A method and device for non-invasive large field-of-view imaging through scattering media

Abstract

The present application belongs to the field of optical scattering imaging technology, and realizes the imaging through scattering medium based on optical memory effect.When the imaging target exceeds the memory effect range, the imaging failure is usually caused, thereby limiting the imaging field of view.The present application proposes a method and device for realizing non-invasive large field of view imaging through scattering medium based on PSF estimation and correlation operation.When the target is too large to exceed the memory effect range, the phase recovery algorithm is used to reconstruct the image of the sub-target O i of the target O hidden in the scattering medium, the positions of all sub-targets are obtained based on the point spread function estimation and autocorrelation and cross-correlation operation, the image of the sub-targets is spliced according to the positions, and the non-invasive large field of view imaging is realized.The present application can also realize non-invasive large field of view super-resolution imaging beyond the memory effect range, and can simultaneously consider the imaging field of view and imaging resolution.The present application has wide application in the fields of biomedical imaging and optical microscopic imaging.

Description

Technical fields:

[0001] This invention belongs to the field of optical scattering imaging technology, specifically relating to a method and device for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations. Background technology:

[0002] In traditional optical imaging, what you see is what you get. However, when light passes through a scattering medium, the obtained image information degrades into a speckle image similar to noise. Therefore, scattering media are often regarded as an obstacle to traditional imaging.

[0003] However, speckle images that appear noisy on the surface often contain information about hidden targets. Therefore, researchers have made considerable efforts based on the principles of scattering imaging. Currently, scattering imaging techniques involve wavefront shaping, transfer matrix measurement, optical phase conjugation, and speckle correlation imaging based on the memory effect. Among these, speckle correlation imaging based on the memory effect has attracted widespread attention because it does not require invasive imaging systems.

[0004] However, speckle correlation imaging based on the memory effect is limited by the memory effect of the scattering medium in terms of field of view, and can usually only reconstruct single targets smaller than the memory effect range. Once the target size exceeds the memory effect range, this method cannot recover the image information of the hidden target, resulting in reconstruction failure. Therefore, for speckle correlation imaging based on the memory effect to achieve broader development and wider application, it is urgent to develop a technical solution that overcomes the limitations of the memory effect and has a larger imaging field of view. Summary of the Invention:

[0005] To address the shortcomings and problems of existing technologies, this invention proposes a method and apparatus for non-invasive large field-of-view imaging through scattering media based on PSF prediction and correlation calculations. This invention achieves non-invasive large field-of-view imaging through scattering media using point spread function (PSF) prediction and correlation calculations. It solves the problem of limited field of view in non-invasive speckle correlation imaging and achieves super-resolution imaging on this basis. This invention does not require invasive "guide stars" or prior information, and the imaging field of view is not limited by memory effects. It has the advantages of a large imaging field of view and high imaging resolution, simultaneously achieving both. It has wide applications in biomedical imaging and optical microscopy.

[0006] To achieve the above objectives, the present invention is implemented through the following solution:

[0007] A method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations is proposed, which utilizes a phase retrieval algorithm to reconstruct sub-targets O hidden in the scattering medium. iBased on the image, the orientation of all sub-targets is obtained through point spread function prediction and autocorrelation and cross-correlation operations. The sub-targets are then image-stitched according to their orientations to achieve non-invasive large field-of-view imaging of target O. The specific steps are as follows:

[0008] Step 1: Scan the target O hidden in the scattering medium using active illumination. The target O is illuminated by n scans, then:

[0009]

[0010] Among them, O i It is the sub-target of the target O to be imaged that is illuminated during the i-th scan, and each sub-target O i All targets are within their respective memory effect ranges. After n scans of illumination, target O exceeds the memory effect range of the scattering imaging system; the area array photodetector records all sub-targets O. i The speckle image of reflected light passing through the scattering medium. i ;

[0011] Step 2: Extract the speckle image I corresponding to each sub-target using the phase retrieval algorithm. i Reconstructing the sub-target image from autocorrelation O i ';

[0012] Step 3: Determine the two adjacent sub-targets O i With O i+1 Is the distance between them smaller than the diameter of the memory effect range? If the two sub-targets O i With O i+1 If the memory effect ranges of the two sub-targets overlap, then the speckle image I of the two sub-targets i with I i+1 There is an information correlation between them. The reconstructed images O' of the two sub-targets are determined by estimating the point spread function (PSF) of their respective speckle images and calculating the autocorrelation and cross-correlation of the PSF of the two sub-targets. i With O' i+1 The relative position and direction information between them;

[0013] Step 4: The reconstructed image with correct relative position and orientation information obtained in Step 3 is shifted and superimposed to obtain the correctly stitched target O, avoiding the limitation of the memory effect of the scattering medium and realizing non-invasive large field-of-view imaging through the scattering medium.

