the j-th pixel point in frame image i, is the value of the -th pixel point after high-resolution processing of frame image i, D is the downsampling matrix, H is the blurring matrix, λ is the penalty parameter, is the regularization function.

[0048] Optionally,

[0049] where nor is the normalization parameter, (u, v) is the coordinate of the pixel point in the image, and s is the variance estimation value obtained according to frame image i.

[0050] The present invention relates to a high-frame-rate imaging method for improving the vertical resolution of an image. The method includes: during the process of blowing air into the patient's cornea, collecting dynamic deformation images of the corneal cross-section during the deformation of the patient's cornea; determining the target area of the dynamic deformation image, where the target area is the smallest area including the cornea; intercepting the target area in each frame of the dynamic deformation image to form consecutive frame images of the target area; performing high-resolution processing on the consecutive frame images of the target area; and evaluating the corneal biomechanical properties of the consecutive frame images after high-resolution processing. This method determines the target area of the dynamic deformation image during the deformation of the patient's cornea, and the target area is the smallest area including the cornea. By performing high-resolution processing on the consecutive frame images of the target area, the data accuracy of the corneal area in the consecutive frame images is improved, and the accuracy of the evaluation result of the corneal biomechanical properties of the consecutive frame images after high-resolution processing is ensured. Description of the Drawings

[0051] Figure 1Schematic flowchart of a high - frame - rate imaging method for improving the vertical resolution of an image provided by an embodiment of the present application;

[0052] Figure 2 Schematic diagram of a dynamic deformation image provided by an embodiment of the present application;

[0053] Figure 3 Schematic diagram of rows and columns in the first - frame dynamic deformation image provided by an embodiment of the present application;

[0054] Figure 4 Another schematic diagram of rows and columns in the first - frame dynamic deformation image provided by an embodiment of the present application. Detailed implementation manners

[0055] To better explain the present invention for easy understanding, the present invention will be described in detail below with reference to the accompanying drawings through specific implementation manners.

[0056] When an image acquisition device acquires an image, it acquires an image of the patient's eye position. However, the cornea is a small part of the eye, which makes each frame of the image acquired by the image acquisition device include a large amount of images of non - cornea regions. This increases the amount of noise data in each frame of the image, not only increasing the computing power resource requirements for subsequent corneal biomechanical property analysis, but also the accuracy problem of corneal region data brought by the noise data will affect the accuracy of corneal biomechanical property analysis.

[0057] To improve the analysis accuracy, the present invention provides a high - frame - rate imaging method for improving the vertical resolution of an image. The method includes: during the process of blowing air onto the patient's cornea, acquiring dynamic deformation images of the corneal cross - section during the deformation of the patient's cornea; determining a target region of the dynamic deformation image, where the target region is the smallest region including the cornea; intercepting the target region in each frame of the dynamic deformation image to form a continuous - frame image of the target region; performing high - resolution processing on the continuous - frame image of the target region; and evaluating the corneal biomechanical properties of the continuous - frame image after high - resolution processing. This method determines the target region of the dynamic deformation image during the deformation of the patient's cornea, and the target region is the smallest region including the cornea. By performing high - resolution processing on the continuous - frame image of the target region, the data accuracy of the corneal region in the continuous - frame image is improved, ensuring the accuracy of the evaluation result of the corneal biomechanical properties of the continuous - frame image after high - resolution processing.

[0058] Refer to Figure 1 , this embodiment provides a high - frame - rate imaging method for improving the vertical resolution of an image, and its implementation process is as follows:

[0059] 101, during the process of blowing air onto the patient's cornea, acquire dynamic deformation images of the corneal cross - section during the deformation of the patient's cornea.

[0060] Among them, the dynamic deformation image is a grayscale image.

[0061] The dynamic deformation image of the corneal cross-section can be collected by Corvis ST. Corvis ST applies an air pulse to the center of the patient's cornea and simultaneously collects consecutive frame images of the entire process of the patient's cornea deformation caused by blowing using Scheimpflug high-speed imaging technology. That is to say, the method provided in this embodiment performs high-resolution imaging on the pictures taken by Corvis ST using an ultra-high-speed Scheimpflug camera. Since the images taken by Corvis are grayscale images, the dynamic deformation image is a grayscale image. For example, any frame of the dynamic deformation image is as Figure 2 shown.

[0062] 102. Determine the target area of the dynamic deformation image.

[0063] The target area is the smallest area including the cornea.

[0064] After obtaining the dynamic deformation image in step 101, the collecting doctor will mark the corneal area in the first frame of the dynamic deformation image to determine the accurate position of the corneal area. At the same time, the corneal area in the first frame of the dynamic deformation image will also be identified through existing image recognition technology, and then the target area will be determined according to the relationship between the recognized corneal area and the marked corneal area in step 102.

[0065] The implementation process of step 102 is as follows:

[0066] 102-1. Obtain the first frame of the dynamic deformation image.

