Fast phase estimation method of line-scan structured light and super-resolution fast reconstruction method

Abstract

The application discloses a kind of line scanning structured light fast phase estimation method, super-resolution fast reconstruction method, phase estimation method includes S1, obtains original strip image, using two digital gratings with known phase difference to modulate them, obtain the phase modulation strip chart with same phase difference;S2, threshold segmentation is carried out to phase modulation strip chart, and signal region is obtained;S3, according to the ratio of the intensity average value of two signal regions and the ratio of the cosine intensity of two modulation phases, an equation is constructed;S4, equation is solved, and modulation phase is obtained.The intensity average value ratio after threshold segmentation of the phase modulation strip chart with known phase difference is used to quickly solve modulation phase, without repeated iteration, and the method is used in the super-resolution reconstruction method of the spatial domain solving half spectrum component and carrying out real part reconstruction of the application, and the reconstruction speed is greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of optical image reconstruction, and more specifically, relates to a fast phase estimation method and a super-resolution fast reconstruction method for line-scan structured light. Background Technology

[0002] Traditional wide-field super-resolution structured light microscopy typically uses cosine fringe illumination, leveraging the moiré fringe effect to achieve a 2x resolution improvement. However, because image contrast in wide-field imaging is significantly affected by background signals, it is generally used for imaging thin samples with limited background signal. Virtual structured detection (VSD) shifts the traditional structured illumination modulation from the illumination end to the detection end, simplifying the illumination structure and eliminating the need for multiple scans. A narrow digital grating can block a portion of the detection signal, primarily containing background, achieving physical optical tomography. By introducing a virtual slit to block background scattering, VSD further extends the application of super-resolution structured light to three-dimensional thick tissue samples.

[0003] Existing line-scan super-resolution imaging based on VSD technology employs a linear array detector, capturing a two-dimensional raw frame at each scan position x in the line-scan imaging process. This detection method is called off-axis array detection. The reconstruction process involves first performing digital structure modulation on the raw frame to reconstruct an equivalent wide-field structure modulation raw image. Then, traditional wide-field super-resolution algorithms are used for reconstruction, including:

[0004] S1. The sample is modulated three times using illumination structures with three different phases, resulting in three original modulation images. These three original modulation images, when transformed to the frequency domain, form a system of linear equations, the coefficients of which are determined by the modulation phase. S2. The three spectral components are separated by solving the system of linear equations. S3, using a Wiener filter to enhance the high-frequency signal; S4, shifting the three spectra to their correct positions before the frequency shift; S5, adding and fusing the three spectral components to obtain the extended super-resolution spectrum; S6, performing an inverse Fourier transform to obtain the spatial domain super-resolution result. In S4, the shift distance is the spatial frequency of the illumination fringes, and this shift distance is often not an integer number of pixels, making direct shifting in the frequency domain impossible. According to the Fourier shift theorem, frequency domain shifts can be converted into spatial domain phase shifts, and spatial domain phase shifts are not limited by pixels. Therefore, higher-order sub-spectrums are generally used... After transforming to the spatial domain, multiply by the corresponding phase, and then transform back to the frequency domain to achieve frequency domain shift.

[0005] To correctly solve for the spectral components, accurate coefficients of the equations are required, meaning accurate estimation of the modulation phase is necessary. In an online scanning imaging optical path, ideally, the line spot is located at the center of the virtual slit, so the modulation phase is the same as the phase of the digital grating and does not require estimation. However, due to issues with optical path adjustment accuracy and system stability, there is a random deviation between the line spot and the center of the virtual slit in the imaging optical path, causing a change in the phase of the equivalent wide-field structure modulation. Specifically:

[0006] Let d be the offset distance between the line spot and the center of the virtual slit. The position of the line spot represents the sampling point position of the sample, therefore the sample is also offset by a distance d. The image distribution of the equivalent wide-field structure modulation on the sample surface is then:

[0007]

[0008] in, The phase of the digital grating used in virtual structure probe modulation is represented by x, the sample coordinates by S, and PSF. il Let represent the illumination point spread function, represent the line spot, and k0 represent the spatial frequency of the stripe. The line spot offset can be relatively viewed as the offset in the opposite direction of the digital grating. Substituting x′=xd into equation (1), the offset is converted to the digital grating:

[0009]

[0010] It can be seen that the modulation phase of the equivalent wide-field structure modulation Phase with digital grating The relationship between them is:

[0011]

[0012] In actual optical paths, the offset distance d is unknown, which leads to It becomes an unknown parameter.

