An interferometric image processing method and system

Abstract

The application relates to an interferometric image processing method and system. The method comprises: acquiring N frames of interferometric images, the N frames of interferometric images being interferometric images collected for a same measured object, and the N frames of interferometric images having sub-pixel offsets therebetween, N being an integer greater than 1; mapping the N frames of interferometric images to a high-resolution grid and performing fusion according to the sub-pixel offset amounts between the N frames of interferometric images, to obtain a high-resolution interferometric image, wherein the resolution of the high-resolution grid is higher than the resolution of the N frames of interferometric images; and performing phase extraction on the high-resolution interferometric image to obtain a phase distribution of the measured object. Thus, by performing high-resolution grid fusion on multiple frames of interferometric images having sub-pixel offsets therebetween, the spatial resolution of the interferometric images can be improved, and a phase distribution with a higher spatial resolution can be obtained.

Description

Technical Field

[0001] This application relates to the field of optical measurement technology, and in particular to an interferometric image processing method and system. Background Technology

[0002] In semiconductor manufacturing, precision optical processing, and other fields, interferometers (such as the Fizeau interferometer) are widely used to measure the surface morphology of wafers and the surface shape of optical components. By analyzing the phase distribution of the interference fringe pattern, the interferometer can extract information corresponding to the surface shape of the measured surface. However, due to the limitation of the interferometer's image sensor pixel size, the spatial frequencies that can be resolved in a single frame of interference image are limited. Therefore, it is necessary to improve the spatial resolution of the interference image. Summary of the Invention

[0003] This application provides an interferometric image processing method, system, electronic device, and storage medium, which can improve the spatial resolution of interferometric images.

[0004] In a first aspect, embodiments of this application provide an interferometric image processing method, comprising: acquiring N frames of interferometric images, wherein the N frames of interferometric images are interferometric images acquired for the same object under test, and the N frames of interferometric images have sub-pixel offsets, where N is an integer greater than 1; mapping the N frames of interferometric images to a high-resolution grid and fusing them according to the sub-pixel offsets between the N frames of interferometric images to obtain a high-resolution interferometric image, wherein the resolution of the high-resolution grid is higher than the resolution of the N frames of interferometric images; and performing phase extraction on the high-resolution interferometric image to obtain the phase distribution of the object under test. Thus, by mapping N frames of interferometric images with sub-pixel offsets to a higher-resolution grid and fusing them, the spatial resolution of the fused interferometric image is no longer limited by the pixel size of a single frame of interferometric image. Furthermore, by performing phase extraction based on the high-resolution interferometric image, a phase distribution with higher spatial resolution can be obtained.

[0005] In one optional implementation, mapping the N frames of interferometric images to a high-resolution grid and fusing them includes: for each pixel among multiple pixels in the N frames of interferometric images, determining a fusion weight for that pixel based on a spatial distance weight, a fringe contrast weight, and a phase consistency weight; and performing weighted fusion of the multiple pixels mapped to the high-resolution grid based on the fusion weight of each pixel among the multiple pixels; wherein the phase consistency weight is used to characterize the degree of consistency of the phase at the pixel position among the N frames of interferometric images. Thus, the fusion weight simultaneously considers three dimensions: spatial distance, fringe contrast, and phase consistency, allowing for the reasonable utilization of information from multiple frames of interferometric images for fusion.

[0006] In one optional implementation, the fusion weight of a pixel is the product of its spatial distance weight, stripe contrast weight, and phase consistency weight. Thus, when the weight of any one of the three dimensions is low, the fusion weight decreases accordingly, achieving joint constraints across the three dimensions.

[0007] In one alternative implementation, the spatial distance weight of a pixel is determined based on the distance between the pixel's mapped position on the high-resolution grid and its neighboring grid points, and this spatial distance weight decreases as the distance increases. Therefore, pixels whose mapped positions are closer to the grid points contribute more to the fusion result.

[0008] In one alternative implementation, the spatial distance weight is a Gaussian function with the distance as the independent variable. Thus, the spatial distance weight decays smoothly with distance in the form of a Gaussian function.

[0009] In one alternative implementation, the stripe contrast weight of a pixel is determined based on the local gradient magnitude at the pixel's location, and the stripe contrast weight increases as the local gradient magnitude increases. Therefore, pixels with high stripe contrast and high information content receive a higher weight in the fusion process.

[0010] In one alternative implementation, the phase consistency weight of the pixel is determined based on the phase difference between the pixel's phase position and the frames, and the phase consistency weight decreases as the phase difference increases. Therefore, pixels with phase consistency across frames have a higher weight in the fusion process.

[0011] In one alternative implementation, the phase consistency weight is the absolute value of the cosine of half the phase difference. Therefore, near the phase transition location, even if the grayscale values ​​of pixels differ significantly, the phase consistency weight will automatically decrease, reducing fusion artifacts.

[0012] In one optional implementation, the process of mapping N frames of interferometric images onto a high-resolution grid and fusing them to obtain a high-resolution interferometric image further includes: determining a phase gradient field based on the phase of the sampled positions in the high-resolution grid; solving the Poisson equation using the phase gradient field as a constraint to obtain a continuous phase field; and determining the pixel values ​​of the unsampled positions in the high-resolution grid according to the continuous phase field to fill the high-resolution interferometric image. Thus, the unsampled positions in the high-resolution grid are filled using the continuity constraint of the interferometric phase, and the filling result is physically reasonable.

[0013] In one optional implementation, solving the Poisson equation to obtain a continuous phase field includes: discretizing the Poisson equation into a system of linear equations; and solving the system of linear equations using an iterative method to obtain the continuous phase field. Thus, by solving the system of linear equations using an iterative method, a continuous phase field can be obtained on a large-scale, high-resolution grid.

[0014] In one alternative implementation, the iterative method is the conjugate gradient method. Therefore, the conjugate gradient method has a faster convergence speed and is suitable for solving discretized sparse linear equation systems.

[0015] In one optional implementation, determining the pixel values ​​at unsampled locations in the high-resolution grid based on the continuous phase field includes: generating the pixel values ​​at the unsampled locations based on the continuous phase field and the interference intensity function. Thus, the pixel values ​​generated by the continuous phase field conform to the physical laws of interferometric imaging.