[0067] 102-2. Obtain the marked edge of the cornea in the first frame of the dynamic deformation image.

[0068] This marked edge is marked by the collecting doctor. That is, after obtaining the dynamic deformation image in step 101, the collecting doctor marks the corneal area in the first frame of the dynamic deformation image, and this mark can be obtained in step 102-2.

[0069] Since the marked edge is marked by the collecting doctor, it can be considered that the area included in the marked edge is the accurate corneal area.

[0070] However, the doctor's marking may have deviations caused by reasons such as hand shaking. Therefore, there may be some corneal areas outside the marked area. Therefore, the area included in the marked edge is the accurate corneal area, but not necessarily the entire corneal area. Subsequently, the complete corneal area will be determined based on the marked edge and the corneal area identified through existing image recognition technology.

[0071] 102-3. Identify the recognized edge of the cornea in the first frame of the dynamic deformation image.

[0072] The recognition edge is recognized by existing image recognition technology, that is, after obtaining the dynamic deformation image in step 101, the corneal region in the initial image is recognized by existing image recognition technology, and then the edge of this region is obtained. Therefore, in step 102-3, this edge can be obtained.

[0073] Since the recognition edge is automatically recognized by image recognition technology, there may be recognition errors due to recognition accuracy. Therefore, the recognition edge will be used as a reference to adjust the labeled edge, and finally an accurate corneal region is obtained.

[0074] 102-4. Form a labeled pixel set M according to the labeled edge, and form a recognition pixel set D according to the recognition edge.

[0075] As Figure 3 shown, the pixel point in the upper left corner of the first-frame dynamic deformation image can be used as the pixel point in row 0 and column 0. From this, the column number increases to the right, and the row number increases downward, and then the row numbers and column numbers of all pixel points in the first-frame dynamic deformation image are obtained.

[0076] It should be noted that Figure 3 the rows and columns in are only for illustration, and in the actual application process, they are determined according to the actual pixel situation in the image.

[0077] Since the corneal region in the first-frame dynamic deformation image is an irregular curve with a certain width, the labeled edge and the recognition edge are only the edges of the corneal region obtained by different methods. As Figure 3 shown, the obtained pixel points are those in row 5 and column 7 and row 7 and column 7. However, the pixel points between the two pixel points (such as the pixel point in row 6 and column 7) are also pixel points in the corneal region. Therefore, in step 102-4, the coordinates of all pixel points of the labeled edge and the recognition edge will be determined (such as the coordinates are (column number, row number)), and then by column, all the coordinates corresponding to the row numbers in the coordinates of each column (that is, the column coordinates remain unchanged, and only the row coordinates are supplemented) are used as the filling coordinates (for example, the pixel point coordinates of the labeled edge in the 7th column are (7,5) and (7,7), and the coordinates with all row numbers between 5 and 7 are (7,6) as the supplementary coordinates). In this way, all the coordinates recognized by the labeled edge and their corresponding supplementary coordinates form the labeled pixel set M, and all the coordinates recognized by the recognition edge and their corresponding supplementary coordinates form the recognition pixel set D. For example, for the 7th column in the labeled edge, it has three coordinates in the labeled pixel set M, which are (7,5), (7,6), and (7,7) respectively.

[0078] In this way, the labeled pixel set M contains all the coordinates of the region surrounded by the labeled edge, and the recognition pixel set D contains all the coordinates of the region surrounded by the recognition edge.

[0079] 102-5, the set of differential pixels I = (D ∪ M) - (D ∩ M).

[0080] Since the labeled edge and the recognized edge are obtained by different methods, the labeled edge is obtained by doctor labeling, and the recognized edge is obtained by image recognition technology. Different methods may produce errors. Therefore, the regions enclosed by the labeled edge and the recognized edge are not necessarily the same region, that is, the regions enclosed by the labeled edge and the recognized edge may have non-overlapping parts.

[0081] Thus, D ∪ M is all the pixel points of the regions enclosed by the labeled edge and the recognized edge. D ∩ M is the pixel points common to the regions enclosed by the labeled edge and the recognized edge. (D ∪ M) - (D ∩ M) is the pixel points of the non-overlapping parts of the regions enclosed by the labeled edge and the recognized edge.

[0082] In the actual application process, there may be three relationships between the regions enclosed by the labeled edge and the recognized edge. The first is that the two are exactly the same (at this time, the set of differential pixels I is an empty set), which means that the regions enclosed by the labeled edge and the recognized edge are the same, that is, the edges of the labeled and recognized corneal regions are very accurate, which is normal. The second is that they are partially the same (at this time, the set of differential pixels I is non-empty, but I is smaller than D or M), which means that there is a deviation between the labeled edge and the recognized edge. The overlapping part can be regarded as the main part of the patient's cornea, and the non-overlapping part can be regarded as the edge part of the patient's corneal region. This deviation may be caused by insufficient accuracy of edge part recognition, which is also normal. The third is that they are completely different (at this time, the set of differential pixels I is non-empty, but I = (D ∪ M)), which means that the labeled edge and the recognized edge are two independent regions. This situation is either a labeling error, or a recognition error, or both labeling and recognition are incorrect, which is abnormal. Therefore, if the third situation occurs, this process can be exited and the high-frame-rate imaging method provided in this embodiment for improving the vertical resolution of the image can be re-executed.