[0013] In existing line-scan super-resolution structured light reconstruction algorithms, the modulation phase is generally estimated using the cross-correlation (CC) method. Right now:

[0014]

[0015] Theoretically when With modulation phase The cross-correlation value is maximized when the signals are consistent. Therefore, the phase of the cosine wave is continuously changed during the iteration process. The cosine phase corresponding to the maximum cross-correlation value is the modulation phase. This method obtains relatively accurate phases, but because it requires iteration, the calculation process is very time-consuming and cannot achieve rapid reconstruction.

[0016] Furthermore, it can be observed that existing super-resolution reconstruction procedures require first reconstructing the original frames from off-axis array probes at different scanning positions x using different digital grating phases to obtain the equivalent wide-field structured illumination structure modulation strip map. Since the modulation phase differs at different scanning positions x, serial calculations are needed along different scanning positions for the original strip image, which is slow. After obtaining the equivalent wide-field structured illumination structure modulation strip map, traditional wide-field super-resolution reconstruction is then performed. However, the traditional wide-field super-resolution reconstruction process requires repeated Fourier transforms and calculations using three orders of spectral components, making it extremely complex and time-consuming. These factors contribute to the complexity and time-consuming nature of existing line-scan super-resolution structured light reconstruction procedures, placing significant pressure on the processing of large sample data. Summary of the Invention

[0017] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a fast phase estimation method and a fast super-resolution reconstruction method for line-scan structured light. The purpose is to solve the problems of complex reconstruction process, large computational load, and time-consuming modulation phase estimation in line-scan super-resolution structured light imaging technology based on virtual structure detection.

[0018] To achieve the above objectives, according to one aspect of the present invention, a fast phase estimation method for line-scan super-resolution structured light is provided, comprising the following steps:

[0019] S1, Obtain the original stripe image L1()~L n (), n≥2 and are integers, x represents the scanning position, n represents the nth row of the off-axis array detection, and the known phase difference is used respectively. Two constant phase digital gratings and For the original strip image L1()~L n () is modulated to obtain a phase difference of (). Phase modulation stripe and and They represent and phase, and They represent and The modulation phase,

[0020] S2, regarding the phase modulation stripe diagram and Threshold segmentation is performed to obtain the signal region. and

[0021] S3, construct an equation based on the ratio of the average intensity of the two signal regions and the ratio of the cosine intensity of the two modulation phases;

[0022] S4, Solve the equation to obtain the modulation phase.

[0023] The above technical solution uses two digital gratings with a known phase difference to modulate the original strip image, thereby obtaining a phase-modulated strip image with a known modulation phase difference. Then, by using the ratio of the average intensity after thresholding and the ratio of the cosine intensity of the two modulation phases, an equation can be constructed, which can quickly solve for the modulation phase without repeated iterations. Thresholding can extract the effective signal and eliminate estimation errors, ultimately achieving fast and accurate phase estimation.

[0024] Preferably, in S1:

[0025]

[0026] Where sgn(·) represents the sign function, n=1:υT, v represents the number of digital modulation periods contained in the virtual slit, and T represents the period of the digital grating. Ensure that the virtual slits are symmetrically distributed around the lighting points.

[0027] Preferably, in S1:

[0028]

[0029] Preferably, the specific process of threshold segmentation in S2 is as follows:

[0030]

[0031]

[0032] Where Thh is the segmentation threshold, Where Max and Min represent the original strip images L1() to L1() respectively. n () represents the strongest and weakest pixel values ​​in the average image, where TempOffset is a constant used to adjust the threshold size.