[0016] In one optional implementation, mapping N frames of interferometric images to a high-resolution grid and fusing them includes: extracting the phase of each frame of the N frames of interferometric images to obtain the phase of each frame; determining the global phase difference of each frame of interferometric images relative to a reference frame, wherein the reference frame is one frame in the N frames of interferometric images; compensating the phase of each frame of interferometric images based on the global phase difference to perform phase alignment of the N frames of interferometric images; and mapping the phase-aligned N frames of interferometric images to the high-resolution grid and fusing them. This eliminates inter-frame phase drift, allowing the N frames of interferometric images to be fused under a unified phase reference, ensuring the phase continuity of the fusion result.

[0017] In one optional implementation, determining the global phase difference of each interferometric image relative to the reference frame includes: determining the average of the differences between the phase of each interferometric image and the phase of the reference frame in the overlapping region as the global phase difference of that interferometric image relative to the reference frame. Thus, by averaging the phase differences in the overlapping region, the influence of local noise on the global phase difference can be suppressed.

[0018] In one optional implementation, the sub-pixel offsets between the N interferometric images are determined as follows: based on the phase correlation method, the sub-pixel offsets between two adjacent interferometric images are determined; based on the sub-pixel offsets between two adjacent interferometric images, the sub-pixel offset of each interferometric image relative to a reference frame is determined, wherein the reference frame is one of the N interferometric images. Thus, by estimating the offsets between adjacent frames using the phase correlation method and accumulating them, the sub-pixel offsets of each interferometric image relative to the reference frame can be obtained.

[0019] In one optional implementation, determining the sub-pixel offset between two adjacent interferometric images in the N frames of interferometric images based on the phase correlation method includes: performing frequency domain transformation on each of the two adjacent interferometric images and calculating the cross-power spectrum; performing an inverse frequency domain transformation on the cross-power spectrum to obtain a phase correlation function; and locating the peak value of the phase correlation function to obtain the sub-pixel offset between the two adjacent interferometric images. Therefore, by calculating the cross-power spectrum in the frequency domain and locating the peak value at the sub-pixel level, the inter-frame sub-pixel offset can be accurately estimated.

[0020] In one optional implementation, the phase extraction of the high-resolution interferometric image to obtain the phase distribution of the measured object includes: performing a frequency domain transformation on the high-resolution interferometric image to obtain a target spectrum; extracting a first-order spectral component from the target spectrum; filtering and performing an inverse frequency domain transformation on the first-order spectral component to obtain an analytical signal; determining the wrapped phase based on the analytical signal, and unwrapping the wrapped phase to obtain the phase distribution.

[0021] In one optional implementation, extracting a first-order spectral component from the target spectrum includes: sub-pixel localization of peaks in the target spectrum; determining a bandpass filter based on the peak positions after sub-pixel localization; and extracting the first-order spectral component from the target spectrum using the bandpass filter. Thus, the bandpass filter is adaptively determined according to the peak position, enabling accurate extraction of the first-order spectral component.

[0022] In one alternative implementation, the pixel spacing of the high-resolution grid is 1 / N of the pixel spacing of the N frames of the interferometric image. Thus, the spatial resolution of the fused interferometric image is N times that of a single frame of the interferometric image.

[0023] Secondly, embodiments of this application provide an interferometric image processing system, comprising: an acquisition unit configured to acquire N frames of interferometric images, wherein the N frames of interferometric images are interferometric images acquired for the same object under test, and the N frames of interferometric images have sub-pixel offsets, where N is an integer greater than 1; a fusion unit configured to map the N frames of interferometric images to a high-resolution grid and fuse them according to the sub-pixel offsets between the N frames of interferometric images to obtain a high-resolution interferometric image, wherein the resolution of the high-resolution grid is higher than the resolution of the N frames of interferometric images; and an extraction unit configured to perform phase extraction on the high-resolution interferometric image to obtain the phase distribution of the object under test.

[0024] In an optional implementation, the fusion unit is further configured to: determine a fusion weight for each pixel among a plurality of pixels in the N-frame interferometric images, based on a spatial distance weight, a fringe contrast weight, and a phase consistency weight; and perform weighted fusion of the plurality of pixels mapped to the high-resolution grid based on the fusion weight of each pixel among the plurality of pixels; wherein the phase consistency weight is used to characterize the degree of consistency of the phase at the pixel location among the N-frame interferometric images.

[0025] In an alternative implementation, the system further includes a filling unit configured to: determine a phase gradient field based on the phase of the sampled locations in the high-resolution grid; solve the Poisson equation with the phase gradient field as a constraint to obtain a continuous phase field; and determine the pixel values ​​of the unsampled locations in the high-resolution grid according to the continuous phase field to fill the high-resolution interferometric image.

[0026] In an optional implementation, the system further includes an alignment unit configured to: extract the phase of each interferometric image in the N frames to obtain the phase of each interferometric image; determine the global phase difference of each interferometric image relative to a reference frame, wherein the reference frame is one of the N frames of interferometric images; and compensate the phase of each interferometric image according to the global phase difference to perform phase alignment of the N frames of interferometric images.

[0027] Thirdly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of the first aspect of this application.

[0028] Fourthly, embodiments of this application provide an electronic device including a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the method of the first aspect of this application. Attached Figure Description

[0029] Figure 1 A flowchart of an interference image processing method according to an example embodiment of this application is shown; Figure 2 A schematic diagram of pixel mapping fusion of multi-frame interferometric images according to an example embodiment of this application is shown; Figure 3 A schematic diagram of vacancy filling in a high-resolution mesh according to an example embodiment of this application is shown; Figure 4 A schematic diagram of the structure of an interferometric image processing system according to an example embodiment of this application is shown; Figure 5A schematic diagram of the structure of an electronic device according to an example embodiment of this application is shown; Figure 6 A schematic diagram of the structure of a chip module according to an example embodiment of this application is shown. Detailed Implementation

[0030] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Although the description of the present invention is presented in conjunction with preferred embodiments, this does not mean that the features of the invention are limited to these embodiments. On the contrary, the purpose of describing the invention in conjunction with embodiments is to cover other options or modifications that may be derived based on the claims of the present invention. To provide a thorough understanding of the invention, many specific details will be included in the following description. The invention may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of the invention, some specific details will be omitted in the description.

[0031] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0032] The following explains some of the terms used in this application.

[0033] Interference image: An image acquired and recorded by an interferometer, showing an interference fringe pattern. The interference fringes are formed by the interference between the measurement light reflected from the object under test and the reference light. The pixel value (grayscale value) of each pixel in the interference image changes periodically with the phase difference between the measurement light and the reference light. Therefore, by extracting the phase from the interference image, the phase distribution corresponding to the surface morphology of the object under test can be obtained.