[0083] 102-6, determine the change threshold according to the gray values of each element in M, D, and I.

[0084] The implementation process of step 102-6 is as follows:

[0085] 201, determine the first mean μ1 and the first standard deviation σ1 of the gray values of all elements in M.

[0086] In step 201, the mean (i.e., the first mean μ1) of the gray values of the pixel points (i.e., all elements in M) of the region enclosed by the labeled edge and the standard deviation (i.e., the first standard deviation σ1) of the gray values of the pixel points (i.e., all elements in M) of the region enclosed by the labeled edge will be determined.

[0087] The first mean μ1 represents the average gray value of the region enclosed by the labeled edge, and the first standard deviation σ1 represents the degree of variation of the gray values of the region enclosed by the labeled edge.

[0088] 202. Determine the second mean μ2 and the second standard deviation σ2 of the gray values of all elements in D.

[0089] In step 202, the mean (i.e., the second mean μ2) of the gray values of the pixel points in the region enclosed by the recognized edge (i.e., all elements in D) and the standard deviation (i.e., the second standard deviation σ2) of the gray values of the pixel points in the region enclosed by the recognized edge (i.e., all elements in D) will be determined.

[0090] The second mean μ2 represents the average gray value of the region enclosed by the recognized edge, and the second standard deviation σ2 represents the degree of variation of the gray values of the region enclosed by the recognized edge.

[0091] 203. If I is an empty set, determine the change threshold as

[0092] If I is an empty set, it means that both the labeled edge and the recognized edge are very accurate. Then, the average gray value μ1 of the region enclosed by the labeled edge should be the same as the average gray value μ2 of the region enclosed by the recognized edge, and the degree of variation σ1 of the gray values of the region enclosed by the labeled edge should be the same as the degree of variation σ2 of the gray values of the region enclosed by the recognized edge. Considering the possible uncontrollable errors in actual applications, in this step, when I is an empty set, the change threshold

[0093] is the average gray value of the corneal region (including the region enclosed by the labeled edge and the region enclosed by the recognized edge), is the degree of variation of the gray values of the corneal region (including the region enclosed by the labeled edge and the region enclosed by the recognized edge), is the ratio of the variation of the gray values of the corneal region (including the region enclosed by the labeled edge and the region enclosed by the recognized edge) to the average gray value of the corneal region (including the region enclosed by the labeled edge and the region enclosed by the recognized edge). This ratio represents the gray value fluctuation of the corneal region (including the region enclosed by the labeled edge and the region enclosed by the recognized edge). The larger this value is, the greater the gray value fluctuation of the corneal region (including the region enclosed by the labeled edge and the region enclosed by the recognized edge). This fluctuation is the real fluctuation of the patient's cornea. The change thresholds Δ of different patients may be the same or different. Subsequently, determining the target region based on the change threshold Δ can achieve the determination of personalized target regions, ensuring that the target region of each user conforms to the situation of that user itself and improving the accuracy of the target region.

[0094] 204. If I is a non-empty set, determine the third mean μ3 and the third standard deviation σ3 of the grayscale values of all elements in I. Determine the change threshold according to the relationship between σ1 and σ2, as well as μ3 and σ3.

[0095] Here is the case of M and D, that is, the set of difference pixels I is non-empty, but I is smaller than D or M, as Figure 3 shown in the figure.

[0096] In step 204, the mean (i.e., the third mean μ3) of the grayscale values of the pixel points (i.e., all elements in I) that mark the non-coincident part of the minimum circle and identify the minimum circle, and the standard deviation (i.e., the third standard deviation σ3) of the grayscale values of the pixel points (i.e., all elements in I) covered by the identified minimum circle are determined. Determine the change threshold according to the relationship between σ1 and σ2, as well as μ3 and σ3.

[0097] The third mean μ3 represents the average grayscale value of the non-coincident part of the area enclosed by the marked edge and the area enclosed by the identified edge, and the third standard deviation σ3 represents the degree of change of the grayscale values of the non-coincident part of the area enclosed by the marked edge and the area enclosed by the identified edge.

[0098] In addition, there are two relationships between σ1 and σ2. One is σ1 = σ2. This relationship indicates that although there are differences between the area enclosed by the marked edge and the area enclosed by the identified edge, the degree of change of the grayscale values of each pixel point in the area enclosed by the marked edge and the area enclosed by the identified edge is the same. The other is σ1 ≠ σ2. This relationship indicates that there are differences between the area enclosed by the marked edge and the area enclosed by the identified edge, and the degree of change of the grayscale values of each pixel point in the area enclosed by the marked edge and the area enclosed by the identified edge is also different.