[0014] Furthermore, in step 3, the specific steps for determining the relative position and orientation information of the two reconstructed sub-target images are as follows:

[0015] Step 3.1: Predict the point spread function: from the speckle image I i Compared with the reconstructed sub-target image O' obtained in step 2 iUsing Wiener deconvolution algorithm to predict speckle image I i Point spread function PSF' i :

[0016] PSF i =Deconv(I i ,O' i )

[0017] Where Deconv represents the deconvolution operation;

[0018] Step 3.2: Determine the reconstructed sub-target image O' i With O' i+1 The relative positions between them: calculate PSF' respectively i Autocorrelation and PSF' i and PSF' i+1 The cross-correlation of the two-dimensional coordinates of the respective correlation peaks is determined, and the difference is used to obtain the image O' representing the sub-target. i With O' i+1 Shift vector of relative positions

[0019]

[0020] In the above formula, ★ represents the correlation operation, and position{} represents the two-dimensional coordinates of the correlation peak position;

[0021] Step 3.3: Determine the reconstructed sub-target image O' i Orientation information: will reconstruct the sub-target image O' i The direction can be flipped left and right, up and down, and simultaneously flipped left, right, up and down, as well as the sub-target image O' i This yields the sub-target image O'. i The four different directional states; then from the speckle image I i and sub-target image O' i The PSF' corresponding to the four directions is estimated from the four different directional states. i Then, by four estimated PSF' i Compared with adjacent speckle images I i+1 Deconvolution reconstruction yields four deconvolution images. These four deconvolution images are then combined with the sub-target image O' obtained from phase retrieval. i+1 By comparing the shapes, O' can be determined. i The correct direction;

[0022] Step 3.4: Reconstruct the two sub-target images O' with the correct orientation i With O' i+1 According to the shift vector in step 3.2 Perform shifting and superposition;

[0023] Step 3.5: Repeat steps 3.1 to 3.4 for all sub-targets to determine the relative positional relationship and orientation information of all sub-targets.

[0024] A method for improving imaging resolution through non-invasive large field-of-view imaging based on PSF prediction and correlation calculations is disclosed. Based on the aforementioned method for non-invasive large field-of-view imaging through scattering media using PSF prediction and correlation calculations, a Gaussian fitting relocation method is employed to obtain a high-resolution image that breaks the diffraction limit of the imaging system. The relative position and orientation information of the high-resolution images of sub-targets are determined and stitched together, enabling both large field-of-view imaging and super-resolution imaging of the target. The specific steps are as follows:

[0025] Step 1: Repeat steps 1 to 2 in the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations to obtain the i-th sub-target O under the j-th sparse illumination. i_j low-resolution image O' i_j Then, the Gaussian fitting relocation method is used to process the low-resolution image to obtain a high-resolution image that breaks through the diffraction limit of the imaging system.

[0026] Step 2: Calculate the point spread function (PSF) of the imaging system, and compare PSF' with the m acquired speckle images I. i_j Each image undergoes deconvolution, followed by Gaussian fitting and relocalization to obtain m high-resolution images.

[0027] Step 3: Extract m high-resolution images By superimposing corresponding positions, a high-resolution image of the sub-target that breaks through the diffraction limit is obtained.

[0028] Step 4: Convert the high-resolution images of each sub-target obtained in Step 3 into... Following steps 3 and 4 of the method for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations, non-invasive large field-of-view super-resolution imaging through a scattering medium is achieved.

[0029] Furthermore, in step 2, the method for calculating the point spread function PSF' is compared with the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations. i The calculation method is the same.