[0034] Subpixel offset: An offset of less than one pixel. In this application, the subpixel offset between N frames of interferometric images refers to a relative displacement of less than one pixel between the imaging areas corresponding to each frame of interferometric images.

[0035] Wrapping and unwrapping: The phase obtained directly from phase extraction is usually truncated within the principal value interval of one period (e.g., [-π, π)), which is called the wrapped phase; the process of restoring the wrapped phase to a continuous phase is called unwrapping.

[0036] Phase correlation method: A method for estimating the translation between two images based on the cross-power spectrum in the frequency domain. The phase difference between the spectra of two translated images corresponds to the translation amount. Therefore, by performing an inverse frequency domain transform on the cross-power spectrum and locating the peak, the translation between the two images can be determined.

[0037] First-order spectral component: In the spectrum of an interference image, the spectral component generated by the carrier frequency of the interference fringes, whose center frequency corresponds to the spatial frequency of the interference fringes. Filtering and inverse frequency domain transformation of the first-order spectral component can yield an analytical signal for phase extraction.

[0038] Analytical signal: A complex signal constructed from a real signal that contains only positive frequency components. The phase of an analytic signal is the same as the phase of the real signal.

[0039] Optical Path Difference (OPD): The difference between the optical path length of the measured light and that of the reference light. The OPD corresponds to the surface morphology of the measured object and can be calculated from the phase distribution.

[0040] The technical solutions of this application can be applied to interferometric measurement scenarios. In fields such as semiconductor manufacturing and precision optical processing, interferometers (such as Fizeau interferometers, Michelson interferometers, and Twyman-Green interferometers) can be used to measure the surface morphology of wafers, the surface shape of optical components, and the surface of micro-electro-mechanical systems (MEMS) devices. The image sensor of the interferometer (such as a CCD or CMOS sensor) acquires the interference image, and phase extraction is performed on the interference image to obtain the phase distribution corresponding to the surface morphology of the object being measured, thereby obtaining optical path difference information.

[0041] Due to the limitations of the pixel size of image sensors, the highest spatial frequency that a single frame of an interference image can resolve is limited. For example, if the pixel size of the image sensor is 50 μm, according to the sampling theorem, the smallest fringe spacing that a single frame of an interference image can resolve is twice the pixel size, i.e., 100 μm. When the surface of the object under test has high-frequency characteristics and the spatial frequency of the interference fringes is close to or exceeds the above sampling limit, aliasing will occur, leading to phase extraction failure.

[0042] It should be noted that general multi-frame super-resolution methods are difficult to directly apply to interferometric images. This is because interference fringes have strong periodicity, and phase transition positions exist in the interferometric image (e.g., 2π truncation positions enclosing the phase). On both sides of the phase transition position, the gray values ​​of pixels differ significantly, but the actual phase is continuous. If multiple interferometric images are fused only according to spatial proximity, pixels with large gray value differences but continuous actual phases are easily fused incorrectly, resulting in artifacts near the phase transition positions. Furthermore, during multi-frame acquisition, factors such as laser power fluctuations, ambient temperature changes, and mechanical vibrations can cause inter-frame phase drift. If the phase references of each interferometric image frame are inconsistent, the phase of the fused result will be discontinuous. This application's embodiments address the above-mentioned characteristics of interferometric images by performing multi-frame fusion, as detailed in the descriptions of the embodiments below.

[0043] Figure 1 A flowchart of an interferometric image processing method according to an example embodiment of this application is shown. This method can be executed, for example, by an interferometric image processing system, which may be a computing device connected to an interferometer, or a processing module integrated into the interferometer; this application does not limit this to the specific implementation. Figure 1 As shown, the method includes: Step S101: Obtain N frames of interference images.

[0044] The N interferometric images are interferometric images acquired for the same object under test, and there is a sub-pixel offset between the N interferometric images, where N is an integer greater than 1. The sub-pixel offset between the interferometric images ensures that the sampling positions of the same area under test in different frames are staggered, providing a sampling basis for subsequent fusion on a higher resolution grid.

[0045] In one implementation, N frames of interferometric images are acquired using a displacement stage: the displacement stage moves the image sensor of the object under test or the interferometer, and interferometric images are acquired at multiple acquisition positions. The displacement between two adjacent acquisitions is less than one pixel, thus obtaining N frames of interferometric images with sub-pixel offsets between them.

[0046] In another implementation, the relative position between the object under test and the interferometer changes slightly due to factors such as environmental vibration, and there may also be sub-pixel shifts between the interferometric images acquired at different acquisition times. These N frames of interferometric images can also be used as inputs for embodiments of this application.

[0047] Furthermore, the sub-pixel offset between N frames of interferometric images can be obtained through various methods. In one method, the sub-pixel offset is determined by the displacement feedback information from the acquisition device, such as a displacement stage; that is, the displacement of the displacement stage is directly read when acquiring each frame of the interferometric image, and the sub-pixel offset between each frame is determined accordingly. In another method, the sub-pixel offset is estimated based on the acquired interferometric images. This application does not limit the method for obtaining the sub-pixel offset.

[0048] In one implementation, the subpixel offset is determined based on the phase correlation method: The subpixel offset between two adjacent interferometric images in N frames is determined using the phase correlation method; then, based on the subpixel offsets between adjacent interferometric images, these offsets are sequentially accumulated to determine the subpixel offset of each interferometric image relative to a reference frame. The reference frame is one of the N interferometric images, for example, the first interferometric image in the N frames.

[0049] Optionally, determining the sub-pixel offset between two adjacent interferometric images based on the phase correlation method includes: performing frequency domain transformation on each of the two adjacent interferometric images to obtain two spectra, and calculating the cross-power spectrum of the two spectra; performing an inverse frequency domain transformation on the cross-power spectrum to obtain the phase correlation function; and locating the peak of the phase correlation function in sub-pixel form, with the peak position being the sub-pixel offset between the two adjacent interferometric images. The frequency domain transformation can be, for example, a Fourier transform, and the inverse frequency domain transformation can be, for example, an inverse Fourier transform.