[0099] For the case of σ1 = σ2, the change threshold can be determined where is the average grayscale value of the corneal region (including the area enclosed by the marked edge and the area enclosed by the identified edge), σ2 is the degree of change of the grayscale values of the corneal region (including the area enclosed by the marked edge and the area enclosed by the identified edge) (actually represents the degree of change of the grayscale values of the corneal region (including the area enclosed by the marked edge and the area enclosed by the identified edge), but because σ1 = σ2, so here σ2 is directly used to represent the degree of change of the grayscale values of the corneal region (including the area enclosed by the marked edge and the area enclosed by the identified edge). In practical applications, σ1 can also be used to represent the degree of change of the grayscale values of the corneal region (including the area enclosed by the marked edge and the area enclosed by the identified edge), that is does not affect the change threshold Δ), Characterizes the gray value fluctuations in the corneal region (including the region enclosed by the marked edge and the region enclosed by the recognition edge).

[0100] Characterizes the gray value fluctuations in the non-overlapping part of the region enclosed by the marked edge and the region enclosed by the recognition edge.

[0101] The change threshold Δ is the ratio of the gray value fluctuation in the overlapping part of the region enclosed by the marked edge and the region enclosed by the recognition edge to the gray value fluctuation in the non-overlapping part. This ratio characterizes the difference between the gray value fluctuation in the main part of the patient's cornea and the gray value fluctuation in the edge part. This difference is the true difference of the patient's cornea. The differences of different patients may be the same or different. Taking this difference as the change threshold Δ can also reflect the individualization of different patients. Subsequently, based on the change threshold Δ, the target region can be determined, realizing the determination of the personalized target region, ensuring that the target region of each user conforms to the situation of that user itself, and improving the accuracy of the target region.

[0102] For the case of σ1≠σ2, the change threshold can be determined Wherein, Is the gray value fluctuation situation of the region enclosed by the marked edge. Since the marked edge is marked by the collecting doctor and is considered accurate, the region enclosed by the marked edge will be used as the main corneal region to determine the change threshold. Characterizes the proportion of the difference in the standard deviation between the region enclosed by the recognition edge and the region enclosed by the marked edge to the standard deviation of the region enclosed by the recognition edge. The larger this proportion, the greater the difference between the standard deviations of the region enclosed by the recognition edge and the region enclosed by the marked edge. Therefore, Characterizes the degree of gray value change between the region enclosed by the recognition edge and the region enclosed by the marked edge. At Based on this, considering the degree of gray value change between the region enclosed by the recognition edge and the region enclosed by the marked edge Adjust the region enclosed by the marked edge, and obtain the gray value fluctuation situation of the adjusted corneal region (i.e., ).

[0103] Characterizes the gray value fluctuations in the non-overlapping part of the region enclosed by the marked edge and the region enclosed by the recognition edge.

[0104] The change threshold Δ is the ratio of the gray value fluctuation of the adjusted corneal region to the gray value fluctuation of the non-overlapping part. The adjusted corneal region is taken as the main part of the patient's cornea. In this way, this ratio characterizes the difference between the gray value fluctuation of the main part of the patient's cornea and the gray value fluctuation of the edge part. This difference is the real difference of the patient's cornea. The differences of different patients may be the same or different. Taking this difference as the change threshold Δ can also reflect the personalization of different patients. Subsequently, determining the target region based on the change threshold Δ can realize the determination of the personalized target region, ensure that the target region of each user conforms to the situation of that user itself, and improve the accuracy of the target region.

[0105] Based on the above description, the change threshold can accurately reflect the gray value change of the corneal region of the current patient.

[0106] 102-7. Determine the target region of the dynamic deformation image according to the change threshold.

[0107] The implementation process of step 102-7 is as follows:

[0108] 301. Classify each element in M by column to form the first labeled pixel point set D1 of each column.

[0109] As can be seen from step 102-4, M is a set formed by the coordinates of all labeled edge recognitions and their corresponding supplementary coordinates. In step 301, taking columns as identifiers, the coordinates in the same column in M will form a D1. For example, Figure 4 as shown, for the 7th column, since the coordinates of the 7th column in M are (7, 5), (7, 6), and (7, 7), therefore, the D1 of the 7th column includes: (7, 5) (i.e., Figure 4 point A1 in Figure 4 ), (7, 6), and (7, 7) (i.e.,

[0110] point A2 in

[0111] ).