[0030] A device for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations is used to implement the above-mentioned method. The device includes a target, an aperture stop, a scattering medium, and an area array photodetector. The aperture stop is placed between the target and the area array photodetector. The scattering medium is in close contact with one side of the aperture stop. The planes containing the area array photodetector and the scattering medium are parallel to the plane containing the target, and their centers coincide with the normal of the target. The distance between the scattering medium and the target is greater than the distance between the scattering medium and the area array photodetector.

[0031] Furthermore, the scattering medium can be frosted glass, diffuse reflection walls, biological tissue, clouds, turbid water, or artificial scattering media.

[0032] This invention utilizes PSF prediction and correlation calculations to achieve non-invasive, large field-of-view super-resolution imaging on a suitable device. First, it reconstructs images of sub-targets hidden in scattering media using a phase retrieval algorithm. Then, it predicts the corresponding PSF and performs autocorrelation and cross-correlation calculations to obtain the orientation of each sub-target image. Based on the obtained orientation, the sub-target images are stitched together, thereby achieving non-invasive, large field-of-view imaging of the entire target hidden in the scattering medium and exceeding the memory effect range, with significantly improved imaging quality. This invention is suitable for imaging applications requiring scattering through biological tissues, clouds, smoke, turbid liquids, and other turbid scattering media. Furthermore, it does not require invasive "guides" or prior information, and the imaging field of view is not limited by the memory effect, offering advantages such as a large imaging field of view and high imaging resolution, simultaneously achieving both. It has wide applications in biomedical imaging and optical microscopy. Attached image description:

[0033] Figure 1 This is a schematic diagram of the device of the present invention;

[0034] Figure 2 These are the speckle image, sub-target image, and point spread function image of the present invention;

[0035] Figure 3 Reconstruct the relative positions of the nine sub-target images O'1-O'9 from (a1) to (a8);

[0036] Figure 4 The result of overlaying and stitching nine sub-target images O'1-O'9 reconstructed from (a1) to (a8);

[0037] Figure 5 The results are for a method to achieve non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations;

[0038] Figure 6(a)-(i) represent non-invasive super-resolution imaging of sub-targets, and (g) represents a super-resolution imaging method based on PSF prediction and correlation calculation to achieve non-invasive large field-of-view imaging and improve imaging resolution.

[0039] Figure 7 (a) shows non-invasive large field-of-view imaging; (b) shows non-invasive large field-of-view super-resolution imaging with improved resolution.

[0040] In the diagram: 1. Target; 2. Aperture stop; 3. Scattering medium; 4. Area array photodetector.

[0041] Figure 1 In the diagram, the white solid circle in target 1 represents the memory effect range, the numbers at the corresponding positions indicate the order of the sub-targets O1-O9 that were scanned and illuminated in sequence, and the arrows indicate the order of correlation calculations between the estimated point spread function PSF1-PSF9.

[0042] Figure 2 Nine speckle images I1-I9 were acquired in (a1)-(a9); nine sub-target images O'1-O'9 were reconstructed in (b1)-(b9); and the corresponding point spread functions PSF'1-PSF'9 were estimated in (c1)-(c9). Detailed implementation method:

[0043] The present invention will now be described in further detail with reference to the accompanying drawings.

[0044] It should be noted that, for ease of description, the descriptions of direction in the following text are consistent with the directions in the accompanying drawings, but they do not limit the structure of the present invention.

[0045] like Figures 1-6 As shown, this invention discloses a method for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and correlation calculations, which utilizes a phase retrieval algorithm to reconstruct sub-targets O hidden in the scattering medium. i Based on the image, the orientation of all sub-targets is obtained through point spread function prediction and autocorrelation and cross-correlation operations. The sub-targets are then image-stitched according to their orientations to achieve non-invasive large field-of-view imaging of target O. The specific steps are as follows:

[0046] Step 1: Scan the target O hidden in the scattering medium using active illumination. The target O is illuminated by n scans, then:

[0047]

[0048] Among them, O i It is the sub-target of the target O to be imaged that is illuminated during the i-th scan, and each sub-target O iAll targets are within their respective memory effect ranges. After n scans of illumination, target O exceeds the memory effect range of the scattering imaging system; the area array photodetector records all sub-targets O. i The speckle image of reflected light passing through the scattering medium. i .