[0050] For example, for the adjacent k-th frame interferometric image I k Interference image I of frame k+1 k+1 Perform two-dimensional Fourier transforms to obtain the spectrum F. k and spectrum F k+1 ; Calculate the cross-power spectrum P(u,v)=F k ·conj(F k+1 ) / |F k ·conj(F k+1 )|, where conj(·) represents taking the conjugate; performing an inverse Fourier transform on the cross-power spectrum yields the phase correlation function c(m,n), which peaks at positions corresponding to the inter-frame shift; sub-pixel localization of the peak positions is achieved through quadratic polynomial interpolation, thus obtaining the sub-pixel offset between the k-th and k+1-th interferometric images; the sub-pixel offsets between each pair of adjacent interferometric images are sequentially accumulated to obtain the sub-pixel offset δ of each interferometric image relative to the reference frame. k =(δ k x , δ k y ).

[0051] Optionally, sub-pixel localization of the peak of the phase correlation function can also be performed using the centroid method or Gaussian fitting method, etc. This application does not limit the sub-pixel localization method.

[0052] Therefore, by calculating the cross-power spectrum and locating the peak in the frequency domain, the offset estimation is not affected by the gray-scale changes and noise of the interference fringes, and the offset estimation results with sub-pixel accuracy can be obtained through sub-pixel positioning.

[0053] Step S102: Based on the sub-pixel offset between N frames of interferometric images, map the N frames of interferometric images to a high-resolution grid and fuse them to obtain a high-resolution interferometric image.

[0054] The high-resolution grid has a higher resolution than the N-frame interferometric images. After establishing the high-resolution grid, for each frame of the interferometric image, based on the sub-pixel offset of that frame relative to the reference frame, each pixel of that frame is mapped to its corresponding position on the high-resolution grid. Then, the multiple pixels mapped to the high-resolution grid are fused to obtain the high-resolution interferometric image. Because there is a sub-pixel offset between the interferometric frames, the positions of pixels from different frames mapped to the high-resolution grid are staggered. Therefore, a denser spatial sampling can be obtained on the high-resolution grid than on a single frame of the interferometric image, thus improving the spatial resolution of the fused high-resolution interferometric image.

[0055] In one implementation, the pixel pitch of the high-resolution grid is 1 / N of the pixel pitch of the N-frame interferometric images, that is, the pixel size of the high-resolution grid is 1 / N of the pixel size of a single-frame interferometric image.

[0056] For example, if the pixel size of a single frame of interference image is 50 μm and N is 10, then the pixel size of the high-resolution grid is 5 μm; correspondingly, the minimum resolvable fringe spacing can be reduced from 100 μm in a single frame to 10 μm.

[0057] Furthermore, phase alignment can be performed on the N frames of interferometric images. During multi-frame acquisition, factors such as laser power fluctuations, ambient temperature changes, and mechanical vibrations can cause global phase drift between frames. If the phase references of each interferometric image are inconsistent, the phase of the fused result will be discontinuous. Therefore, phase alignment is performed on the N frames of interferometric images, and the phase-aligned interferometric images are then fused. Phase alignment includes: extracting the phase of each interferometric image from the N frames to obtain the phase of each frame; determining the global phase difference of each interferometric image relative to the reference frame; and compensating for the phase of each interferometric image based on the global phase difference to obtain the phase-aligned interferometric images.

[0058] In one implementation, determining the global phase difference of each interferometric image relative to a reference frame includes: determining the average of the differences between the phase of each interferometric image and the phase of the reference frame in the overlapping region as the global phase difference of that interferometric image relative to the reference frame.

[0059] Optionally, phase extraction can be performed on each frame of the interferometric image. For example, a single-frame phase extraction method based on Fourier transform can be used to obtain the wrapper phase of the interferometric image in that frame. The global phase difference characterizes the overall phase shift between the interferometric image in that frame and the reference frame caused by factors such as light source fluctuations and environmental changes. It is independent of the phase distribution of the interferometric image in that frame in space, so it can be statistically obtained in the overlapping region between frames. By averaging the phase difference in the overlapping region, the influence of local noise on the global phase difference can be suppressed.

[0060] For example, for the k-th frame of the interferometric image I k Perform single-frame phase extraction to obtain the wrapper phase. k After spatially aligning the k-th frame interferometric image with the reference frame, calculate the wrap phase. k Wrapping phase with reference frame The difference between the k-th frame interferometric image and the reference frame is calculated by taking the average of the differences in the overlapping regions between frames. k That is, ΔΦ k =mean( k - 0); then, phase compensation is performed on the k-th frame of the interferometric image, i.e. k aligned= k -ΔΦ k This yields the phase-aligned interferometric image of the k-th frame. It should be noted that the global phase difference between the reference frame and itself is zero, and the phase remains unchanged after compensation.

[0061] Optionally, in calculating the global phase difference ΔΦ k At that time, first determine the phase difference at each position ( k - 0) Perform principal value transformation before taking the mean, i.e., ΔΦ k =mean(wrap( k - 0), where wrap(·) represents principal phase squaring. Primitive phase squaring refers to folding the phase difference at each position into a principal phase interval (e.g., [-π, π)). Since the wrapped phase is truncated within the principal phase interval, the difference between the wrapped phases of two frames at the same position on either side of the 2π truncation point may deviate from the actual inter-frame phase difference by an integer multiple of 2π. Primitive phase squaring folds the differences at each position into the same principal phase interval, eliminating jumps that are integer multiples of 2π, thus avoiding the impact of the wrapped phase truncation on the global phase difference calculation.

[0062] Therefore, by extracting the phase of each frame of the interferometric image, calculating the global phase difference, and compensating for it, the interferometric images of each frame can be fused under a unified phase reference. This can eliminate the influence of inter-frame phase drift on the fusion result and ensure the phase continuity of the fusion result.

[0063] In one implementation, the fusion in S102 includes weighted fusion: for each pixel among multiple pixels in the N-frame interferometric image, the fusion weight of the pixel is determined; based on the fusion weight of each pixel among multiple pixels, multiple pixels at neighboring positions of the same grid point mapped to the high-resolution grid are weighted and superimposed, and the result of the weighted superposition is used as the pixel value of the grid point, thereby obtaining the high-resolution interferometric image.

[0064] Figure 2 A schematic diagram illustrating pixel mapping fusion of multi-frame interferometric images according to an example embodiment of this application is shown. Figure 2 As shown, taking N=4 as an example, frames 1 to 4 have sub-pixel offsets. The sub-pixel offset of each frame's interferometric image relative to the reference frame can be decomposed into X-direction components and Y-direction components. Based on the sub-pixel offset of each frame's interferometric image, each pixel in frames 1 to 4 is mapped to different positions on the high-resolution grid. By performing weighted fusion on multiple pixels in adjacent positions of the same grid point mapped to the high-resolution grid, the pixel value at that grid point can be obtained. Figure 2 In the high-resolution grid, each cell represents the weighted and fused pixel value using shades of gray.