[0112] It should be noted that in this embodiment, each coordinate and pixel point have a one-to-one correspondence relationship. Each coordinate corresponds to a unique pixel point, and each pixel point corresponds to a unique coordinate. Therefore, the minimum row number and maximum row number of all pixel points in D1 of each column in step 302 are also the row minimum value and row maximum value of all coordinates in D1 of each column. Without special instructions, a pixel point can be understood as the coordinate of the pixel point, and the coordinate can also be understood as the pixel point corresponding to the coordinate. The two actually correspond to the same object.

[0113] 303 , determine the pixel points in the upper row of the minimum number of rows in each column and the pixel points in the lower row of the maximum number of rows as added pixel points, and form a second annotated pixel point set D2 of each column by the added pixel points and D1 .

[0114] For example, move the pixel point with the minimum row number 5 in the 7th column to the pixel point in the next row (i.e., the pixel point in the 7th column and the 4th row, such as Figure 4 A0 in the figure) and the pixel point with the largest number of rows down one row (i.e., the 7th column and the 8th row, such as Figure 4 Point A3 in the figure is determined as an additional pixel point, and the additional pixel point and D1 form a second annotated pixel point set D2 of each column. For example, D2 includes: (7, 4) (i.e. Figure 4 Point A0), (7, 5) (i.e. Figure 4 Points A1), (7, 6), (7, 7) (i.e. Figure 4 Point A2) and (7, 8) (i.e. Figure 4 point A3).

[0115] That is to say, for any column, its D2 includes all elements in D1, and has two more elements than D1, which are the pixel values ​​of the previous row and the pixel values ​​of the next row. That is, D1 is expanded outward by one pixel to form D2.

[0116] 304 , determining the changed pixel points of each column according to D1 , D2 and the change threshold of each column.

[0117] For any column, the changed pixel points of any column will be determined according to D1, D2 and the change threshold of any column in step 304. The specific implementation process is as follows:

[0118] 304-1, determine a fourth mean μ4 of the grayscale values ​​of all elements in D1 of any column.

[0119] In step 304 - 1 , the average value (ie, fourth average value μ4 ) of the grayscale values ​​of the pixel points covered by any column of the corneal region (ie, all elements in D1 ) is determined.

[0120] The fourth mean μ4 represents the average grayscale value of the area covered by any column of corneal regions.

[0121] 304-2, determine the fifth mean μ5 of the grayscale values ​​of all elements in D2 of any column.

[0122] In step 304-2, the average value (ie, the fifth average value μ5) of the grayscale values ​​of the pixel points (ie, all elements in D2) covered by one row extending outward from any column of the corneal area is determined.

[0123] The fifth mean μ5 represents the average gray value of the area covered by one row extending outward from any column of corneal areas.

[0124] 304 - 3, if μ5 > μ4 and then determine all the pixel points in D1 of any column as the changing pixel points of any column. If μ5 ≤ μ4, or, μ5 > μ4 but then determine the changing pixel points according to D1, D2 of any column and the change threshold.

[0125] where α1 is a preset gray - level threshold, and α1 ∈ (0, 1).

[0126] μ5 > μ4 indicates that the average gray - level value of the area covered by expanding the corneal region of any column by one row is smaller than the average gray - level value of the area covered by the corneal region of any column. This is caused by the increase in the gray - level values of the pixel points included in the additional row of expansion. The gray - level value outside the corneal region will be significantly greater than the gray - level value inside the corneal region. Therefore, this increase is due to including pixel points in the non - corneal region. μ5 ≤ μ4 indicates that the average gray - level value of the area covered by expanding the corneal region of any column by one row is smaller than or equal to the average gray - level value of the area covered by the corneal region of any column. This is caused by the decrease or equality of the gray - level values of the pixel points included in the additional row of expansion. The gray - level value outside the corneal region will be significantly greater than the gray - level value inside the corneal region. Therefore, this decrease or equality is considered to include pixel points in the corneal region and does not include pixel points in the non - corneal region.

[0127] It indicates the proportion of the increase in the average gray - level value of the area covered by expanding the corneal region of any column by one row to the average gray - level value of the area covered by the corneal region of any column. The larger this proportion, the greater the degree of the non - corneal region included in the area covered by expanding the corneal region of any column by one row.

[0128] μ5 > μ4 and It indicates that the area covered by expanding the corneal region of any column by one row includes the non - corneal region, and the proportion of the non - corneal region included in the area covered by expanding the corneal region of any column by one row is relatively large. At this time, it is considered that the corneal region of any column is the smallest region including the entire corneal region. If it is one more pixel point larger, it will include more non - corneal regions. Therefore, the expansion of the corneal region of any column can be stopped, and the corneal region of any column (i.e., all pixel points in D1 of any column) is determined as the changing pixel points of any column.

[0129] μ5 ≤ μ4 indicates that the area covered by expanding the corneal region of any column by one row does not include the non - corneal region. μ5 > μ4 but It is described that the area covered by expanding any column of the corneal region by one row includes the non-corneal region, but the proportion of the included non-corneal region is small. It is necessary to determine whether to stop expanding any column of the corneal region according to D1, D2 and the change threshold, and then determine the changed pixel points, that is, the changed pixel points will be determined according to D1, D2 and the change threshold of any column.

