A speckle image reconstruction method and device based on two-way feedback fusion

Abstract

This application provides a speckle image reconstruction method and apparatus based on dual-path feedback fusion. The method includes: preprocessing the original speckle image; directly performing phase retrieval processing on the preprocessed image to obtain a first image; performing adaptive image enhancement processing on the preprocessed image to obtain an enhanced image; performing phase retrieval processing on the enhanced image to obtain a second image; and fusing the first image and the second image to output a reconstructed target image. This application innovatively proposes a dual-input parallel synchronous transmission architecture, breaking the traditional single-path serial limitation and achieving a synergistic improvement in imaging efficiency and reconstruction accuracy. Adaptive front-end enhancement of the speckle image dynamically optimizes processing parameters based on real-time image features, improving the robustness of this application's speckle image reconstruction technology to dynamically changing scattering environments.

Description

Technical Field

[0001] This application relates to the field of computational optical imaging and image processing technology, and in particular to a method and apparatus for speckle image reconstruction based on dual-path feedback fusion. Background Technology

[0002] Speckle correlation imaging, based on the optical memory effect, provides a powerful tool for non-invasive imaging through strongly scattering media. However, its application faces three major challenges when the scattering medium (such as smog, turbulence, and biological tissue with varying concentrations and distributions over time) is in a dynamic state: First, image degradation caused by medium dynamics: dynamic scattering causes the diffusion function of the scattering point of the medium to be time-varying, which means that the acquired speckle image not only has inherent speckle noise, but also has severe non-uniform background illumination and random noise superimposed, significantly reducing the signal-to-noise ratio and directly destroying the input data quality required by the phase retrieval algorithm.

[0003] Secondly, there is the adaptive bottleneck in preprocessing techniques: image enhancement preprocessing is indispensable for improving input quality. However, traditional methods such as histogram equalization, median filtering, and desharpening masks all use fixed parameters. Under dynamic scattering conditions, image features change rapidly, and fixed parameters cannot achieve the best balance between noise suppression and detail enhancement. This can easily lead to insufficient enhancement resulting in residual noise or over-enhancement introducing artifacts, which are then amplified and carried over to the subsequent phase retrieval stage, causing reconstruction failure.

[0004] Finally, there is an inherent contradiction in the system architecture: most existing speckle image processing solutions adopt a single-path serial architecture. This architecture has inherent limitations: if the process is simplified in pursuit of real-time performance, reconstruction accuracy will inevitably be sacrificed; if complex iterations are used in pursuit of high fidelity, the processing speed will be significantly slowed down. In dynamic imaging scenarios, both rapid response to changes and clear restoration of details are required, and this contradiction between "efficiency and quality" cannot be reconciled under a single-path framework. Summary of the Invention

[0005] In view of this, this application provides a speckle image reconstruction method and apparatus based on dual-path feedback fusion, which innovates at the system architecture level, has front-end adaptive processing capabilities, and can intelligently coordinate speed and accuracy to provide a novel speckle image processing solution to meet the demanding imaging requirements of dynamic scattering media.

[0006] Specifically, this application is implemented through the following technical solution: According to the embodiments of this specification, a speckle image reconstruction method based on dual-path feedback fusion is provided, comprising the following steps: Step S1: Preprocess the original speckle image; Step S2: Perform phase recovery processing directly on the preprocessed image to obtain the first image; Step S3: Perform adaptive image enhancement processing on the preprocessed image to obtain an enhanced image, and perform phase recovery processing on the enhanced image to obtain a second image; Step S4: Fuse the first image and the second image to output the target reconstructed image.

[0007] According to embodiments of this specification, a speckle image reconstruction device based on dual-path feedback fusion is also provided, comprising: The preprocessing unit preprocesses the original speckle image; The adaptive enhancement unit performs adaptive image enhancement processing on the preprocessed image to obtain the enhanced image; The phase retrieval unit performs phase retrieval processing directly on the preprocessed image to obtain the first image, and performs phase retrieval processing on the enhanced image to obtain the second image; The fusion unit fuses the first image and the second image to output the target reconstructed image.