[0065] For example, by mapping four 1000×1000 pixel interferometric images with sub-pixel offsets to a high-resolution grid of 4000×4000 pixels with four times the resolution and then performing weighted fusion, a high-resolution interferometric image of 4000×4000 pixels can be obtained.

[0066] In one implementation, for each pixel among multiple pixels in an N-frame interferometric image, a fusion weight is determined based on the pixel's spatial distance weight, fringe contrast weight, and phase consistency weight. The phase consistency weight characterizes the degree of consistency of the pixel's phase across the N-frame interferometric images.

[0067] Optionally, the spatial distance weight is determined based on the distance between the mapped position of a pixel on the high-resolution grid and its neighboring grid points. The spatial distance weight decreases as the distance increases; that is, the closer the mapped position is to a grid point, the greater the contribution of the pixel to the pixel value of that grid point. The fringe contrast weight is determined based on the local gradient magnitude at the pixel's location. The fringe contrast weight increases as the local gradient magnitude increases. The local gradient magnitude represents the fringe contrast at the pixel's location and can be obtained by differential calculation of the gray levels of the pixel's neighborhood. For example, the Sobel operator can be used to calculate the gradients of the pixel's neighborhood in the horizontal and vertical directions. Locations with high interference fringe contrast have more reliable fringe information and contain more phase information, and therefore are given higher weights. Phase consistency weight is determined based on the phase difference between the pixel positions in frames. The phase consistency weight decreases as the phase difference increases. There are phase jump positions in the interferometric image (such as the 2π truncation position that wraps the phase). On both sides of the phase jump position, the gray values ​​of the pixels are different, but the actual phase is continuous. The phase consistency weight decreases as the phase difference between frames increases, so that the fusion weight of pixels near the phase jump position is automatically reduced, thereby avoiding the incorrect fusion of pixels with large differences in gray values ​​but continuous actual phase.

[0068] For example, spatial distance weight w dist =exp(-d² / 2σ²), where d is the distance between the mapped pixel position and the grid point, σ is the scale parameter of the Gaussian function, and the value of σ can be set according to the pixel spacing of the high-resolution grid; the stripe contrast weight w cont =| I k (m,n)| / max(| I|), where | I k (m,n)| represents the local gradient magnitude of the k-th frame of the interferometric image at pixel (m,n), max(| I|) represents the maximum local gradient magnitude of each frame of the interferometric image, used for normalization; phase consistency weight w phase =|cos(Δ k / 2)|,where, Δ k For the corresponding pixel position, the phase difference between the interferometric image in the k-th frame and the reference frame, when Δ k When approaching 0 w phase When Δ approaches 1 k When approaching π (i.e. near the phase transition position) w phase Close to 0.

[0069] Optionally, the pixel fusion weight is the product of the pixel's spatial distance weight, stripe contrast weight, and phase consistency weight, i.e., w = w dist ×w cont ×w phase .

[0070] For example, the pixel values ​​at grid point (x,y) in a high-resolution interferometric image can satisfy: I HR (x,y)=Σ k I k (m,n)·w k (m,n) / Σ k w k (m,n), where I k (m,n) represents the pixel value of the pixel (m,n) mapped to the nearest position of the grid point (x,y) in the k-th frame of the interferometric image, w k (m,n) represents the fusion weight of the pixel, and the denominator is used to normalize the fusion weight.

[0071] It should be noted that this application does not limit the specific form of the fusion weights. In another implementation, the spatial distance weight can also adopt other functional forms that decrease with increasing distance, such as the exponential decay function, B-spline basis function, etc.; the fringe contrast weight can also adopt other measures that can characterize the sharpness of local fringes, such as the local gradient magnitude calculated by the Scharr operator, Prewitt operator, etc.; the phase consistency weight can also adopt other functional forms that decrease with increasing phase difference.

[0072] Furthermore, unsampled locations in the high-resolution grid can be filled. After mapping N frames of interferometric images to the high-resolution grid, there may be unsampled locations (i.e., empty spaces) within the grid. Sampled locations are those mapped to pixels in the high-resolution grid, while unsampled locations are those not mapped to pixels. Unsampled locations have no pixel values ​​and need to be filled to obtain a complete high-resolution interferometric image.

[0073] Figure 3 A schematic diagram illustrating the vacancy filling of a high-resolution mesh according to an example embodiment of this application is shown. Figure 3 As shown, in the weighted fusion high-resolution grid, the sampled positions have the fused pixel values, while the unsampled positions have no pixel values; after filling the gaps, the unsampled positions also have pixel values, thus obtaining a complete high-resolution interferometric image.

[0074] For example, when N is 10 and the pixel pitch of the high-resolution grid is 1 / 10 of the pixel pitch of a single frame of interferometric image, the total number of positions of the high-resolution grid is 10 times the total number of pixels of 10 frames of interferometric image. Even if the pixels of each frame of interferometric image are mapped to completely different positions, the proportion of unsampled positions in the high-resolution grid reaches about 90%.

[0075] In one implementation, vacancy filling includes: extracting the phase of the sampled positions in the high-resolution grid to obtain the phase of the sampled positions, and calculating the phase gradient field of the sampled positions; solving the Poisson equation with the phase gradient field as a constraint to obtain a continuous phase field covering the high-resolution grid; and determining the pixel values ​​of the unsampled positions in the high-resolution grid based on the continuous phase field to fill the high-resolution interferometric image.

[0076] Optionally, the phase is extracted from the sampled locations in the high-resolution grid, and the phase gradient field g= sampled The empty space filling problem is transformed into the following optimization problem: min | -g|²dxdy, that is, finding the continuous phase field that minimizes the deviation from the phase gradient field g at the sampled position. This optimization problem corresponds to solving the Poisson equation. Since the real surface of the measured object is usually continuous, and its phase distribution is also continuous, the filling result obtained by using the phase gradient field as a constraint satisfies the phase continuity constraint, which is physically reasonable. Compared with the filling method that only uses spatial proximity for interpolation (such as bilinear interpolation), the above filling method utilizes the physical continuity constraint of the interference phase, and the filling error of the unsampled positions is smaller.