[0033] Preferably, the equation constructed in S3 is as follows:

[0034]

[0035] Mean represents the average value.

[0036] Preferably, the modulation phase solved in S4 Specifically:

[0037]

[0038] This invention also proposes a super-resolution fast reconstruction method for line-scan structured light, comprising the following steps:

[0039] Step A, obtain the original strip images L1()~L n (), n≥2 and are integers, x represents the scanning position, n represents the nth row of the off-axis array detection;

[0040] Step B, using the known phase difference respectively Two digital gratings and Modulating the original strip image yields a phase difference of... Structural modulation strip diagram and x represents the scan position. and They represent and phase, and The modulation phases are respectively and

[0041] Step C: Construct two spatial domain linear equation sets for phase modulation strip diagrams, solve for the spatial domain distribution of the +1 level spectrum or the -1 level spectrum, and then perform phase shifting in the spatial domain to obtain the spatial domain distribution of the higher-order spectrum after frequency shifting.

[0042] Step D: Take the spatial domain distribution of a higher-order sub-spectrum from step C and perform real part extraction. Multiply the obtained real part by 2 to obtain the super-resolution result.

[0043] It also includes step E, which uses known phase differences as... Two constant phase digital gratings and For the original strip image L1()~L n () is modulated to obtain a phase difference of (). Phase modulation stripe and and They represent and The modulation phase, The phase modulation stripe and Threshold segmentation is performed to obtain the signal region. and An equation is constructed based on the ratio of the average intensity of the two signal regions and the ratio of the cosine intensity of the two modulation phases. Solving the equation yields the modulation phase. Step E is located after step A.

[0044] Preferably, the real part extraction process in step D specifically involves: directly extracting the real part of the spatial domain distribution of the higher-order sub-spectrum after frequency shift; the obtained super-resolution result is a high-speed reconstruction result.

[0045] Preferably, the real part extraction process in step D specifically involves: transforming the frequency-shifted higher-order spectrum spatial domain distribution to the frequency domain, then using a mask to extract the standard half-spectrum, and then transforming it back to the spatial domain to extract the real part; the obtained super-resolution result is a high-contrast reconstruction result.

[0046] Preferably, the real part extraction process in step D specifically involves: transforming the frequency-shifted higher-order sub-spectral spatial domain distribution to the frequency domain, then using a mask to extract the standard half-spectrum and performing Wiener filtering, and then inversely transforming it back to the spatial domain to extract the real part; the obtained super-resolution result is a high-fidelity reconstruction result.

[0047] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0048] (1) This invention uses the ratio of the average intensity of the signal region and the ratio of the cosine intensity of the two modulation phases to construct the equation, which eliminates the need for repeated iterations and is faster.

[0049] (2) Thresholding the phase modulation strip map before using it to construct the solution equation helps to eliminate estimation errors and can achieve both fast and accurate phase estimation.

[0050] (3) During super-resolution reconstruction, rapid spectral separation in the spatial domain avoids repeated Fourier transforms.

[0051] (4) Fast super-resolution reconstruction of the real part can be performed using only one of the +1 level spectrum or the -1 level spectrum, reducing the number of spectra required for reconstruction;

[0052] (5) It has three reconstruction modes: high speed, high contrast and high fidelity, which can respectively achieve the effects of high-speed reconstruction, obtaining high-contrast super-resolution images and obtaining super-resolution images with high fidelity.