[0049] Step 2: Extract the speckle image I corresponding to each sub-target using the phase retrieval algorithm. i Reconstructing the sub-target image O' from autocorrelation i .

[0050] Step 3: Determine the two adjacent sub-targets O i With O i+1 Is the distance between them smaller than the diameter of the memory effect range? If the two sub-targets O i With O i+1 If the memory effect ranges of the two sub-targets overlap, then the speckle image I of the two sub-targets i with I i+1 There is an information correlation between them. The reconstructed images O' of the two sub-targets are determined by estimating the point spread function (PSF) of their respective speckle images and calculating the autocorrelation and cross-correlation of the PSF of the two sub-targets. i With O' i+1 The relative position and orientation information between the two sub-targets O; i With O i+1 The distance between the two sub-targets is greater than the memory effect range, indicating that the memory effect ranges of the two sub-targets do not overlap, and there is no information correlation between the speckle images of the two sub-targets.

[0051] The specific steps for determining the relative position and orientation information of the two reconstructed sub-target images are as follows:

[0052] Step 3.1: Predict the point spread function: from the speckle image I i Compared with the reconstructed sub-target image O' obtained in step 2 i Using Wiener deconvolution algorithm to predict speckle image I i Point spread function PSF' i :

[0053] PSF i =Deconv(I i ,O' i (2)

[0054] Here, Deconv represents the deconvolution operation.

[0055] Step 3.2: Determine the reconstructed sub-target image O' i With O' i+1 The relative positions between them: calculate PSF' respectively iand PSF' i Autocorrelation and PSF' i and PSF' i+1 The cross-correlation of the two-dimensional coordinates of the respective correlation peaks is determined, and the difference is used to obtain the image O' representing the sub-target. i With O' i+1 Shift vector of relative positions

[0056]

[0057] In the above formula, ★ represents the correlation operation, and position{} represents the two-dimensional coordinates of the correlation peak position.

[0058] Step 3.3: Determine the reconstructed sub-target image O' i Directional information: It should be noted that the phase-reconstructed sub-target image O' i The orientation is uncertain; it may involve incorrect orientations such as left-right flipping, up-down flipping, or simultaneous flipping in all directions. However, the reconstructed sub-target image O' i speckle image I i With sub-target image O' i+1 speckle image I i+1 There is a correlation in location information between them. In this case, the sub-target image O' will be reconstructed. i The direction can be flipped left and right, up and down, and simultaneously flipped left, right, up and down, as well as the sub-target image O' i This yields the sub-target image O'. i The four different directional states; then from the speckle image I i and sub-target image O' i The PSF' corresponding to the four directions is estimated from the four different directional states. i Then, by four estimated PSF' i Compared with adjacent speckle images I i+1 Deconvolution reconstruction yields four deconvolution images. These four deconvolution images are then combined with the sub-target image O' obtained from phase retrieval. i+1 By comparing the shapes, O' can be determined. i The correct direction.

[0059] Step 3.4: Reconstruct the two sub-target images O' with the correct orientation i With O' i+1 According to the shift vector in step 3.2 Perform shifting and superposition.

[0060] Step 3.5: Repeat steps 3.1 to 3.4 for all sub-targets to determine the relative positional relationship and orientation information of all sub-targets.

[0061] Step 4: The reconstructed image with correct relative position and orientation information obtained in Step 3 is shifted and superimposed to obtain the correctly stitched target O, avoiding the limitation of the memory effect of the scattering medium and realizing non-invasive large field-of-view imaging through the scattering medium.

[0062] This invention also discloses a method for improving imaging resolution through non-invasive large field-of-view imaging based on PSF prediction and correlation calculations. Based on the aforementioned method for non-invasive large field-of-view imaging through scattering media using PSF prediction and correlation calculations, a high-resolution image exceeding the diffraction limit of the imaging system is obtained using a Gaussian fitting relocation method. The relative position and orientation information of the high-resolution images of the sub-targets are determined and stitched together, enabling both large field-of-view imaging and super-resolution imaging of the target. The specific steps are as follows:

[0063] Step 1: Repeat steps 1 to 2 in the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations to obtain the i-th sub-target O under the j-th sparse illumination. i_j low-resolution image O' i_j Then, the Gaussian fitting relocation method is used to process the low-resolution image to obtain a high-resolution image that breaks through the diffraction limit of the imaging system.