[0077] Optionally, solving the Poisson equation includes: applying boundary conditions to the high-resolution grid, such as Dirichlet or Neumann boundary conditions, where the boundary values ​​can be determined based on the phase of the sampled positions in the high-resolution grid; discretizing the Poisson equation on the high-resolution grid to obtain a system of linear equations; and solving the system of linear equations using an iterative method to obtain a continuous phase field. In one implementation, the iterative method is the conjugate gradient method. With the Dirichlet boundary conditions applied, the coefficient matrix obtained after discretizing the Poisson equation is a symmetric positive definite sparse matrix. The conjugate gradient method can utilize the properties of this matrix to iteratively search along the conjugate direction, making it suitable for solving large-scale sparse linear equation systems with a relatively fast convergence speed. Optionally, the iterative method can also be the Gauss-Seidel iteration method, the successive over-relaxation method, or the multigrid method, etc., which are not limited in this application.

[0078] For example, discretizing the Poisson equation into a sparse linear system of equations L =div(g), where L is the Laplace operator and div(g) is the divergence of the phase gradient field g; the sparse linear equation system is solved iteratively using the conjugate gradient method to obtain the complete and continuous phase field. HR .

[0079] In one implementation, determining the pixel values ​​at unsampled locations based on a continuous phase field includes: generating the pixel values ​​at unsampled locations based on the continuous phase field and an interference intensity function. The interference intensity function describes the relationship between the pixel values ​​and the phase of the interference image.

[0080] For example, the interference intensity function can satisfy I HR =A+B·cos( HR (where A is the background light intensity and B is the modulation index). HR This represents the phase value of the continuous phase field at the corresponding location. For unsampled locations in the high-resolution grid, substituting the phase value of the continuous phase field at that location into the interference intensity function generates the pixel value at that location, thus obtaining a complete and continuous high-resolution interferometric image.

[0081] Step S103: Phase extraction is performed on the high-resolution interferometric image to obtain the phase distribution of the object under test.

[0082] Phase extraction of high-resolution interferometric images can be performed, for example, by performing frequency domain transformation on the high-resolution interferometric image, extracting the spectral components corresponding to the carrier frequencies of the interference fringes in the frequency domain, and then filtering and performing inverse frequency domain transformation on the extracted spectral components to extract the phase and obtain the phase distribution of the measured object.

[0083] In one implementation, phase extraction includes: performing a frequency domain transformation on the high-resolution interferometric image to obtain the target spectrum; extracting a first-order spectral component from the target spectrum; filtering and performing an inverse frequency domain transformation on the first-order spectral component to obtain an analytical signal; extracting the amplitude of the analytical signal to obtain the wrapped phase, and unwrapping the wrapped phase to obtain the phase distribution.

[0084] Optionally, extracting first-order spectral components from the target spectrum includes: sub-pixel localization of peaks in the target spectrum; determining a bandpass filter based on the peak positions after sub-pixel localization; and extracting first-order spectral components from the target spectrum using the bandpass filter. Sub-pixel localization of peaks in the target spectrum can be achieved, for example, using the centroid method, which determines the centroid position based on the amplitude of each spectral point in the peak's neighborhood, and uses this as the peak's sub-pixel position; or using a Gaussian fitting method, which fits a Gaussian surface to the spectral amplitudes in the peak's neighborhood, and uses the fitted peak position as the sub-pixel position. The bandpass filter determined based on the peak positions after sub-pixel localization can be, for example, an adaptive Gaussian bandpass filter, where the center position of the bandpass filter is adaptively determined according to the peak position. Therefore, the bandpass filter can accurately cover the first-order spectral components, avoiding the impact of spectral leakage on phase extraction.

[0085] Optionally, the wrapped phase can be unwrapped, for example, by using a quality-guided unwrapping algorithm; after obtaining the phase distribution, the optical path difference information of the measured object can be calculated based on the phase distribution.

[0086] For example, since the spatial resolution of a high-resolution interferometric image is N times that of a single-frame interferometric image, the frequency domain range of the target spectrum is also expanded accordingly. For instance, based on the pixels of a single-frame interferometric image, the range can be expanded from [-0.5, 0.5] cycles / pixel to [-N / 2, N / 2] cycles / pixel. Therefore, when extracting first-order spectral components from the target spectrum, high-frequency components that exceed the sampling limit of a single-frame interferometric image can be resolved, thus recovering high-frequency fringe information that cannot be resolved by a single-frame interferometric image.

[0087] Therefore, by performing phase extraction on high-resolution interferometric images, a high-resolution phase distribution with N times the spatial resolution can be obtained, thereby acquiring the optical path difference information of the measured object. Furthermore, multi-frame fusion can suppress noise and help improve phase accuracy.

[0088] Figure 4 A schematic diagram of the structure of an interferometric image processing system according to an example embodiment of this application is shown.

[0089] like Figure 4As shown, the interferometric image processing system 400 includes an acquisition unit 401, a fusion unit 402, and an extraction unit 403. The acquisition unit 401 is configured to acquire N frames of interferometric images, wherein the N frames are interferometric images acquired for the same object under test, and there is a sub-pixel offset between the N frames, where N is an integer greater than 1. The fusion unit 402 is configured to map the N frames of interferometric images to a high-resolution grid and fuse them according to the sub-pixel offset between the N frames to obtain a high-resolution interferometric image, wherein the resolution of the high-resolution grid is higher than the resolution of the N frames of interferometric images. The extraction unit 403 is configured to perform phase extraction on the high-resolution interferometric image to obtain the phase distribution of the object under test.

[0090] Optionally, the interference image processing system 400 may further include at least one of an alignment unit 404 and a filling unit 405. Figure 4 The images are shown in dashed boxes. The alignment unit 404 is configured to: extract the phase of each interferometric image in the N frames to obtain the phase of each frame; determine the global phase difference of each frame relative to a reference frame, where the reference frame is one of the N frames; and compensate for the phase of each frame based on the global phase difference to perform phase alignment on the N frames. The filling unit 405 is configured to: determine the phase gradient field based on the phase of the sampled positions in the high-resolution grid; solve the Poisson equation using the phase gradient field as a constraint to obtain a continuous phase field; and determine the pixel values ​​of the unsampled positions in the high-resolution grid based on the continuous phase field to fill the high-resolution interferometric image.

[0091] Optionally, the fusion unit 402 can be further configured to: determine the fusion weight of each pixel among the plurality of pixels in the N-frame interferometric images based on the spatial distance weight, fringe contrast weight, and phase consistency weight of the pixel; and perform weighted fusion of the plurality of pixels mapped to the high-resolution grid based on the fusion weight of each pixel among the plurality of pixels; wherein the phase consistency weight is used to characterize the degree of consistency of the phase of the pixel position among the N-frame interferometric images.