[0053] (6) Combining fast super-resolution reconstruction with fast phase estimation, the reconstruction speed is improved from multiple aspects. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the phase estimation method of the present invention;

[0055] Figure 2 (a) is the phase modulation stripe plot obtained by using virtual structure detection modulation when the background intensity is 10;

[0056] Figure 2 (b) is a modulation comparison diagram under different background intensities and a phase value data diagram estimated by the phase estimation method of the present invention;

[0057] Figure 2 Figure (c) is a comparison of the calculation time of the phase estimation method of the present invention and the traditional CC algorithm for phase estimation. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0059] like Figure 1 As shown, this invention proposes a fast phase estimation method for line-scan structured light, comprising the following steps:

[0060] S1, Obtain the original stripe image L1()~L n (), n≥2 and are integers, x represents the scanning position, n represents the nth row of the off-axis array detection, and the known phase difference is used respectively. Two constant phase digital gratings and For the original strip image L1()~L n () is modulated to obtain a phase difference of (). Phase modulation stripe and and They represent and phase, and They represent and The modulation phase,

[0061] S2, regarding the phase modulation stripe diagram and Threshold segmentation is performed to obtain the signal region. and

[0062] S3, construct an equation based on the ratio of the average intensity of the two signal regions and the ratio of the cosine intensity of the two modulation phases;

[0063] S4, Solve the equation to obtain the modulation phase.

[0064] In S1 of this invention, two constant-phase digital gratings with a known phase difference are constructed to modulate the original strip image, and the phase difference between the two phase-modulated strip images obtained is also... Threshold segmentation is then used to extract the effective signal, eliminating background blur and noise interference. Averaging eliminates estimation errors, and the phase is determined by the ratio of the cosine intensity of the two modulation phases. It eliminates the need for iteration, accelerating phase estimation while maintaining high accuracy. It enables fast and accurate phase estimation, improving the reconstruction speed of line-scan super-resolution structured light based on virtual structure detection.

[0065] Specifically, in S1:

[0066]

[0067] in, and Let represent the phases of the two digital gratings respectively, sgn(·) represent the sign function, n = 1:υT, v represent the number of digital modulation periods contained in the virtual slit, and T represent the period of the digital grating. Ensure that the virtual slits are symmetrically distributed around the lighting points.

[0068] In this invention, the original strip image is modulated using a digital grating to achieve equivalent wide-field structure modulation, thereby obtaining a phase-modulated strip image. Specifically, the two phase-modulated strip images are as follows:

[0069]

[0070] After obtaining the phase modulation stripe map, it is thresholded to extract the signal region, thus preventing the unmodulated background signal from affecting the subsequent phase estimation calculation. Specifically, the thresholding process is as follows:

[0071]

[0072]

[0073] Where Thh is the segmentation threshold, Where Max and Min represent the original strip images L1(x) to L1(x) respectively. n(x) represents the strongest and weakest pixel values ​​in the average image. TempOffset is a constant used to adjust the threshold size, which can be selected according to the actual situation, and is usually set to -0.1. This can avoid setting the threshold too high due to a few bright spots in biological samples, resulting in too little signal area being captured.

[0074] The equations constructed in S3 are as follows:

[0075]

[0076] Mean represents the average value.

[0077] This invention provides a line-scan super-resolution reconstruction algorithm for virtual structure detection. A digital grating, convolved with a probe PSF, is applied to the sample to obtain a cosine-modulated sample distribution. In step S1, the phase-modulated stripe map after digital grating modulation is further processed by an illumination PSF to obtain the image distribution I of the sample surface:

[0078]

[0079] Where S represents the sample, PSF il Let represent the illumination point spread function, and represent the line spot. From this formula, it can be seen that the ratio of the average intensity values ​​of the signal region is also the phase modulation stripe pattern. and The ratio of the cosine intensities of the two modulation phases is used to construct the above equation.

[0080] Finally, the modulation phase can be obtained by solving the above equations. Specifically:

[0081]

[0082] In S1, to simplify the estimation of the modulation phase, the known phase difference used in S1 is further... It can be set to π / 2, in the digital grating phase difference When the value is π / 2, the above equation simplifies to:

[0083]

[0084] Based on the simplified results, the modulation phase can be estimated more quickly. The accuracy and speed of the fast phase estimation method of this invention are verified below.