[0064] If the desired sparse illumination can be further controlled, then the i-th sub-target O in step 1 of the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations can be achieved. i If the light is applied m times by a random sparse lattice, then we have

[0065]

[0066] Among them, O i_j Sub-target O i The result of illumination by the j-th sparse lattice (i.e., the j-th sparse illumination of the i-th sub-target O) i_j Repeat step 2 of the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations to obtain the i-th sub-target O under the j-th sparse illumination. i_j low-resolution image O' i_j Then, the low-resolution image O' is processed using a Gaussian fitting relocation method. i_j To obtain high-resolution images that break the diffraction limit of the imaging system.

[0067] Step 2: Calculate the point spread function (PSF) of the imaging system, and compare PSF' with the m acquired speckle images I. i_j Each image undergoes deconvolution, followed by Gaussian fitting and relocalization to obtain m high-resolution images. Methods for calculating the point spread function (PSF') and methods for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations. i The calculation method is the same.

[0068] Step 3: Extract m high-resolution images By superimposing corresponding positions, a high-resolution image of the sub-target that breaks through the diffraction limit is obtained.

[0069] Step 4: Convert the high-resolution images of each sub-target obtained in Step 3 into... Following steps 3 and 4 of the method for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations, non-invasive large field-of-view super-resolution imaging through a scattering medium is achieved.

[0070] This invention also discloses a device for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations, used to implement the two methods mentioned above. The device includes a target 1, an aperture stop 2, a scattering medium 3, and a surface array photodetector 4. The aperture stop 2 is placed between the target 1 and the surface array photodetector 4, and the scattering medium 3 is in close contact with one side of the aperture stop 2. Figure 1 In the embodiment shown, the scattering medium 3 is located to the left of the aperture stop 2. Placing the scattering medium 3 on the other side of the aperture stop 2 can also achieve the same technical effect as this device. The planes where the area array photodetector 4 and the scattering medium 3 are located are parallel to the plane where the target 1 is located, and the center position coincides with the normal of the target 1. The distance between the scattering medium 3 and the target 1 is greater than the distance between the scattering medium 3 and the area array photodetector 4.

[0071] The scattering medium 3 is frosted glass, diffuse reflection wall, biological tissue, clouds, turbid water, or artificial scattering medium.

[0072] Achieving non-invasive large field-of-view imaging:

[0073] in accordance with Figure 1 Construct the device of the present invention, in Figure 1 In this embodiment, the object distance u from the plane of target 1 to the plane of scattering medium 3 is 94 mm, and the image distance v from the plane of scattering medium 3 to the plane of the area array photodetector 4 is 80 mm. Sub-targets O1-O9 are scanned sequentially, with the scanning illumination order as follows: Figure 1 As shown in the figures. The reflected light from sub-targets O1-O9 passes through a 4.5mm aperture stop and through the scattering medium, and then nine speckle images I1-I9 are acquired by the area array photodetector 4. In this embodiment, an sCMOS camera is used as the area array photodetector. The results are as follows. Figure 2As shown in (a1)-(a9). Then, the speckle images I1-I9 are reconstructed using the phase retrieval algorithm to obtain nine sub-target images O'1-O'9, as follows. Figure 2 As shown in (b1)-(b9), the corresponding speckle image and sub-target image are deconvolved to predict PSF'1-PSF'9, as follows. Figure 2 As shown in (c1)-(c9). According to... Figure 1 The arrows shown in the diagram calculate PSF' respectively. i Autocorrelation and PSF' i and PSF' i+1 The cross-correlation of the two-dimensional coordinates of the respective correlation peaks is determined, and the difference is used to obtain the image O' representing the sub-target. i With O' i+1 The eight shift vectors of relative positions between them like Figure 3 As shown in (a1)-(a8). Then, the orientation information of the reconstructed sub-target images O'1-O'9 is determined. Finally, the sub-target images O'1-O'9 with the correct orientation are respectively... Figure 3 The eight shift vectors (a1)-(a8) shown are superimposed and concatenated, as follows: Figure 4 As shown in (a1)-(a8), the correct overall objective O was obtained, as follows. Figure 5 As shown, non-invasive large field-of-view imaging can be achieved.