[0092] It should be noted that the specific implementations of the operations performed by the acquisition unit 401, fusion unit 402, extraction unit 403, alignment unit 404, and filling unit 405 can be found in the relevant descriptions in the foregoing method embodiments, and will not be repeated here. Each of the above units can be a software unit, a hardware unit, or a combination of both. For example, each unit can be implemented entirely in hardware such as circuits, or at least some units can be implemented in software programs running on the processor, while the remaining (if any) units can be implemented in hardware such as circuits.

[0093] Figure 5 A schematic diagram of the structure of an electronic device according to an example embodiment of this application is shown.

[0094] like Figure 5 As shown, the electronic device 500 may include a memory 501 and a processor 502; optionally, it may also include a communication interface 503. The memory 501, processor 502, and communication interface 503 are connected via one or more communication buses 504.

[0095] The memory 501 stores program instructions and data, and may include read-only memory, random access memory, non-volatile random access memory, etc. The communication interface 503 receives or transmits data, such as receiving N frames of interferometric images acquired by the interferometer. The communication interface 503 may include a transmitting module and / or a receiving module. The processor 502 calls the program instructions stored in the memory 501 to cause the electronic device 500 to execute the interferometric image processing method described in the foregoing method embodiments. The processor 502 may be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0096] In one implementation, the interferometric image processing method described in this application can be executed by a chip or a chip module. The above embodiments can be applied to electronic devices, or chips or chip modules within electronic devices. The chip may include at least one processor for executing program instructions to perform the interferometric image processing method of this application.

[0097] Figure 6A schematic diagram of the structure of a chip module according to an example embodiment of this application is shown. Figure 6 As shown, the chip module 600 may include a communication interface 601 and a chip 602. The communication interface 601 is used for internal communication within the chip module and / or for communication between the chip module and external devices; the communication interface 601 can also be described as a communication module. The chip 602 is used to execute the interferometric image processing method described in the foregoing method embodiments. Optionally, the chip module 600 may further include a storage module 603 and a power module 604. Figure 6 The image is shown in a dashed box; the storage module 603 is used to store data and instructions, and the power module 604 is used to provide power to the chip module.

[0098] The interference image processing method of this application will be described below with reference to Embodiments 1 to 3.

[0099] Example 1

[0100] In this embodiment, N is 4. Four interferometric images with sub-pixel offsets are acquired from the same object under test using an interferometer, and the size of each interferometric image is 1000×1000 pixels.

[0101] In step S101, the four interferometric images are acquired. Further, taking frame 1 as the reference frame, the sub-pixel offsets of frames 2 to 4 relative to frame 1 are estimated based on the phase correlation method: two-dimensional Fourier transforms are performed on two adjacent interferometric images respectively and the cross power spectrum is calculated. The phase correlation function is obtained by performing an inverse Fourier transform on the cross power spectrum. The peak value is located by quadratic polynomial interpolation to obtain the sub-pixel offset between two adjacent frames. The sub-pixel offsets of frames 2 to 4 relative to frame 1 are then accumulated.

[0102] In step S102, further, phase alignment is performed on the four interferometric images: single-frame phase extraction is performed on each interferometric image to obtain the wrapping phase of each frame. The average difference between the wrapping phase of each interferometric image and the wrapping phase of frame 1 in the overlapping region is used as the global phase difference of that frame, and phase compensation is performed on each frame so that the four interferometric images are fused under a unified phase reference. Then, a high-resolution grid of 4000×4000 pixels is established, and each pixel of the four interferometric images is mapped onto the high-resolution grid according to the sub-pixel offset of each frame; for each pixel, its spatial distance weight w is calculated. dist =exp(-d² / 2σ²), stripe contrast weight w cont =| I k (m,n)| / max(| I|) and phase consistency weight w phase =|cos(Δ k / 2)|, take the product of the three as the fusion weight; weighted superposition of multiple pixels mapped to the same grid point in neighboring locations yields a high-resolution interferometric image. Further, the unsampled locations in the high-resolution grid are filled: the phase is extracted from the sampled locations and the phase gradient field is calculated, discretizing the Poisson equation into a sparse linear system of equations L =div(g), the continuous phase field is obtained by iterative solution using the conjugate gradient method, and then based on the interference intensity function I HR =A+B·cos( HR The pixel values ​​of the unsampled locations are generated to obtain a complete 4000×4000 pixel high-resolution interferometric image.

[0103] In step S103, a two-dimensional Fourier transform is performed on the high-resolution interferometric image. In the frequency domain with a 4x extension, the peak of the first-order spectrum is located at the sub-pixel level using the centroid method. An adaptive Gaussian bandpass filter is determined based on the peak position, and the first-order spectrum components are extracted. An analytical signal is obtained by inverse Fourier transform. The wrapped phase is extracted, and a continuous phase distribution is obtained by a quality-guided unwrapping algorithm.

[0104] In this embodiment, a phase consistency dimension is introduced into the fusion weight, and the fusion weight of pixels near the phase transition position is automatically reduced. Through comparison and verification, compared with the fusion method that only uses spatial distance weight, the weighted fusion method in this embodiment has less impact on the stripe width, produces fewer artifacts in the phase transition region, and has smaller phase error.

[0105] Example 2

[0106] In this embodiment, N is 10, and the pixel size of the image sensor of the interferometer is 50 μm. Ten interferometric images with sub-pixel offsets are acquired by the interferometer on the same object under test (e.g., wafer surface).

[0107] In this embodiment, the pixel spacing of the high-resolution grid is 1 / 10 of the pixel spacing of a single-frame interference image, that is, the pixel size is reduced from 50μm to 5μm, and the minimum resolvable fringe spacing is reduced from 100μm to 10μm. The specific implementation of each step is the same as in Embodiment 1, and will not be repeated here.

[0108] In this embodiment, after mapping 10 frames of interferometric images to a high-resolution grid with 10 times the resolution, the proportion of unsampled positions in the high-resolution grid is approximately 90%. In this embodiment, the unsampled positions are filled by solving the Poisson equation with the phase gradient field of the sampled positions as a constraint. Compared with filling methods such as bilinear interpolation that only utilize spatial proximity, the filling method in this embodiment utilizes the physical continuity constraint of the interferometric phase, resulting in a smaller filling error for unsampled positions.

