Spectral imaging method, apparatus and device based on edge error suppression

Abstract

The application provides a spectrum imaging method and device based on edge error suppression, equipment, storage medium and program product, relating to the field of spectrum imaging, comprising: obtaining a detection image formed after a target to be observed is modulated by a metasurface array; identifying edge pixels in the detection image, and selecting target non-edge pixels from the neighborhood of the edge pixels; replacing the modulation function of the edge pixels with the modulation function of the target non-edge pixels, constructing a new modulation matrix of the edge superpixel to which the edge pixels belong; based on the new modulation matrix, performing spectrum reconstruction on the detection signal of the edge superpixel to obtain spectrum information, and generating a complete spectrum image of the target to be observed based on the spectrum information. The application can improve the overall spectrum recovery performance.

Description

Technical Field

[0001] This invention relates to the field of spectral imaging technology, and in particular to a spectral imaging method, apparatus, device, storage medium, and program product based on edge error suppression. Background Technology

[0002] Spectral imaging technology, capable of simultaneously acquiring spatial and spectral information of a target, has become an important technique in fields such as biomedical analysis, semiconductor detection, materials characterization, and remote sensing imaging. With the development of on-chip spectrometers and metasurface technology, integrating numerous miniature spectral units onto image sensors to achieve high spatial and high spectral resolution fusion imaging at the chip level has become a significant development direction.

[0003] On-chip hyperspectral imaging typically combines several metasurface units into a "superpixel," assuming that the incident spectrum within the tiny region corresponding to the superpixel remains consistent. Based on this assumption, the incident spectrum can be reconstructed by utilizing multiple different modulation responses within the superpixel. However, in real-world imaging scenarios, image content often exhibits abrupt changes, especially at object edges and material interfaces. When a superpixel crosses such boundaries, its internal pixels will simultaneously contain light signals from different objects or materials, resulting in a non-uniform incident spectrum. This situation directly undermines the fundamental assumption of "regional spectral consistency" upon which spectral reconstruction relies, leading to spectral aliasing, spectral line distortion, and reconstruction errors, becoming a key issue limiting the quality of on-chip hyperspectral imaging. Summary of the Invention

[0004] This invention provides a spectral imaging method, apparatus, device, storage medium, and program product based on edge error suppression, which solves the defect in the prior art where the accuracy of spectral reconstruction is seriously reduced in areas of spectral abrupt change such as the edge of an object due to the inconsistency of the incident spectrum within the superpixel in on-chip spectral imaging.

[0005] This invention provides a spectral imaging method based on edge error suppression, comprising the following steps:

[0006] Acquire a detection image of the target object after modulation by a metasurface array;

[0007] Identify edge pixels in the detected image and select target non-edge pixels from the neighborhood of the edge pixels;

[0008] The modulation function of the edge pixel is replaced with the modulation function of the target non-edge pixel to construct a new modulation matrix for the edge superpixel to which the edge pixel belongs;

[0009] Based on the new modulation matrix, the detection signal of the edge superpixel is spectrally reconstructed to obtain spectral information, and a complete spectral image of the target to be observed is generated based on the spectral information.

[0010] According to the spectral imaging method based on edge error suppression provided by the present invention, the identification of edge pixels in the detected image includes:

[0011] Calculate the local gradient around each pixel in the probe image;

[0012] Pixels in the probed image whose local gradient is greater than an adaptive threshold are identified as edge pixels.

[0013] According to a spectral imaging method based on edge error suppression provided by the present invention, the step of selecting target non-edge pixels from the neighborhood of the edge pixels includes:

[0014] The neighborhood of the edge pixel is determined with the edge pixel as the center;

[0015] Non-edge pixels are selected from the neighborhood, and a set of non-edge pixels is constructed based on the non-edge pixels;

[0016] Based on preset evaluation criteria, target non-edge pixels are selected from the set of non-edge pixels.

[0017] According to the spectral imaging method based on edge error suppression provided by the present invention, the step of selecting target non-edge pixels from the set of non-edge pixels based on a preset evaluation criterion includes:

[0018] Calculate the spatial distance between each non-edge pixel in the non-edge pixel set and the edge pixel;

[0019] Calculate the signal-to-noise ratio of the detection signal for each non-edge pixel in the set of non-edge pixels;

[0020] Calculate the signal fluctuation variance of each non-edge pixel in the non-edge pixel set within a preset time window;

[0021] Weighting coefficients are set according to the preset evaluation criteria;

[0022] Based on the spatial distance, the signal-to-noise ratio of the detection signal, the signal fluctuation variance, and the weighting coefficient, the score result of each non-edge pixel in the non-edge pixel set is calculated;

[0023] Based on the scoring results, target non-edge pixels are selected from the set of non-edge pixels.

[0024] According to the spectral imaging method based on edge error suppression provided by the present invention, the preset evaluation criterion includes one of the following: distance gradient criterion, signal-to-noise ratio priority criterion, and variance enhancement criterion; the step of setting weight coefficients according to the preset evaluation criterion includes:

[0025] When the target to be observed is a large object with high reflectivity, the weighting coefficients are set according to the distance gradient criterion, wherein the distance gradient criterion is based on the spatial distance to determine the distance weighting coefficients.

[0026] When the target to be observed is a low reflectivity object, the weighting coefficient is set according to the signal-to-noise ratio priority criterion, wherein the signal-to-noise ratio priority criterion is based on the signal-to-noise ratio of the detected signal to determine the signal-to-noise ratio weighting coefficient in the weighting coefficient;

[0027] When the target to be observed is a transparent or semi-transparent object, the weighting coefficients are set according to the variance enhancement criterion, wherein the variance enhancement criterion is based on the neighborhood spectral variance to determine the variance weighting coefficients in the weighting coefficients.