[0074] Achieving non-invasive, large field-of-view super-resolution imaging:

[0075] First, the first sub-target O1 is randomly illuminated by sparse dot matrix j times, resulting in j speckle images I. 1_j In this embodiment, j is set to 500, and any speckle image I is selected. 1_56 Phase retrieval was then performed to obtain a low-resolution sub-image O' 1_56 By combining Gaussian fitting and relocalization processing, high-resolution sub-images that break through the diffraction limit of the imaging system are obtained. Then the acquired speckle image I 1_56 With high-resolution sub-images Perform Wiener deconvolution to calculate the point spread function PSF' of the speckle image. i The point spread function PSF' calculated above i With 500 speckle images I 1_j Each image undergoes deconvolution operation, followed by Gaussian fitting and relocalization processing to obtain 500 high-resolution sub-images. 500 high-resolution sub-images By superimposing the corresponding positions, a sub-target image O1 that breaks the diffraction limit can be obtained, such as... Figure 6 As shown in (a). Repeat the above steps for the remaining 8 sub-targets one by one, performing super-resolution imaging as follows. Figure 6As shown in (b)-(i). Finally, the obtained super-resolution images (6(a)-(i)) are combined with non-invasive large field-of-view imaging to achieve non-invasive large field-of-view super-resolution imaging, as shown in... Figure 6 As shown in (g).

[0076] Taking common scattering media such as frosted glass, diffuse reflection walls, biological tissues, clouds, turbid water, and artificial scattering media as examples, non-invasive large field-of-view imaging can be achieved based on the above scattering media.

[0077] The area array photodetector 4 can also be a CCD, and non-invasive large field-of-view imaging can also be achieved according to the method of the present invention.

[0078] When the target is too large and exceeds the range of the memory effect, the phase retrieval algorithm is used to reconstruct the image of the sub-targets hidden in the scattering medium. Based on the point spread function prediction and autocorrelation and cross-correlation calculations, the orientation of all sub-targets is obtained. The images of the sub-targets are then stitched together according to their orientations to achieve non-invasive large field-of-view imaging. Figure 7 As shown in (a); the method for non-invasive large field-of-view imaging based on PSF prediction and correlation calculation proposed in this invention to improve imaging resolution, uses Gaussian fitting relocation method to obtain high-resolution images that break through the diffraction limit of the imaging system, determines the relative position and orientation information of the high-resolution images of sub-targets and stitches them together, so that the target achieves large field-of-view imaging and super-resolution imaging at the same time, such as... Figure 7 As shown in (b), it can simultaneously take into account both the imaging field of view and the imaging resolution.