[0085] Using the spoke pattern with intensity normalized to 1 as a sample, a flat background was added 0.5 μm below the sample to simulate the degradation of fringe contrast by the background signal in 3D thick sample imaging. Shot noise was added to the raw frames detected by the off-axis array. A 31-cycle digital grating modulation was used to include as much background and noise as possible. The illumination spot was positioned at the exact center of the off-axis array detector, corresponding to an initial phase accuracy of zero.

[0086] The background plate intensity is set to five values ​​ranging from 0 to 10, with an interval of 2.5. The phase modulation stripe pattern of the virtual structure probe modulation when the background intensity is 10 is shown below. Figure 2 As shown in (a), the background has a significant impact, and modulation stripes are almost invisible. The modulation contrast at five different background intensities is shown below. Figure 2 The solid line in (b) shows that the modulation contrast is obtained by calculating the ratio of the spectral intensity of the overlapping region of the modulated high-frequency component and the unmodulated zero-frequency component near the low frequency; the phase estimated using the method of this invention under 5 different background intensities is as follows. Figure 2 As shown by the dashed line in (b), it can be observed that the phase value estimated by the present invention always fluctuates around the correct zero value, and the present invention can still accurately estimate the phase when the modulation contrast drops below 0.02. Figure 2 Figure (c) shows the computation time of the conventional CC algorithm and the method (IR) of this invention, respectively. It can be seen that this invention reduces the computation time for phase estimation from 0.15 seconds for the CC algorithm to 0.0029 seconds, an improvement of approximately 50 times. These results demonstrate that the phase estimation method of this invention has high accuracy and speed.

[0087] This invention also proposes a super-resolution fast reconstruction method for line-scan structured light, characterized by comprising the following steps:

[0088] Step A, obtain the original strip images L1()~L n (), n≥2 and are integers, x represents the scanning position, n represents the nth row of the off-axis array detection;

[0089] Step B, using the known phase difference respectively Two digital gratings and Modulating the original strip image yields a phase difference of... Structural modulation strip diagram and x represents the scan position. and They represent and phase, and The modulation phases are respectively and

[0090] Step C: Construct two spatial domain linear equation sets for phase modulation strip diagrams, solve for the spatial domain distribution of the +1 level spectrum or the -1 level spectrum, and then perform phase shifting in the spatial domain to obtain the spatial domain distribution of the higher-order spectrum after frequency shifting.

[0091] Step D: Take the spatial domain distribution of a higher-order sub-spectrum from step C and perform real part extraction. Multiply the obtained real part by 2 to obtain the super-resolution result.

[0092] It also includes step E, which uses known phase differences as... Two constant phase digital gratings and For the original strip image L1()~L n () is modulated to obtain a phase difference of (). Phase modulation stripe and and They represent and The modulation phase, The phase modulation stripe and Threshold segmentation is performed to obtain the signal region. and An equation is constructed based on the ratio of the average intensity of the two signal regions and the ratio of the cosine intensity of the two modulation phases. Solving the equation yields the modulation phase. Step E is located after step A.

[0093] Existing techniques involve constructing a linear system of equations in the frequency domain, solving for three frequency components, then transforming to the spatial domain for phase shifting, and finally transforming back to the frequency domain. This shifts the higher-order spectral components back to their positions before modulation by cosine fringe structured light illumination. This repeated Fourier transform is one reason for its time-consuming nature. Since the frequency domain equations are linear, performing an inverse Fourier transform on both sides of the equations still results in the equations remaining valid, the coefficient matrix unchanged, and the variables becoming the spatial domain distribution of the spectral components. Therefore, this invention uses two phase differences as... Structural modulation strip diagram and Using the spatial domain distribution as a variable, a system of linear equations in the spatial domain is listed with the same coefficient matrix. The spatial domain distribution of the +1 level spectrum or the -1 level spectrum is directly solved. Then, phase shift is directly performed in the spatial domain, which is equivalent to frequency shift in the frequency domain, avoiding repeated Fourier transforms.