[0028] According to the spectral imaging method based on edge error suppression provided by the present invention, the step of replacing the modulation function of the edge pixel with the modulation function of the target non-edge pixel to construct a new modulation matrix of the edge superpixel to which the edge pixel belongs includes:

[0029] Determine the edge superpixel to which the edge pixel belongs;

[0030] Based on the detected image, the superboundary pixels and internal pixels in the edge superpixels are determined;

[0031] Identify and remove the metasurface units corresponding to the superboundary pixels in the edge superpixels;

[0032] Along the unbounded direction of the edge superpixel, the removed edge superpixel is extended and compensated to obtain a new superpixel;

[0033] The modulation function of the edge pixel is replaced with the modulation function of the target non-edge pixel;

[0034] A new modulation matrix for the new superpixel is determined based on the modulation functions of the replaced edge pixels, the modulation functions of the inner pixels, and the modulation functions of the extended pixels of the new superpixel.

[0035] The present invention also provides a spectral imaging device based on edge error suppression, comprising the following modules:

[0036] The acquisition module is used to acquire the detection image formed by the target being observed after being modulated by the metasurface array;

[0037] The determination module is used to identify edge pixels in the probed image and select target non-edge pixels from the neighborhood of the edge pixels;

[0038] The replacement module is used to replace the modulation function of the edge pixel with the modulation function of the target non-edge pixel to construct a new modulation matrix of the edge superpixel to which the edge pixel belongs;

[0039] The reconstruction module is used to perform spectral reconstruction on the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generate a complete spectral image of the target to be observed based on the spectral information.

[0040] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the spectral imaging method based on edge error suppression as described above.

[0041] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the spectral imaging method based on edge error suppression as described above.

[0042] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the spectral imaging method based on edge error suppression as described above.

[0043] This invention provides a spectral imaging method, apparatus, device, storage medium, and program product based on edge error suppression. The method involves acquiring a detection image of the target object after modulation by a metasurface array; identifying edge pixels in the detection image and selecting non-edge pixels from the neighborhood of the edge pixels; replacing the modulation function of the edge pixels with the modulation function of the non-edge pixels to construct a new modulation matrix for the edge superpixel to which the edge pixels belong; and reconstructing the spectrum of the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generating a complete spectral image of the target object based on the spectral information. This invention solves the technical problem of severely reduced spectral reconstruction accuracy in on-chip spectral imaging in regions with spectral abrupt changes, such as object edges, due to inconsistent incident spectra within superpixels. Compared to existing technologies, this invention effectively improves the spectral reconstruction accuracy in edge regions and achieves high-fidelity restoration of full-field spectral information by replacing the modulation function of the edge pixels with the modulation function of the non-edge pixels. Furthermore, this scheme requires no additional hardware compensation or changes to the metasurface unit array layout. By adaptively reusing the modulation response of non-edge pixels, it can significantly reduce the spectral reconstruction error in the edge region, thereby improving the overall spectral recovery performance while maintaining high spatial resolution. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0045] Figure 1 This is one of the flowcharts of the spectral imaging method based on edge error suppression provided by the present invention.

[0046] Figure 2 This is a schematic diagram of the overall structure of the spectral imaging chip for the spectral imaging method based on edge error suppression provided by the present invention.

[0047] Figure 3 This is a schematic diagram of pixel replacement in the spectral imaging method based on edge error suppression provided by the present invention.

[0048] Figure 4 This is a schematic diagram of the spectral reconstruction process of the spectral imaging method based on edge error suppression provided by the present invention.

[0049] Figure 5 This is a schematic diagram of the extended compensation of the spectral imaging method based on edge error suppression provided by the present invention.

[0050] Figure 6 This is the second schematic diagram of the spectral imaging method based on edge error suppression provided by the present invention.

[0051] Figure 7 This is a schematic diagram of the structure of the spectral imaging device based on edge error suppression provided by the present invention.

[0052] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0054] The following is combined with Figure 1 and Figure 6 The present invention describes a spectral imaging method based on edge error suppression, which is applicable to any spectral imaging based on edge error suppression. The subject executing this method can be an electronic device or a spectral imaging device based on edge error suppression installed in the electronic device. The spectral imaging device based on edge error suppression can be implemented by software, hardware, or a combination of both.

[0055] Figure 1 This is one of the flowcharts of the spectral imaging method based on edge error suppression provided by the present invention, such as... Figure 1 As shown, the method includes the following:

[0056] Step 101: Obtain the detection image of the target to be observed after modulation by the metasurface array.

[0057] It should be noted that, as Figure 2 The overall structure of the spectral imaging chip shown consists of a metasurface layer, an image sensor, and a signal processing unit arranged from top to bottom. The metasurface layer (i.e., the metasurface array) serves as the front-end processing component for incident light, undertaking the task of spectral modulation of the incident light. The image sensor, arranged accordingly, is used to collect the light signal modulated by the metasurface layer and convert it into a data signal. The signal processing unit located in the lower layer receives the data transmitted by the image sensor and then performs subsequent processing operations such as signal analysis and edge error suppression.

[0058] It should be noted that the metasurface array is composed of a large number of periodically arranged "superpixels," each of which integrates multiple nanostructure units (i.e., metasurface units) with preset differences in spectral modulation functions. To achieve efficient coupling between the metasurface array and the detector, a high-precision wafer bonding process is used to align the metasurface array and the CMOS pixel at the sub-micron level, within approximately 0.5 cm. 2 The array integrates millions of spectral units in a compact layout, ensuring a high degree of consistency in the physical parameters of each unit in the large-scale array. When broadband incident light emitted from the target to be observed illuminates this array, each metasurface unit within each superpixel simultaneously weights and modulates the incident spectrum with different weights. The integrated area array photodetector then synchronously records the total light intensity modulated at each spatial location, ultimately forming a two-dimensional signal intensity distribution map (i.e., the detection image).