[0109] In this embodiment, based on the pixels of a single frame of interferometric image, the frequency domain range of phase extraction is expanded from [-0.5, 0.5] cycles / pixel of the single frame of interferometric image to [-5, 5] cycles / pixel, which can resolve high-frequency fringe components that exceed the single frame sampling limit and obtain a phase distribution with 10 times the spatial resolution.

[0110] Example 3

[0111] In this embodiment, the subpixel offset is determined by the displacement feedback information of the displacement stage. The object under test is moved by the displacement stage, and interferometric images are acquired at multiple acquisition positions. The displacement between two adjacent acquisitions is less than one pixel. Based on the displacement feedback information of the displacement stage when acquiring each frame of interferometric image, the subpixel offset of each frame of interferometric image relative to the reference frame is directly determined, eliminating the need for offset estimation based on the interferometric image. The specific implementation of the remaining steps is the same as in Embodiment 1, and will not be repeated here.

[0112] In another implementation scenario, the power of the light source fluctuates during multi-frame acquisition, resulting in global phase drift between the interferometric images of each frame. If the images are fused directly without phase alignment, the phase of the fused result will be discontinuous. However, after phase alignment, the interferometric images of each frame are fused under a unified phase reference, resulting in a smaller deviation between the fused result and the true reference value, and better phase continuity.

[0113] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and units involved are not necessarily essential to this application.

[0114] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly manifested as execution by a hardware processor, or as a combination of hardware and software units within the processor. The software units can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor executes the instructions in the memory, combining them with its hardware to complete the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0115] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements some or all of the steps of the method described in the above method embodiments.

[0116] This application also provides a computer program product, which includes a computer program or instructions. When executed, the computer program or instructions implement some or all of the steps of the method described in the above method embodiments. This computer program product can be a software installation package.

Claims

1. An interferometric image processing method, characterized in that, include: Acquire N frames of interference images, wherein the N frames of interference images are interference images acquired for the same object under test, and the N frames of interference images have sub-pixel offsets, and N is an integer greater than 1; Based on the sub-pixel offset between the N interferometric images, the N interferometric images are mapped to a high-resolution grid and fused to obtain a high-resolution interferometric image, wherein the resolution of the high-resolution grid is higher than the resolution of the N interferometric images; Phase extraction is performed on the high-resolution interferometric image to obtain the phase distribution of the measured object.

2. The method according to claim 1, characterized in that, The step of mapping the N frames of interferometric images to a high-resolution grid and fusing them includes: For each pixel in the N frames of the interferometric image, the fusion weight of the pixel is determined according to the spatial distance weight, fringe contrast weight and phase consistency weight of the pixel. Based on the fusion weight of each of the plurality of pixels, the plurality of pixels mapped to the high-resolution grid are weighted and fused. The phase consistency weight is used to characterize the degree of consistency of the phase at a pixel location among the N frames of interferometric images.

3. The method according to claim 2, characterized in that, The fusion weight of a pixel is the product of its spatial distance weight, stripe contrast weight, and phase consistency weight.

4. The method according to claim 2, characterized in that, The phase consistency weight of a pixel is determined based on the phase difference between the phases of the pixel position and the frames, and the phase consistency weight decreases as the phase difference increases.

5. The method according to claim 4, characterized in that, The phase consistency weight is the absolute value of the cosine of half the phase difference.

6. The method according to claim 1 or 2, characterized in that, The step of mapping the N frames of interferometric images to a high-resolution grid and fusing them to obtain a high-resolution interferometric image further includes: Based on the phase of the sampled locations in the high-resolution grid, the phase gradient field is determined; Using the phase gradient field as a constraint, the Poisson equation is solved to obtain a continuous phase field; Based on the continuous phase field, the pixel values ​​of the unsampled locations in the high-resolution grid are determined to fill the high-resolution interferometric image.

7. The method according to claim 6, characterized in that, Determining the pixel values ​​at unsampled locations in the high-resolution grid based on the continuous phase field includes: The pixel values ​​of the unsampled locations are generated based on the continuous phase field and the interference intensity function.

8. The method according to claim 1 or 2, characterized in that, The step of mapping the N frames of interferometric images to a high-resolution grid and fusing them includes: Phase extraction is performed on each of the N frames of interference images to obtain the phase of each frame of interference image; Determine the global phase difference of each interferometric image relative to a reference frame, wherein the reference frame is one of the N interferometric images; Based on the global phase difference, the phase of each frame of the interferometric image is compensated to perform phase alignment of the N frames of the interferometric image; The phase-aligned N-frame interferometric images are mapped onto the high-resolution grid and then fused.

9. An interferometric image processing system, characterized in that, include: The acquisition unit is configured to acquire N frames of interference images, wherein the N frames of interference images are interference images acquired for the same object under test, and the N frames of interference images have sub-pixel offsets, and N is an integer greater than 1. A fusion unit is configured to map the N interferometric images to a high-resolution grid and fuse them according to the sub-pixel offset between the N interferometric images to obtain a high-resolution interferometric image, wherein the resolution of the high-resolution grid is higher than the resolution of the N interferometric images; The extraction unit is configured to perform phase extraction on the high-resolution interferometric image to obtain the phase distribution of the measured object.

10. The system according to claim 9, characterized in that, The fusion unit is further configured to: For each pixel in the N frames of the interferometric image, the fusion weight of the pixel is determined according to the spatial distance weight, fringe contrast weight and phase consistency weight of the pixel. Based on the fusion weight of each of the plurality of pixels, the plurality of pixels mapped to the high-resolution grid are weighted and fused. The phase consistency weight is used to characterize the degree of consistency of the phase at a pixel location among the N frames of interferometric images.

11. The system according to claim 9 or 10, characterized in that, The system further includes a filling unit configured to: Based on the phase of the sampled locations in the high-resolution grid, the phase gradient field is determined; Using the phase gradient field as a constraint, the Poisson equation is solved to obtain a continuous phase field; Based on the continuous phase field, the pixel values ​​of the unsampled locations in the high-resolution grid are determined to fill the high-resolution interferometric image.

12. The system according to claim 9 or 10, characterized in that, The system further includes an alignment unit configured to: Phase extraction is performed on each of the N frames of interference images to obtain the phase of each frame of interference image; Determine the global phase difference of each interferometric image relative to a reference frame, wherein the reference frame is one of the N interferometric images; Based on the global phase difference, the phase of each frame of the interferometric image is compensated to perform phase alignment of the N frames of the interferometric image.

Images