[0094] Secondly, existing technologies fuse three spectral components, two of which are higher-order spectral components. However, since the captured image is represented by real numbers, and the spectrum of real numbers has conjugate symmetry, only half of the spectrum is effective. Using two higher-order spectra for reconstruction during the fusion process is wasteful and increases the time for spectrum processing and fusion. In contrast, this invention only needs to solve for one of the higher-order spectral components (i.e., only solve for the spatial domain distribution of the +1 or -1 order spectrum), and then perform real part processing on its spatial domain distribution for reconstruction, reducing the number of spectra required for reconstruction and the reconstruction time.

[0095] Furthermore, this invention also utilizes the aforementioned fast phase estimation method, which eliminates the need for repeated iterations and achieves faster speed. In summary, this invention improves reconstruction speed in three ways: avoiding repeated Fourier transforms, reducing the number of spectra required for reconstruction, and accelerating the estimation of the modulation phase.

[0096] Specifically, in step A, x = 1:N-1, where N represents the total number of pixels in the sample.

[0097] In step B:

[0098]

[0099]

[0100] n = 1:υT, where v represents the number of digital modulation periods contained in the virtual slit, and T represents the period of the digital grating. Ensure that the virtual slits are symmetrically distributed around the lighting points.

[0101] Furthermore, This makes the calculation in step C simpler, further simplifying the solution process for the spatial domain distribution of higher-order spectra and the estimation of the modulation phase. The specific reasons are explained below. During the solution process, since digital gratings can modulate negative values ​​into the image, compared to traditional structured light modulation, the image does not have a zero-order spectrum; the spectrum only has ±1 orders superimposed. Therefore, the spatial domain linear equations constructed in step C are:

[0102]

[0103] Then, the spatial domain distribution of the +1 order spectrum can be obtained: Spatial domain distribution of the -1 level spectrum:

[0104] The +1 level spectrum is frequency shifted, which is based on the above spatial domain distribution multiplied by... The -1 level spectrum is frequency shifted by multiplying the spatial domain distribution described above by [missing information]. k0 = N / T, where k represents the modulation frequency and k represents the frequency coordinate.

[0105] The spatial domain distribution of the obtained higher-order sub-spectrum is as follows:

[0106]

[0107] Substituting the digital grating modulation from step B into the above equation yields:

[0108]

[0109] The above equation combines virtual structured light reconstruction and structured light super-resolution reconstruction into one. However, since different scanning positions x have different digital grating modulations and different phase shift coefficients (i.e., different calculation coefficients), the above equation needs to be calculated serially along different scanning positions x, making the process still relatively cumbersome. When the phase interval... At this time, the digital grating modulation and phase shift coefficients can be simplified, and the calculated coefficients at different scanning positions x become consistent, which allows for the calculation of the original strip image L. n () Parallel computation of data at different x locations significantly reduces the computational difficulty of steps C, such as constructing, solving, and phase shifting spatial domain linear equations, thereby further improving the reconstruction speed.

[0110] Specifically, The spatial domain distribution of the higher-order sub-spectrum obtained in step C is as follows:

[0111]

[0112] Based on this, taking a digital grating period T of 4, commonly used in line-scan virtual structured light systems, as an example, when the virtual slit contains one digital modulation period, that is... At this point, the spatial domain distribution of the higher-order sub-spectrum is as follows:

[0113]

[0114] It can be seen that the formula no longer includes the position information x, so even if the scanning position x changes, it is no longer necessary to calculate line by line.

[0115] Since this invention can directly use the real part of the spatial domain distribution of a higher-order spectrum for reconstruction, the reconstruction process simply involves multiplying this real part by 2 to obtain the super-resolution reconstructed image. The frequency-shifted higher-order spectrum used in the reconstruction includes a standard half-spectrum and a small amount of opposing spectrum on the other side of the coordinate axis, and is modulated by the frequency-shifted OTF. The standard half-spectrum refers to the spectrum on the side containing super-resolution information at the spectral boundary and half of the spectrum on the boundary line. The boundary line is the perpendicular line from the origin to the line connecting the origin and the modulation frequency point on the spectrum. Therefore, in step D, three different methods can be used to process the spatial domain distribution of the higher-order spectrum obtained in step C by taking the real part and obtaining the super-resolution result.