[0059] Understandably, the detected image is not an ordinary image that reflects the intuitive shape of the target, but a "coded image" in which each pixel value represents the inner product of the original incident spectrum and a specific metasurface modulation function (i.e., the modulation function corresponding to the metasurface unit inside the superpixel). Its information form is a highly compressed and non-intuitive mixed signal.

[0060] Step 102: Identify edge pixels in the probe image and select target non-edge pixels from the neighborhood of the edge pixels.

[0061] It should be noted that edge detection can be performed on the probe image to identify edge pixels in the probe image. Specifically, gradient operators (such as Sobel, Roberts, or arbitrary algorithms based on local changes) can be used to calculate the local gradient G(x,y) around pixel p(x,y) in the probe image. When G(x,y)>T (T is a threshold adaptively determined according to the noise level of the probe image), the pixel can be marked as an edge pixel.

[0062] Step 103: Replace the modulation function of the edge pixel with the modulation function of the target non-edge pixel to construct a new modulation matrix of the edge superpixel to which the edge pixel belongs.

[0063] Understandably, because different pixels (i.e., metasurface units in superpixels) in a metasurface array can share certain types of modulation functions, this reusable structure makes the modulation responses between neighboring pixels predictable, thus ensuring the physical feasibility of alternative solutions. Specifically, the metasurface array pre-defines multiple compatible metasurface unit layouts during the design phase. The physical structural parameters of metasurface units of the same type are completely unified, and complementary and interconnected response logic is formed between units of different spectral bands. In this way, edge superpixels can directly retrieve the modulation functions of the same template units of their neighboring superpixels to replace the modulation functions of the edge pixels, achieving spectral inversion without modifying the hardware.

[0064] In specific implementations, such as Figure 3 As shown in the diagram, this schematic illustrates the process of replacing the modulation function of edge pixels: the pixels marked with grid fill are the edge pixels to be replaced, and their respective regions are edge superpixels; the pixels marked with "selected non-edge pixels" on the left are the target non-edge pixels. By retrieving the modulation function of the target non-edge pixel, the original modulation function of the edge pixel is directly replaced. Combined with the modulation functions of other pixels within the edge superpixel, a new modulation matrix for the edge superpixel can be constructed. This replacement method relies on the cell multiplexing characteristics of metasurface arrays, and can optimize the modulation performance of edge pixels without hardware adjustments.

[0065] Step 104: Based on the new modulation matrix, perform spectral reconstruction on the detection signal of the edge superpixel to obtain spectral information, and generate a complete spectral image of the target to be observed based on the spectral information.

[0066] It should be noted that the detection signal of the edge superpixel refers to the detection signal corresponding to the edge superpixel in the detection image. During the spectral reconstruction stage, the detection signal of the edge superpixel and the new modulation matrix need to be linearly solved (e.g., using inversion strategies such as least squares method or sparse reconstruction algorithm) to obtain high-precision spectral information of the region corresponding to the edge superpixel. Simultaneously, combined with the conventional spectral reconstruction results of superpixels in other non-edge regions of the metasurface array, a complete spectral image of the target object is finally generated.

[0067] In specific implementations, such as Figure 4 The schematic diagram of the spectral reconstruction process shown first involves inputting a detection signal S, which is then passed to a modified modulation matrix M' (i.e., the new modulation matrix). Subsequently, a spectral inversion algorithm is used to process the detection signal of the new modulation matrix, and finally, the reconstructed spectrum is output. This process realizes the transformation from a measurement signal to an accurate spectral result.

[0068] This invention provides a spectral imaging method, apparatus, device, storage medium, and program product based on edge error suppression. The method involves acquiring a detection image of the target object after modulation by a metasurface array; identifying edge pixels in the detection image and selecting non-edge pixels from the neighborhood of the edge pixels; replacing the modulation function of the edge pixels with the modulation function of the non-edge pixels to construct a new modulation matrix for the edge superpixel to which the edge pixels belong; and reconstructing the spectrum of the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generating a complete spectral image of the target object based on the spectral information. This invention solves the technical problem of severely reduced spectral reconstruction accuracy in on-chip spectral imaging in regions with spectral abrupt changes, such as object edges, due to inconsistent incident spectra within superpixels. Compared to existing technologies, this invention effectively improves the spectral reconstruction accuracy in edge regions and achieves high-fidelity restoration of full-field spectral information by replacing the modulation function of the edge pixels with the modulation function of the non-edge pixels. Furthermore, this scheme requires no additional hardware compensation or changes to the metasurface unit array layout. By adaptively reusing the modulation response of non-edge pixels, it can significantly reduce the spectral reconstruction error in the edge region, thereby improving the overall spectral recovery performance while maintaining high spatial resolution.

[0069] Based on any of the above embodiments, step 102 includes:

[0070] Calculate the local gradient around each pixel in the probe image;

[0071] Pixels in the probed image whose local gradient is greater than an adaptive threshold are identified as edge pixels.

[0072] It should be noted that calculating the local gradient around each pixel in the probe image essentially involves a differential quantification analysis of the neighborhood signal distribution of each pixel within the probe image. The magnitude of the gradient value is positively correlated with the degree of abrupt change in the spectral or spatial information of the pixel location: a higher gradient value indicates a more significant difference in light intensity and spectral composition between the pixel and its neighboring pixels, meaning that the location is more likely to be in the edge region of the observed target; a lower gradient value indicates a more uniform signal distribution in the pixel's neighborhood, corresponding to a non-edge region of the target. Therefore, edge pixels in the probe image can be determined based on the local gradient around each pixel. Specifically, an adaptive threshold can be set as the quantification standard for edge determination according to the actual situation. When the local gradient of a pixel is greater than the adaptive threshold, the pixel can be directly determined as an edge pixel.