[0130] Among them, the process of determining the changed pixel points according to D1, D2 and the change threshold is as follows:

[0131] 1. Determine the change set D3 = D2 - D1.

[0132] D3 is the two pixel points increased by expanding any column of the corneal region by one row.

[0133] 2. Determine the sixth mean value μ6 and the sixth standard deviation σ6 of the gray values of all elements in D3.

[0134] The sixth mean value μ6 is the mean value of the gray values of the two pixel points increased by expanding any column of the corneal region by one row. The sixth standard deviation σ6 is the standard deviation of the gray values of the two pixel points increased by expanding any column of the corneal region by one row.

[0135] 3. If then determine all pixel points in D1 of any column as the changed pixel points of any column. If then update all pixel points in D1 of any column to all pixel points in D2 of any column, and repeat the steps of determining the pixel points in the upper row of the minimum number of rows and the pixel points in the lower row of the maximum number of rows of each column as the increased pixel points, and forming the second labeled pixel point set D2 of each column from the increased pixel points and D1 (that is, repeating step 303) and subsequent steps.

[0136] Among them, α2 is the preset change accuracy, Δ is the change threshold (that is, obtained in step 102-6), μ1 is the first mean value of the gray values of all elements in M, and σ1 is the first standard deviation of the gray values of all elements in M.

[0137] is the gray value fluctuation situation of the area enclosed by the labeled edge, is the gray value fluctuation situation of the two pixel points increased by expanding any column of the corneal region by one row. is the ratio of the change degree of the gray value of the area enclosed by the labeled edge to the gray value fluctuation of the two pixel points increased by expanding any column of the corneal region by one row. α2×Δ characterizes the normal fluctuation situation of the corneal region and non-corneal region of the current patient.

[0138] If It indicates that the fluctuation of the gray value brought by the two pixel points added when any column of the corneal region is expanded outward by one row is greater than the normal fluctuation of the patient itself. Then it is considered that any column of the corneal region is the smallest region including the entire corneal region. Expanding any column of the corneal region outward by one row will bring abnormal gray value fluctuations. Therefore, the expansion of any column of the corneal region can be stopped, and any column of the corneal region (i.e., all pixel points in D1 of any column) is determined as the changed pixel points.

[0139] If It indicates that the fluctuation of the gray value brought by the two pixel points added when any column of the corneal region is expanded outward by one row is not greater than the normal fluctuation of the patient itself. Then it is considered that any column of the corneal region is not the smallest region including the entire corneal region. Expanding any column of the corneal region outward by one row will not have too much impact on the gray value fluctuation. Compared with any column of the corneal region, the region covered by expanding any column of the corneal region outward by one row includes a more complete corneal region. Therefore, it will expand outward by one row on the basis of the region covered by expanding any column of the corneal region outward by one row to find the smallest region that accurately includes the entire corneal region. Therefore, all pixel points in D1 of any column will be updated to all pixel points in D2 of any column, and steps 303 and subsequent steps will be repeatedly executed until the smallest region including the entire corneal region is found.

[0140] In addition, α1 is a preset empirical value determined according to the imaging accuracy. The higher the imaging accuracy, the larger the value of α1, so that the possibility of is reduced. More non-corneal region pixels between the region covered by expanding any column of the corneal region outward by one row and any column of the corneal region are required to determine the changed pixel points, and the determined changed pixel points are more accurate.

[0141] α2 is a preset empirical value determined according to the imaging accuracy. The higher the imaging accuracy, the larger the value of α2, and the larger α2×Δ is. So that the possibility of is reduced. A larger pixel value fluctuation brought by the region covered by expanding any column of the corneal region outward by one row is required to determine the changed pixel points, and the determined changed pixel points are more accurate.

[0142] 305. Determine the smallest rectangle containing the changed pixel points of all columns as the target region.

[0143] In step 305, the smallest rectangle containing the changed pixel points of all columns can be directly determined as the target region.

[0144] However, in the specific implementation, it is very difficult for the patient to ensure absolute stillness. Therefore, a redundancy value α3 (e.g., α3 = 3) of the pixel value change caused by patient movement can be estimated in advance, and the changed pixel points in each column are respectively expanded upward and downward by α3 pixel points. The smallest rectangle containing the expanded changed pixel points of all columns is determined as the target area.

[0145] Among them, α3 is a preset redundancy value, and α3 is determined according to patient attributes, such as according to the patient's self-control ability.

[0146] However, α3 ∈ [0, α4], where α4 is the maximum redundancy value, and α4 = min{R u , R d}-R n , R u is the minimum value of the number of pixel points between the center point of the changed pixel points in each column and the upper edge of the first-frame dynamic deformation image, and R d is the minimum value of the number of pixel points between the center point of the changed pixel points in each column and the lower edge of the first-frame dynamic deformation image, and R n is the maximum value of half of the total number of changed pixel points in each column.