[0116] In some embodiments, a high-speed reconstruction mode is adopted, ignoring the opposing spectrum and the modulation of the frequency-shifted OTF, and directly taking the real part in the spatial domain of the higher-order sub-spectrum after frequency shift, which can be reconstructed more quickly, and the obtained super-resolution result is a high-speed reconstruction result. Regardless of whether the higher-order spectrum obtained in step D is a +1 level spectrum or a -1 level spectrum: D HS (x) represents the high-speed reconstruction result.

[0117] In other embodiments, a high-contrast reconstruction mode is employed. The spatial domain distribution of the frequency-shifted higher-order sub-spectrum is transformed to the frequency domain, a standard half-spectrum is extracted using a mask, and then inversely transformed back to the spatial domain to extract the real part. The mask extracts a standard half-spectrum located on one side of the coordinate axis that contains the super-resolution spectrum, which can remove the opposing spectrum. At the same time, the frequency-shifted OTF has a low-frequency suppression effect, resulting in reconstruction with lower noise and higher contrast. The obtained super-resolution result is a high-contrast reconstruction result. When the higher-order spectrum is the +1 level spectrum:

[0118] When the higher-order spectrum is the -1 level spectrum:

[0119] D HC (x) represents the high-contrast reconstruction result, and Re represents taking the real part. This refers to a photomask.

[0120] In some other embodiments, a high-fidelity reconstruction mode is employed. The spatial domain distribution of the frequency-shifted higher-order sub-spectrum is transformed to the frequency domain. A standard half-spectrum is truncated using a mask and subjected to Wiener filtering, then transformed back to the spatial domain to obtain the real part. This method can remove the modulation of the frequency-shifted OTF after removing the opposing spectrum, obtaining the true half-spectrum. The reconstruction result has high fidelity, and the resulting super-resolution result is a high-fidelity reconstruction. When the higher-order spectrum is the +1 level spectrum: When the higher-order spectrum is the -1 level spectrum:

[0121]

[0122] D HF (x) represents the high-fidelity reconstruction result, and Re represents taking the real part. This indicates a mask, and Wiener() indicates Wiener filtering.

[0123] More specifically, for line-scan structured light processing, when the virtual modulation is along the scanning direction x, k0 is a point on the horizontal axis in the two-dimensional frequency domain diagram. The mask used for high-speed reconstruction and high-contrast reconstruction can be:

[0124] When rebuilding at +1 level:

[0125] When rebuilding at level -1:

[0126] Where, k x Represents the horizontal coordinate in the two-dimensional frequency domain.

[0127] This method employs three different modes during reconstruction: a high-speed reconstruction mode, where the entire process is performed in the spatial domain, resulting in rapid reconstruction; a high-contrast mode, utilizing the natural modulation of the frequency-shifted OTF (Optical Transfer Function) to obtain a high-contrast super-resolution image; and a high-fidelity mode, using Wiener filtering to remove the influence of the system's OTF, yielding a super-resolution image with high fidelity. This fast reconstruction method of the real part of the half-spectrum of super-resolution structured light improves the efficiency of structured light super-resolution image reconstruction, saving time and computational resources, and greatly contributing to further expanding the application areas of super-resolution structured light imaging.

[0128] Specifically, it can be observed that all three different reconstruction results are related to the modulation phase. As described in formula (3) in the background art, in practical cases Since these are unknown parameters, step E employs the same steps as the previously disclosed fast phase estimation method to estimate the modulation phase. Step E can be performed at any stage after step A, and the modulation phase is estimated quickly and accurately through step E. This can further improve the speed of this reconstruction method. Furthermore, another set of initial phases is used in step E, which is also... Furthermore, a digital grating with constant phase is used to modulate the original strip image in step A, resulting in a phase-modulated strip image. The phase difference of this set of digital gratings is... and The values ​​can be the same or different. According to the previous description in the instruction manual, when they are all π / 2, the calculation process of steps C, D and E can be simplified to the minimum.