[0073] Understandably, the adaptive threshold is not a fixed value; it can be dynamically generated based on the global gradient statistics of the probed image (such as gradient mean and variance) or the local neighborhood gradient distribution. Since spurious gradient signals caused by imaging noise and random crosstalk are usually small in value and discretely distributed, their gradient values ​​are much lower than the peak values ​​of the gradients in the true edge regions. By dynamically dividing the adaptive threshold, the boundary between the "effective gradient signal" and the "interference gradient signal" can be accurately defined, directly filtering out pixels corresponding to spurious gradients with values ​​below the threshold, thus preventing such interference signals from being misjudged as edge pixels.

[0074] In a practical implementation, gradient operators (such as Sobel, Roberts, or any algorithm based on local changes) can be used to calculate the local gradient G(x,y) around pixel p(x,y) in the probe image. When G(x,y)>T (T is a threshold adaptively determined according to the noise level of the probe image), the pixel can be marked as an edge pixel.

[0075] The spectral imaging method based on edge error suppression provided in this invention constructs a new modulation matrix adapted to edge superpixels by accurately locating edge pixels. This can fundamentally cancel spectral distortion caused by boundary diffraction and pixel crosstalk, effectively improve the reconstruction accuracy of edge regions, and achieve high-fidelity spectral imaging across the entire field of view.

[0076] Based on any of the above embodiments, step 103 includes:

[0077] Determine the edge superpixel to which the edge pixel belongs;

[0078] Based on the detected image, the superboundary pixels and internal pixels in the edge superpixels are determined;

[0079] Identify and remove the metasurface units corresponding to the superboundary pixels in the edge superpixels;

[0080] Along the unbounded direction of the edge superpixel, the removed edge superpixel is extended and compensated to obtain a new superpixel;

[0081] The modulation function of the edge pixel is replaced with the modulation function of the target non-edge pixel;

[0082] A new modulation matrix for the new superpixel is determined based on the modulation functions of the replaced edge pixels, the modulation functions of the inner pixels, and the modulation functions of the extended pixels of the new superpixel.

[0083] It should be noted that boundary pixels refer to pixels located at the edge of the physical contour of the target being observed. Their spectral signals contain characteristics of both the target body and the background, and are susceptible to cross-boundary interference, leading to imaging errors. Superboundary pixels are pixels in the edge superpixels that extend beyond the physical boundary of the target and belong entirely to the background area. Their corresponding metasurface units introduce invalid background signals, causing confusion in the spectral features within the superpixel, which needs to be eliminated through identification and removal. Inner pixels are pixels in the edge superpixels that are within the physical boundary of the target. Their spectral signals are stable and can accurately reflect the characteristics of the target body, requiring no additional correction.

[0084] It should be noted that when a superpixel crosses an object boundary, the super-boundary pixels it contains will cause spectral signal cross-interference with the background pixels, directly leading to errors such as edge imaging distortion and blurring. Furthermore, because edge pixels are located at the object's edge, their modulation functions are easily affected by boundary noise, making it difficult to accurately reflect the true spectral characteristics. Therefore, it is necessary to first determine the edge superpixel to which the edge pixel belongs, then determine the super-boundary pixels and internal pixels within the edge superpixel based on the detected image. For super-boundary pixels, their corresponding metasurface units are identified and removed to block cross-boundary signal interference. Simultaneously, compensation is extended along the non-cross-boundary direction to complete the spatial structure of the removed superpixel, resulting in a new superpixel. For edge pixels, their modulation functions are replaced with the modulation functions of the target non-edge pixels, introducing error-free, high-quality modulation features. Finally, the modulation functions of the replaced edge pixels, internal pixels, and extended pixels of the new superpixel are fused to determine a new modulation matrix. This solves the cross-boundary interference problem of super-boundary pixels and optimizes the modulation performance of edge pixels, achieving precise suppression of edge errors.

[0085] In specific implementations, such as Figure 5 As shown, when a superpixel crosses an object boundary, it needs to be removed from the superpixel pool. After removal, the region shape is no longer square, and the number of sampling points is reduced, making it difficult to meet the requirements of spectral inversion. To restore sufficient sampling, compensation can be achieved by extending along the direction that did not cross the boundary. For example, if 8 rows remain after removal, it can be extended to 12 columns in another direction, forming 8... The new superpixel is 12. The extended portion comes from a spectrally consistent neighboring region, which can complete the modulation matrix and stabilize the inversion.

[0086] The spectral imaging method based on edge error suppression provided in this invention distinguishes between boundary pixels, super-boundary pixels, and internal pixels and performs differentiated processing to eliminate super-boundary pixel interference, optimize edge pixel modulation functions, and complete super-pixel structures. This effectively suppresses edge imaging distortion and error diffusion, thereby improving the clarity and accuracy of spectral imaging.

[0087] Figure 6 This is the second schematic diagram of the spectral imaging method based on edge error suppression provided by the present invention, as shown below. Figure 6As shown, step 102 also includes steps 1021 to 1023:

[0088] Step 1021: Determine the neighborhood of the edge pixel with the edge pixel as the center.

[0089] In the specific implementation, a square region window (i.e., a neighborhood, which can be 3×3 or 5×5) is defined with the edge pixel as the center and the period length of the superpixel to which the edge pixel belongs as the side length. Then, a set of non-edge pixels is selected from this neighborhood. This ensures that the modulation functions corresponding to the non-edge pixels in the non-edge pixel set are all based on relatively single and stable spectral signals, thereby guaranteeing spectral resolution. The window size of this neighborhood is strongly bound to the superpixel period length, which avoids the problem of insufficient non-edge pixel sample size due to an excessively small window range, and also prevents the introduction of heterogeneous spectral signal interference from distant pixels due to an excessively large window range. At the same time, the fully covered superpixel structure ensures that the metasurface units corresponding to the non-edge pixels are all in a uniform light field modulation environment. The consistency of their output signals can significantly reduce the computational complexity of subsequent target non-edge pixel screening and improve the efficiency and accuracy of edge error suppression.