[0147] That is to say, the redundancy value cannot be greater than the minimum value of the pixel points that can be redundant from the center point of the corneal area to the upper and lower sides of the first-frame dynamic deformation image, ensuring that there are enough pixel values for redundancy.

[0148] 103. Intercept the target area in each frame of the dynamic deformation image to form a continuous frame image of the target area.

[0149] In step 103, an existing scheme is used to intercept the target area in each frame of the dynamic deformation image, and the target areas are sorted according to the frame number to obtain a continuous frame image of the target area.

[0150] 104. Perform high-resolution processing on the continuous frame image of the target area.

[0151] Although the continuous frame image of the target area collected in step 103 can reduce the area of the invalid area in the image, it cannot guarantee the resolution of the image, nor can it guarantee the data accuracy of the corneal area in the continuous frame image. Through step 104, the resolution of each frame of the image in the continuous frame image can be improved, and the data accuracy of the corneal area in the continuous frame image can be improved.

[0152] In step 104, high-resolution processing is performed on each frame of the continuous frame image of the target area.

[0153] For any frame image i in the continuous frame image of the target area, its high-resolution processing process is obtained based on the processing function, and the function is:

[0154] Among them, i is the frame image identifier in the consecutive frame images of the target area.

[0155] j is the pixel point identifier in the frame image i. For example, the pixel in the upper left corner of the frame image of the target area is used as the first pixel, and then all the pixels in the grayscale image are numbered in sequence from left to right and from top to bottom, that is, the pixel point identifier.

[0156] is the value of the j-th pixel point in the frame image i.

[0157] is the value of the j-th pixel point after high-resolution processing of the frame image i.

[0158] D is the downsampling matrix.

[0159] λ is the penalty parameter.

[0160] is the regularization function, which is used for related processing of adding regularization rules in high-resolution processing.

[0161] H is the blurring matrix.

[0162] Among them, nor is the normalization parameter, (u, v) is the coordinate of the pixel point in the image, and s is the variance estimation value obtained according to the frame image i.

[0163] Such as

[0164] H s is the parameter form of H, W s，ρ is W H，ρ of the parameter form, W H，ρ =(H T H + ρC) -1 H T , ρ is the regularization parameter, and C is the eigenvalue ||ω|| 2 circulant matrix, ω=(ω1, ω2) represents the spatial frequency of Gaussian noise.

[0165] μ is the regularization parameter that changes iteratively in non-descending order, and N is the maximum value of the pixel points in the frame image i.

[0166] 105, evaluate the corneal biomechanical properties of the consecutive frame images after high-resolution processing.

[0167] This step adopts the existing corneal biomechanical property evaluation scheme, such as evaluating the central corneal thickness, deformation amplitude, deformation speed, curvature, etc., which will not be elaborated here.

[0168] This embodiment provides a high-frame-rate imaging method for improving the vertical resolution of an image. During the process of blowing air onto a patient's cornea, dynamic deformation images of the corneal cross-section during the deformation of the patient's cornea are collected; a target region of the dynamic deformation image is determined, where the target region is the smallest region including the cornea; the target region in each frame of the dynamic deformation image is intercepted to form a continuous frame image of the target region; high-resolution processing is performed on the continuous frame image of the target region; and biomechanical characteristics of the cornea are evaluated for the continuous frame image after high-resolution processing. This method determines the target region of the dynamic deformation image during the deformation of the patient's cornea, and this target region is the smallest region including the cornea. By performing high-resolution processing on the continuous frame image of the target region, the data accuracy of the corneal region in the continuous frame image is improved, ensuring the accuracy of the evaluation results of the biomechanical characteristics of the cornea for the continuous frame image after high-resolution processing.

[0169] It should be clear that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present invention.

[0170] It should also be noted that the exemplary embodiments mentioned in the present invention describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above steps, that is, the steps can be executed in the order mentioned in the embodiments, or different from the order in the embodiments, or several steps can be executed simultaneously.

[0171] Finally, it should be noted that the above-described embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high frame rate imaging method for improving the vertical resolution of an image, characterized in that: The method comprises: During the process of blowing air toward the patient's cornea, a dynamic deformation image of the corneal cross section during the deformation of the patient's cornea is collected; wherein the dynamic deformation image is a grayscale image; Determining a target area of ​​the dynamically deformable image, wherein the target area is a minimum area including the cornea; intercepting a target area in each frame of the dynamically deformed image to form a continuous frame image of the target area; Performing high-resolution processing on the continuous frame images of the target area; The biomechanical properties of the cornea were evaluated on the continuous frame images after high-resolution processing.