[0079] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations, characterized in that, Reconstructing a target hidden in a scattering medium using a phase retrieval algorithm O sub-targets O i Based on the image, the orientation of all sub-targets is obtained using point spread function prediction and autocorrelation and cross-correlation operations. The sub-targets are then image-stitched according to their orientations to make the target... O The specific steps to achieve non-invasive large field-of-view imaging are as follows: Step 1: Use active illumination to scan targets hidden in scattering media. ,Target O quilt The illumination during the second scan is then: in, The target to be imaged No. The sub-targets that are scanned and illuminated, each sub-target Within their respective memory effect ranges, the target is illuminated after n scans. Exceeding the memory effect range of the scattering imaging system; the area array photodetector records all sub-targets. Speckle image of reflected light passing through a scattering medium ; Step 2: Extract the speckle image corresponding to each sub-target using a phase retrieval algorithm. Reconstructing sub-target images from autocorrelation ; Step 3: Determine the two adjacent sub-targets and Is the distance between the two sub-targets smaller than the diameter of the memory effect range? and If the memory effect ranges of the two sub-targets overlap, then the speckle images of the two sub-targets... and There is an information correlation between them, which is achieved by predicting the point spread function of the speckle images of the two sub-targets. The reconstructed images of the two sub-targets are determined by calculating the autocorrelation and cross-correlation of their point spread functions. and The relative position and direction information between them; Step 3.1: Predict the point spread function: from the speckle image The reconstructed sub-target image obtained in step 2 Predicting speckle images using Wiener deconvolution algorithm Point spread function : in, This represents the deconvolution operation; Step 3.2: Determine the reconstructed sub-target image and The relative positions between them: calculate separately autocorrelation and and The cross-correlation of the data was used to determine the two-dimensional coordinates of the respective correlation peaks, and the difference was calculated to obtain the image representing the sub-target. and Shift vector of relative positions : In the above formula, ★ represents the relevant operation. Two-dimensional coordinates representing the location of the relevant peak; Step 3.3: Determine the reconstructed sub-target image Orientation information: will reconstruct sub-target images The direction can be flipped left and right, up and down, and simultaneously flipped left, right, up and down, as well as the sub-target image. It itself yields the sub-target image. Four different directional states; then from the speckle image and sub-target images The four directions corresponding to the predicted states are obtained from the four different directional states. Then, from four estimates Compared with adjacent speckle images Deconvolution reconstruction yields four deconvolution images. These four deconvolution images are then combined with the sub-target image obtained from phase retrieval. By comparing their shapes, we can determine The correct direction; Step 3.4: Reconstruct the two sub-target images with the correct orientation. and According to the shift vector in step 3.2 Perform shifting and superposition; Step 3.5: Repeat steps 3.1 to 3.4 for all sub-targets to determine the relative positional relationships and orientation information of all sub-targets; Step 4: Shift and overlay the reconstructed images with correct relative position and orientation information obtained in Step 3 to obtain the correctly stitched target image. O This avoids the limitations of the memory effect of the scattering medium and enables non-invasive large field-of-view imaging through the scattering medium.

2. A method for improving imaging resolution by achieving non-invasive large field-of-view imaging based on PSF prediction and correlation calculation, based on the method for achieving non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and correlation calculation as described in claim 1, characterized in that... A high-resolution image that breaks the diffraction limit of the imaging system is obtained by using the Gaussian fitting relocation method. The relative position and orientation information of the high-resolution images of the sub-targets are determined and stitched together, enabling super-resolution imaging of the target while achieving a large field of view. The specific steps are as follows: Step 1: Repeat steps 1 to 2 of the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations to obtain the result. Secondary sparse illumination Sub-goals low-resolution images Then, the Gaussian fitting relocation method is used to process the low-resolution image to obtain a high-resolution image that breaks through the diffraction limit of the imaging system. ; Step 2: Calculate the point spread function of the imaging system ,Will and A collection of speckle images Each element undergoes deconvolution, followed by Gaussian fitting and relocalization to obtain... A high-resolution image ; Step 3: Put m A high-resolution image By superimposing corresponding positions, a high-resolution image of the sub-target that breaks through the diffraction limit is obtained. ; Step 4: Convert the high-resolution images of each sub-target obtained in Step 3 into... Following steps 3 and 4 of the method for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations, non-invasive large field-of-view super-resolution imaging through a scattering medium is achieved.

3. The method for improving imaging resolution based on PSF prediction and correlation calculation to achieve non-invasive large field-of-view imaging according to claim 2, characterized in that, In step 2, the point spread function The calculation method and the point spread function of the method for non-invasive large field-of-view imaging through scattering media based on PSF prediction and related calculations. The calculation method is the same.

4. A device for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations, used to implement the method described in any one of claims 1 to 3, characterized in that, The device includes a target (1), an aperture stop (2), a scattering medium (3), and an array photodetector (4). The aperture stop (2) is placed between the target (1) and the array photodetector (4). The scattering medium (3) is attached to one side of the aperture stop (2). The planes of the array photodetector (4) and the scattering medium (3) are parallel to the plane of the target (1), and their centers coincide with the normal of the target (1). The distance between the scattering medium (3) and the target (1) is greater than the distance between the scattering medium (3) and the array photodetector (4).

5. The device for non-invasive large field-of-view imaging through a scattering medium based on PSF prediction and related calculations according to claim 4, characterized in that, The scattering medium (3) is frosted glass, diffuse reflection wall, biological tissue, cloud, turbid water or artificial scattering medium.

Images