[0090] Step 1022: Filter out non-edge pixels from the neighborhood and construct a non-edge pixel set based on the non-edge pixels.

[0091] It should be noted that non-edge pixels refer to pixels in the neighborhood that are made of the same uniform material as the object (i.e. the target to be observed). These pixels do not have significant abrupt changes in spectral characteristics, gray values ​​and texture distribution with the pixels in their surrounding neighborhood, and exhibit a high degree of spatial-spectral consistency.

[0092] Step 1023: Select target non-edge pixels from the set of non-edge pixels based on preset evaluation criteria.

[0093] It should be noted that the preset evaluation criteria may include at least one of the following: the spatial distance between non-edge pixels and edge pixels, the signal-to-noise ratio (SNR) of the detection signal of non-edge pixels, and the stability of the modulation function corresponding to the non-edge pixels. The "spatial distance between non-edge pixels and edge pixels" refers to the geometric distance between the superpixel corresponding to the candidate non-edge pixel and the superpixel to which the edge pixel belongs in the metasurface array. It is usually expressed in units of superpixel period length (e.g., "1 period length" represents adjacent superpixels). Its core function is to measure the similarity of their local optical field environments; the closer the distance, the stronger the optical field consistency, thus reducing the error in modulation matrix replacement. The "detection signal SNR" refers to the ratio of the effective spectral encoded signal to the noise signal in the detector output signal corresponding to the non-edge pixel. The calculation formula is "SNR = 10 × lg (effective signal power / noise power)". This indicator directly reflects the proportion of effective information in the signal. A high SNR means that the signal is less affected by noise interference and can more accurately characterize the modulation pattern of the superpixel. "Modulation function stability" refers to the degree of fluctuation of the modulation function of the superpixel corresponding to the non-edge pixel within a preset time window. It is usually measured by the variance of the modulation coefficient. The smaller the variance, the higher the stability, which means that the modulation rule of the superpixel is more reliable and the edge error compensation is less likely to fail due to signal drift when its modulation function is reused.

[0094] In practice, all non-edge pixels in the non-edge pixel set can be scored based on preset evaluation criteria (such as spatial distance, signal-to-noise ratio, and modulation function stability), and finally the most suitable target non-edge pixel can be selected from the non-edge pixel set based on the scoring results.

[0095] The spectral imaging method based on edge error suppression provided in this embodiment can accurately lock the non-edge pixels with the best light field compatibility with the edge pixels from the non-edge pixel set through preset evaluation criteria. This provides a highly reliable reference benchmark for the correction of the edge superpixel modulation matrix, effectively reduces the spectral coding distortion in the edge region, and improves the spectral resolution and spatial consistency of the overall imaging.

[0096] Based on any of the above embodiments, step 1023 includes:

[0097] Calculate the spatial distance between each non-edge pixel in the non-edge pixel set and the edge pixel;

[0098] Calculate the signal-to-noise ratio of the detection signal for each non-edge pixel in the set of non-edge pixels;

[0099] Calculate the signal fluctuation variance of each non-edge pixel in the non-edge pixel set within a preset time window;

[0100] Weighting coefficients are set according to the preset evaluation criteria;

[0101] Based on the spatial distance, the signal-to-noise ratio of the detection signal, the signal fluctuation variance, and the weighting coefficient, the score result of each non-edge pixel in the non-edge pixel set is calculated;

[0102] Based on the scoring results, target non-edge pixels are selected from the set of non-edge pixels.

[0103] It should be noted that the weighting coefficients can be set according to the specific imaging scenario. Specifically, in low-light scenarios, the weight of the signal-to-noise ratio of the detection signal can be increased to prioritize the selection of non-edge pixels with strong noise resistance to avoid modulation matrix errors caused by low-light noise. In dynamic scenarios, the weight of signal fluctuation variance can be increased to focus on selecting pixels with high modulation function stability and avoid the influence of signal drift caused by target motion. For large object imaging scenarios, since the metasurface array covers a wider object area, the spatial light field consistency between edge pixels and surrounding non-edge pixels is stronger. The weight of spatial distance can be appropriately increased to prioritize the selection of non-edge pixels that are adjacent to edge pixels to ensure the local adaptability of the modulation rules. For small object imaging scenarios, the number of superpixels occupied by the object in the metasurface array is small, and the number of available non-edge pixel samples around the edge pixels is limited. The weight of spatial distance can be reduced, and the weights of the signal-to-noise ratio of the detection signal and the signal fluctuation variance can be increased to prioritize ensuring that the selected non-edge pixels have high-quality signals and stable modulation characteristics, and to avoid error compensation failure due to insufficient number of non-edge pixels in the non-edge pixel set.

[0104] In the specific implementation, the geometric center of the superpixel to which the edge pixel belongs is taken as the reference origin (x0, y0), and the geometric center of the superpixel to which the non-edge pixel belongs is taken as the target point (xi, yi). The superpixel period length L of the metasurface array is used as the basic unit of measurement, and the normalized spatial distance is calculated using the Euclidean distance formula, which is: After normalization, the spatial distance between adjacent superpixels is 1, and the spatial distance between diagonally adjacent superpixels is... This value can directly reflect the similarity of the local light field environment of two superpixels; the smaller the distance, the stronger the consistency of the light field.