2. The method according to claim 1, characterized in that: The determining the target area of ​​the dynamically deformed image includes: Get the first frame of dynamic deformation image; Acquire the marked edge of the cornea in the first frame of the dynamically deformed image; Identifying the edge of the cornea in the first frame of the dynamically deformed image; Forming a marked pixel set M according to the marked edge, and forming a recognized pixel set D according to the recognized edge; Difference pixel set I = (D∪M)-(D∩M); Determine the change threshold according to the gray value of each element in M, D and I; A target area of ​​the dynamically deformed image is determined according to the change threshold.

3. The method according to claim 2, characterized in that Determining the change threshold according to the gray value of each element in M, D and I includes: Determine the first mean μ1 and the first standard deviation σ1 of the grayscale values ​​of all elements in M; Determine the second mean μ2 and the second standard deviation σ2 of the gray values ​​of all elements in D; If I is an empty set, the change threshold is determined as If I is not an empty set, then determine the third mean μ3 and the third standard deviation σ3 of the grayscale values ​​of all elements in I; determine the change threshold value according to the relationship between σ1 and σ2, as well as μ3 and σ3.

4. The method according to claim 3, characterized in that The change threshold is determined according to the relationship between σ1 and σ2, and μ3 and σ3. include: If σ1=σ2, then the change threshold is determined as If σ1≠σ2, then the change threshold is determined as 5. The method according to claim 2, characterized in that: The step of determining the target area of ​​the dynamically deformed image according to the change threshold comprises: Classify each element in M ​​by column to form a first labeled pixel point set D1 of each column; Determine the minimum and maximum number of rows of all pixels in D1 in each column; The pixel points with the smallest number of rows in each column that are one row above and the pixel points with the largest number of rows that are one row below are determined as added pixel points, and the second annotated pixel point set D2 of each column is formed by the added pixel points and D1; Determine the changed pixel points of each column according to D1, D2 and the change threshold of each column; The smallest rectangle containing all the changed pixels in the columns is determined as the target area.

6. The method according to claim 5, characterized in that The step of determining the changed pixel points of each column according to D1, D2 and the change threshold of each column includes: For any column, determine a fourth mean μ4 of the grayscale values ​​of all elements in D1 of the column; Determine the fifth mean μ5 of the gray values ​​of all elements in D2 of any column; If μ5>μ4 and Then all the pixels in D1 of any column are determined to be the changed pixels of any column; wherein α1 is a preset grayscale threshold, α1∈(0,1); If μ5≤μ4, or, μ5>μ4 but The changed pixel point is determined according to D1, D2 of any column and the change threshold.

7. The method according to claim 6, characterized in that The determining the changed pixel point according to D1 and D2 of any column and the change threshold comprises: Determine the change set D3 = D2 - D1; Determine a sixth mean μ6 and a sixth standard deviation σ6 of the gray values ​​of all elements in D3; like Then all the pixels in D1 of any column are determined to be the changed pixels in any column; wherein α2 is the preset change accuracy, Δ is the change threshold, μ1 is the first mean of the grayscale values ​​of all elements in M, and σ1 is the first standard deviation of the grayscale values ​​of all elements in M; like Then all the pixel points in D1 of any column are updated to all the pixel points in D2 of any column, and the steps of determining the pixel points in the upper row of the minimum number of rows in each column and the pixel points in the lower row of the maximum number of rows as added pixel points are repeated, and forming the second labeled pixel point set D2 of each column by the added pixel points and D1, and subsequent steps.

8. The method according to claim 5, characterized in that The step of determining the minimum rectangle containing all columns of changed pixels as the target area includes: The changed pixels of each column are expanded upward and downward by α3 pixels respectively; wherein α3 is a preset redundancy value, which is determined according to the patient attributes, α3∈[0,α4], α4 is the maximum redundancy value, α4=min{R u ,R d }-R n , R u is the minimum value of the number of pixels between the center point of each column of changed pixels and the upper edge of the first frame of the dynamically deformed image, R d is the minimum value of the number of pixels between the center point of each column of changed pixels and the lower edge of the first frame of the dynamic deformation image, R n The maximum value of half of the total number of changed pixels in each column; The smallest rectangle containing the enlarged changed pixels in all columns is determined as the target area.

9. The method according to claim 1, characterized in that: The step of performing high-resolution processing on the continuous frame images of the target area comprises: For any frame image in the continuous frame images of the target area, the following function is used Perform high-resolution processing; Wherein, i is the frame image identifier in the continuous frame images of the target area, j is the pixel point identifier in the frame image i, is the value of the jth pixel in frame image i, is the value of the jth pixel after high-resolution processing of frame image i, D is the downsampling matrix, H is the blur matrix, λ is the penalty parameter, is a regular function.

10. The method according to claim 9, characterized in that Said Among them, nor is the normalization parameter, (u, v) is the coordinate of the pixel in the image, and s is the variance estimate obtained based on the frame image i.

Images