[0129] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fast phase estimation method for line-scan structured light, characterized in that, Includes the following steps: S1, Obtain the original strip image ~ , And it is an integer. Indicates the scan position. The first off-axis array detection Line, respectively using the known phase difference as Two constant phase digital gratings and For the original strip image ~ Modulation is performed, and the phase difference is also obtained. Phase modulation stripe and , and They represent and phase, and They represent and The modulation phase, ; S2, regarding the phase modulation stripe diagram and Threshold segmentation is performed to obtain the signal region. and ; S3, an equation is constructed based on the ratio of the average intensity of the two signal regions and the ratio of the cosine intensity of the two modulation phases; the specific equation is as follows: ； in, This indicates calculating the average value; S4, Solve the equation to obtain the modulation phase. Specifically: ； when When, let the modulation phase Revised to .

2. The method according to claim 1, characterized in that, In S1: ； in, Represents a symbolic function. , This indicates the number of digital modulation cycles contained in the virtual slit. The period of the digital grating is indicated. Ensure that the virtual slits are symmetrically distributed around the lighting points.

3. The method according to claim 2, characterized in that, In S1: 。 4. The method according to claim 1, characterized in that, The specific process of threshold segmentation in S2 is as follows: ， ； in, The threshold for segmentation, ;in, and These represent the original strip images. ~ The strongest and weakest pixel values ​​in the average image. It is a constant used to adjust the threshold value.

5. A super-resolution fast reconstruction method using line-scan structured light, characterized in that, Includes the following steps: Step A: Obtain the original strip image ~ , And it is an integer. Indicates the scan position. The first off-axis array detection OK; Step B, using the known phase difference respectively Two digital gratings and Modulating the original strip image yields a phase difference of... Structural modulation strip diagram and , Indicates the scan position. and They represent and phase, and The modulation phases are respectively and ; Step C: Construct two spatial domain linear equation sets for phase modulation strip diagrams, solve for the spatial domain distribution of the +1 level spectrum or the -1 level spectrum, and then perform phase shifting in the spatial domain to obtain the spatial domain distribution of the higher-order spectrum after frequency shifting. Step D: Take the spatial domain distribution of a higher-order sub-spectrum from step C and perform real part extraction. Multiply the obtained real part by 2 to obtain the super-resolution result. It also includes step E, which uses known phase differences as... Two constant phase digital gratings and For the original strip image ~ Modulation is performed, and the phase difference is also obtained. Phase modulation stripe and , and They represent and The modulation phase, For the phase modulation strip pattern and Threshold segmentation is performed to obtain the signal region. and An equation is constructed based on the ratio of the average intensity of the two signal regions and the ratio of the cosine intensity of the two modulation phases. The equation is then solved to obtain the modulation phase. Step E is located after step A.

6. The method according to claim 5, characterized in that, The real part extraction process in step D specifically involves directly extracting the real part of the spatial domain distribution of the higher-order spectrum after frequency shift; the obtained super-resolution result is the high-speed reconstruction result.

7. The method according to claim 5, characterized in that, The real part extraction process in step D specifically involves: transforming the frequency-shifted higher-order spectrum spatial domain distribution to the frequency domain, then using a mask to extract the standard half-spectrum, and then inversely transforming it back to the spatial domain to extract the real part; the obtained super-resolution result is a high-contrast reconstruction result.

8. The method according to claim 5, characterized in that, The real part extraction process in step D specifically involves: transforming the frequency-shifted higher-order spectrum spatial domain distribution to the frequency domain, then using a mask to extract the standard half-spectrum and performing Wiener filtering, and then inversely transforming it back to the spatial domain to extract the real part; the obtained super-resolution result is a high-fidelity reconstruction result.

Images