[0105] In the specific implementation, for the detector output signal corresponding to non-edge pixels, N consecutive sample values ​​{S1,S2,...,SN} are collected within a preset sampling time. The signal-to-noise ratio (SNR) is calculated through the following steps: ① Calculate the mean of the sample values, which represents the reference strength of the effective signal; ② Calculate the noise power, which is the average of the sum of squares of the deviations between the sample values ​​and the mean; ③ Calculate the effective signal power based on the reference strength and the average value, and finally obtain the signal-to-noise ratio of the detection signal through logarithmic transformation.

[0106] In the specific implementation, the signal fluctuation variance directly characterizes the stability of the superpixel modulation function corresponding to non-edge pixels. The calculation object is the core modulation output parameters of the superpixel within a preset time window T (such as total light intensity signal and phase modulation coefficient). The specific steps are as follows: ① Collect M consecutive modulation parameter sample values ​​{F1,F2,...,FM} within the time window T; ② Calculate the mean of the sample values; ③ Calculate the signal fluctuation variance of the sample values ​​based on the mean. The smaller the variance, the lower the fluctuation degree of the modulation function and the higher the stability.

[0107] In the specific implementation, after calculating the spatial distance, signal-to-noise ratio of the detection signal, and signal fluctuation variance within a preset time window for each non-edge pixel in the non-edge pixel set, these calculation results are multiplied by the corresponding weight coefficients and then summed to obtain the score result for each non-edge pixel.

[0108] The spectral imaging method based on edge error suppression provided in this embodiment calculates the spatial distance between non-edge pixels and edge pixels, the signal-to-noise ratio of the detection signal, and the signal fluctuation variance within a preset time window. It also sets weight coefficients based on preset evaluation criteria adapted to the characteristics of the target to be observed. Then, it calculates pixel scores based on the above parameters and weight coefficients to screen target non-edge pixels. This method can specifically strengthen the weight of non-edge pixels with high feature matching with edge pixels, effectively suppressing edge imaging errors caused by spatial distance deviation, signal noise interference, and signal fluctuation.

[0109] Based on any of the above embodiments, the preset evaluation criterion includes one of the distance gradient criterion, signal-to-noise ratio priority criterion, and variance enhancement criterion; the step of setting weight coefficients according to the preset evaluation criterion includes:

[0110] When the target to be observed is a large object with high reflectivity, the weighting coefficients are set according to the distance gradient criterion, wherein the distance gradient criterion is based on the spatial distance to determine the distance weighting coefficients.

[0111] When the target to be observed is a low reflectivity object, the weighting coefficient is set according to the signal-to-noise ratio priority criterion, wherein the signal-to-noise ratio priority criterion is based on the signal-to-noise ratio of the detected signal to determine the signal-to-noise ratio weighting coefficient in the weighting coefficient;

[0112] When the target to be observed is a transparent or semi-transparent object, the weighting coefficients are set according to the variance enhancement criterion, wherein the variance enhancement criterion is based on the neighborhood spectral variance to determine the variance weighting coefficients in the weighting coefficients.

[0113] It should be noted that for the distance gradient criterion, there is often a wide transition zone in the edge area of large objects with high reflectivity. Spatial distance is the key factor affecting the correlation between edge pixels and non-edge pixels. Therefore, the spatial distance weight is given a high priority to avoid error diffusion in the transition zone. For the signal-to-noise ratio priority criterion, the detection signal of low-reflectivity objects is weak and the noise proportion is high. Noise interference is the main reason for the blurring of edge details. Therefore, the proportion of the signal-to-noise ratio weight is preferentially increased to strengthen the dominant role of the effective signal in weight calculation. For the variance enhancement criterion, the boundary signal of transparent and translucent objects is easily interfered by the background. The spectral variance of the neighborhood can reflect the degree of feature difference between pixels. The variance weight is given a high priority to accurately distinguish the pixel features of the object boundary and the background.

[0114] In specific implementation, the implementation method of the distance gradient criterion is as follows: calculate the Euclidean distance d between non-edge pixels and edge pixels, and divide the distance into three intervals according to the width of the edge transition zone of large objects with high reflectivity, namely the near-edge area (d ≤ d1), the middle-edge transition area (d1 < d ≤ d2), and the far-edge smoothing area (d > d2); secondly, assign initial distance weights w with decreasing gradients to the three intervals d1 (0.7 - 0.9), w d2 (0.4 - 0.6), w d3 (0.1 - 0.3), ensuring that the weight proportion of non-edge pixels closer to the edge is higher; finally, allocate the signal-to-noise ratio weight w according to a fixed proportion s (0.2 - 0.3), the variance weight w v (0.1 - 0.2).

[0115] In specific implementation, the implementation method of the signal-to-noise ratio priority criterion is as follows: calculate the signal-to-noise ratio SNR of non-edge pixels, and set the signal-to-noise ratio threshold SNR0 (such as 30); secondly, based on the signal-to-noise ratio priority principle, assign a high signal-to-noise ratio weight w s1 (0.6 - 0.8) to pixels with SNR ≥ SNR0, and assign a low signal-to-noise ratio weight w s2 (0.3 - 0.5) to pixels with SNR < SNR0; finally, configure the distance weight w according to a fixed proportion d (0.2 - 0.3) and the variance weight w v (0.1 - 0.2).

[0116] In specific implementation, calculate the signal fluctuation variance Var within the neighborhood of non-edge pixels, and set the variance threshold Var0 (such as 0.8); secondly, based on the variance enhancement principle, assign a high variance weight w v1 (0.7 - 0.9) to pixels with Var ≥ Var0 (large boundary feature differences), and assign a low variance weight w to pixels with Var < Var0 (small feature differences)v2 (0.4-0.6); Finally, allocate distance weights w according to the proportion. d (0.1-0.2), signal-to-noise ratio weight w s (0.1-0.2).

[0117] The spectral imaging method based on edge error suppression provided by this invention precisely suppresses edge errors of various targets by matching exclusive preset evaluation criteria to targets with different properties, quantifying and binding spatial distance, signal-to-noise ratio, variance and weight coefficients, improving imaging resolution and feature restoration, and ensuring the accuracy and stability of imaging in complex scenes.

[0118] The spectral imaging device based on edge error suppression provided by the present invention will be described below. The spectral imaging device based on edge error suppression described below can be referred to in correspondence with the spectral imaging method based on edge error suppression described above. Figure 7 As shown, the spectral imaging device based on edge error suppression includes:

[0119] The acquisition module 10 is used to acquire the detection image formed by the target being observed after being modulated by the metasurface array;

[0120] The determining module 20 is used to identify edge pixels in the detected image and select target non-edge pixels from the neighborhood of the edge pixels;

[0121] Replacement module 30 is used to replace the modulation function of the edge pixel with the modulation function of the target non-edge pixel to construct a new modulation matrix of the edge superpixel to which the edge pixel belongs;

[0122] The reconstruction module 40 is used to perform spectral reconstruction on the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generate a complete spectral image of the target to be observed based on the spectral information.

[0123] Optionally, the determining module 20 is further configured to:

[0124] Calculate the local gradient around each pixel in the probe image;

[0125] Pixels in the probed image whose local gradient is greater than an adaptive threshold are identified as edge pixels.

[0126] Optionally, the determining module 20 is further configured to:

[0127] The neighborhood of the edge pixel is determined with the edge pixel as the center;

[0128] Non-edge pixels are selected from the neighborhood, and a set of non-edge pixels is constructed based on the non-edge pixels;

[0129] Based on preset evaluation criteria, target non-edge pixels are selected from the set of non-edge pixels.

[0130] Optionally, the determining module 20 is further configured to:

[0131] Calculate the spatial distance between each non-edge pixel in the non-edge pixel set and the edge pixel;

[0132] Calculate the signal-to-noise ratio of the detection signal for each non-edge pixel in the set of non-edge pixels;

[0133] Calculate the signal fluctuation variance of each non-edge pixel in the non-edge pixel set within a preset time window;

[0134] Weighting coefficients are set according to the preset evaluation criteria;

[0135] Based on the spatial distance, the signal-to-noise ratio of the detection signal, the signal fluctuation variance, and the weighting coefficient, the score result of each non-edge pixel in the non-edge pixel set is calculated;

[0136] Based on the scoring results, target non-edge pixels are selected from the set of non-edge pixels.

[0137] Optionally, the preset evaluation criteria include one of the distance gradient criterion, signal-to-noise ratio priority criterion, and variance enhancement criterion; the determining module 20 is further configured to:

[0138] When the target to be observed is a large object with high reflectivity, the weighting coefficients are set according to the distance gradient criterion, wherein the distance gradient criterion is based on the spatial distance to determine the distance weighting coefficients.

[0139] When the target to be observed is a low reflectivity object, the weighting coefficient is set according to the signal-to-noise ratio priority criterion, wherein the signal-to-noise ratio priority criterion is based on the signal-to-noise ratio of the detected signal to determine the signal-to-noise ratio weighting coefficient in the weighting coefficient;

[0140] When the target to be observed is a transparent or semi-transparent object, the weighting coefficients are set according to the variance enhancement criterion, wherein the variance enhancement criterion is based on the neighborhood spectral variance to determine the variance weighting coefficients in the weighting coefficients.

[0141] Optionally, the replacement module 30 is further configured to:

[0142] Determine the edge superpixel to which the edge pixel belongs;

[0143] Based on the detected image, the superboundary pixels and internal pixels in the edge superpixels are determined;

[0144] Identify and remove the metasurface units corresponding to the superboundary pixels in the edge superpixels;

[0145] Along the unbounded direction of the edge superpixel, the removed edge superpixel is extended and compensated to obtain a new superpixel;

[0146] The modulation function of the edge pixel is replaced with the modulation function of the target non-edge pixel;

[0147] A new modulation matrix for the new superpixel is determined based on the modulation functions of the replaced edge pixels, the modulation functions of the inner pixels, and the modulation functions of the extended pixels of the new superpixel.

[0148] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute a spectral imaging method based on edge error suppression. This method includes: acquiring a detection image of the target object after modulation by a metasurface array; identifying edge pixels in the detection image and selecting non-edge pixels of the target from the neighborhood of the edge pixels; replacing the modulation function of the edge pixels with the modulation function of the non-edge pixels of the target to construct a new modulation matrix for the edge superpixel to which the edge pixels belong; and based on the new modulation matrix, performing spectral reconstruction on the detection signal of the edge superpixel to obtain spectral information, and generating a complete spectral image of the target object based on the spectral information.

[0149] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0150] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the spectral imaging method based on edge error suppression provided by the above methods. The method includes: acquiring a detection image of the target to be observed after modulation by a metasurface array; identifying edge pixels in the detection image and selecting target non-edge pixels from the neighborhood of the edge pixels; replacing the modulation function of the edge pixels with the modulation function of the target non-edge pixels to construct a new modulation matrix of the edge superpixel to which the edge pixels belong; performing spectral reconstruction on the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generating a complete spectral image of the target to be observed based on the spectral information.

[0151] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the spectral imaging method based on edge error suppression provided by the above methods. The method includes: acquiring a detection image of the target to be observed after modulation by a metasurface array; identifying edge pixels in the detection image and selecting non-edge pixels of the target from the neighborhood of the edge pixels; replacing the modulation function of the edge pixels with the modulation function of the non-edge pixels of the target to construct a new modulation matrix of the edge superpixel to which the edge pixels belong; performing spectral reconstruction on the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generating a complete spectral image of the target to be observed based on the spectral information.

[0152] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0153] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A spectral imaging method based on edge error suppression, characterized in that, include: Acquire a detection image of the target object after modulation by a metasurface array; The metasurface array is composed of a large number of periodically arranged superpixels, and each superpixel integrates multiple metasurface units with preset differences in spectral modulation functions. Identify edge pixels in the detected image and select target non-edge pixels from the neighborhood of the edge pixels; The edge pixels are pixels located at the physical boundary of the target to be observed; The modulation function of the edge pixel is replaced by the modulation function of the target non-edge pixel to construct a new modulation matrix for the edge superpixel corresponding to the edge pixel; each pixel value in the probe image represents the inner product of the original incident spectrum and the modulation function; Based on the new modulation matrix, the detection signal of the edge superpixel is spectrally reconstructed to obtain spectral information, and a complete spectral image of the target to be observed is generated based on the spectral information. The step of replacing the modulation function of the edge pixel with the modulation function of the target non-edge pixel to construct a new modulation matrix for the edge superpixel corresponding to the edge pixel includes: Determine the edge superpixel corresponding to the edge pixel; Based on the detected image, the superboundary pixels and internal pixels in the edge superpixels are determined; the superboundary pixels are pixels that extend beyond the target physical boundary and belong entirely to the background area; the internal pixels are pixels that are within the target physical boundary. Identify and remove the metasurface units corresponding to the superboundary pixels in the edge superpixels; Along the unbounded direction of the edge superpixel, the removed edge superpixel is extended and compensated to obtain a new superpixel; The modulation function of the edge pixel is replaced with the modulation function of the target non-edge pixel; A new modulation matrix for the new superpixel is determined based on the modulation functions of the replaced edge pixels, the modulation functions of the inner pixels, and the modulation functions of the extended pixels of the new superpixel.

2. The spectral imaging method based on edge error suppression according to claim 1, characterized in that, The process of identifying edge pixels in the detected image includes: Calculate the local gradient around each pixel in the probe image; the local gradient is used to perform differential quantization analysis on the neighborhood signal distribution of each pixel in the probe image; Pixels in the probed image whose local gradient is greater than an adaptive threshold are identified as edge pixels.

3. The spectral imaging method based on edge error suppression according to claim 1, characterized in that, Selecting the target non-edge pixel from the neighborhood of the edge pixel includes: The neighborhood of the edge pixel is determined with the edge pixel as the center; Non-edge pixels are selected from the neighborhood, and a set of non-edge pixels is constructed based on the non-edge pixels; Based on preset evaluation criteria, target non-edge pixels are selected from the set of non-edge pixels.

4. The spectral imaging method based on edge error suppression according to claim 3, characterized in that, The step of selecting target non-edge pixels from the set of non-edge pixels based on preset evaluation criteria includes: Calculate the spatial distance between each non-edge pixel in the non-edge pixel set and the edge pixel; Calculate the signal-to-noise ratio of the detection signal for each non-edge pixel in the set of non-edge pixels; Calculate the signal fluctuation variance of each non-edge pixel in the non-edge pixel set within a preset time window; Weighting coefficients are set according to the preset evaluation criteria; Based on the spatial distance, the signal-to-noise ratio of the detection signal, the signal fluctuation variance, and the weighting coefficient, the score result of each non-edge pixel in the non-edge pixel set is calculated; Based on the scoring results, target non-edge pixels are selected from the set of non-edge pixels.

5. The spectral imaging method based on edge error suppression according to claim 4, characterized in that, The preset evaluation criteria include one of the following: distance gradient criterion, signal-to-noise ratio priority criterion, and variance enhancement criterion; the step of setting weight coefficients according to the preset evaluation criteria includes: The distance gradient criterion determines the distance weight coefficient in the weight coefficients based on the spatial distance priority. The signal-to-noise ratio priority criterion determines the signal-to-noise ratio weighting coefficients in the weighting coefficients based on the signal-to-noise ratio of the detected signal. The variance enhancement criterion is based on determining the variance weight coefficients in the weight coefficients according to the neighborhood spectral variance.

6. A spectral imaging device based on edge error suppression, characterized in that, include: The acquisition module is used to acquire the detection image formed by the target being observed after being modulated by the metasurface array; The metasurface array is composed of a large number of periodically arranged superpixels, and each superpixel integrates multiple metasurface units with preset differences in spectral modulation functions. The determination module is used to identify edge pixels in the probed image and select target non-edge pixels from the neighborhood of the edge pixels; The edge pixels are pixels located at the physical boundary of the target to be observed; The replacement module is used to replace the modulation function of the edge pixel with the modulation function of the target non-edge pixel to construct a new modulation matrix of the edge superpixel corresponding to the edge pixel; each pixel value in the probe image represents the inner product of the original incident spectrum and the modulation function; The reconstruction module is used to perform spectral reconstruction on the detection signal of the edge superpixel based on the new modulation matrix to obtain spectral information, and generate a complete spectral image of the target to be observed based on the spectral information; The replacement module is used for: Determine the edge superpixel corresponding to the edge pixel; Based on the detected image, the superboundary pixels and internal pixels in the edge superpixels are determined; the superboundary pixels are pixels that extend beyond the target physical boundary and belong entirely to the background area; the internal pixels are pixels that are within the target physical boundary. Identify and remove the metasurface units corresponding to the superboundary pixels in the edge superpixels; Along the unbounded direction of the edge superpixel, the removed edge superpixel is extended and compensated to obtain a new superpixel; The modulation function of the edge pixel is replaced with the modulation function of the target non-edge pixel; A new modulation matrix for the new superpixel is determined based on the modulation functions of the replaced edge pixels, the modulation functions of the inner pixels, and the modulation functions of the extended pixels of the new superpixel.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the spectral imaging method based on edge error suppression as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the spectral imaging method based on edge error suppression as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the spectral imaging method based on edge error suppression as described in any one of claims 1 to 5.

Images