[0008] According to embodiments of this specification, an electronic device is also provided, comprising: a processor; and a computer-readable storage medium storing computer program instructions, which, when executed by the processor, cause the processor to perform a speckle image processing method as described above.

[0009] According to embodiments of this specification, a computer-readable storage medium is also provided, on which a computer program is stored, the computer program being executed by a processor as described above in a speckle image processing method.

[0010] The embodiments of this application have at least the following beneficial effects: First, it achieves a synergistic improvement in imaging efficiency and reconstruction accuracy; This application employs a dual-path parallel processing architecture: the first path directly performs phase retrieval on the preprocessed image, ensuring extremely low processing latency and meeting real-time requirements; the second path first performs adaptive image enhancement and then phase retrieval, significantly improving the quality of the input image and thus obtaining high-precision reconstruction results. By fusing the results from the two paths, the final output target reconstructed image retains both the high efficiency and contour fidelity of the first path and incorporates the rich detail information of the second path, thereby simultaneously optimizing processing speed and imaging quality within the same methodological framework. This resolves the core contradiction in traditional serial architectures where speed results in unclear images, or clear images result in slow speed.

[0011] Second, it enhances adaptability and robustness to dynamic scattering environments; This application introduces an adaptive image enhancement module that dynamically adjusts enhancement parameters based on real-time characteristics of the preprocessed image, such as signal-to-noise ratio, edge gradient, and noise ratio, achieving an adaptive balance between noise suppression and detail enhancement. Compared to conventional enhancement methods using fixed parameters, this application maintains stable and excellent preprocessing results under different scattering conditions or image degradation levels. It effectively avoids phase recovery failures caused by residual noise due to insufficient enhancement or artifacts introduced by over-enhancement, thereby greatly improving the robustness of the method in practical applications involving dynamic scattering media such as turbulence, smoke, and biological tissue.

[0012] Third, by integrating the advantages of both paths through a fusion mechanism, high-quality output can be obtained; This application embodiment fuses different reconstruction results from two paths, assigning fusion weights based on the local image quality of each pixel location. This preserves the overall structural clarity of the first reconstruction result while absorbing the subtle features recovered after enhancement from the second reconstruction result. Compared to the independent output of either single path, the fused output has a higher peak signal-to-noise ratio and structural similarity, resulting in superior visual effects, richer details, and a more uniform background.

[0013] Fourth, modular design facilitates expansion and integration; Each step in the embodiments of this application is functionally independent and has a clearly defined interface: preprocessing, adaptive enhancement, phase retrieval, and fusion can all be optimized or replaced with other known algorithms without affecting the overall architecture. Therefore, the embodiments of this application can not only serve as a complete speckle imaging solution, but also be embedded into existing imaging systems as enhancement or preprocessing sub-modules, possessing broad engineering applicability and industrial application prospects. Attached Figure Description

[0014] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Some specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings in an exemplary and non-limiting manner. The same reference numerals in the drawings indicate the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of a speckle image reconstruction method based on dual-path feedback fusion. Figure 2 This is an overall flowchart illustrating a speckle image reconstruction method based on dual-path feedback fusion, as shown in an exemplary embodiment of this application. Figure 3 This is a schematic diagram of the phase recovery process in a speckle image reconstruction method based on dual-path feedback fusion, as illustrated in an exemplary embodiment of this application. Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment of this application; Figure 5 This is a block diagram of a speckle image reconstruction apparatus based on dual-path feedback fusion, as illustrated in an exemplary embodiment of this application. Detailed Implementation

[0015] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of one or more embodiments of this application as detailed in the appended claims.

[0016] This application proposes a globally optimized dual-input feedback speckle image processing method. The core of its technical solution lies in constructing a hierarchical and collaborative processing framework, which organically comprises a unified preprocessing stage, a dual-input parallel processing stage, and an intelligent fusion output stage. In the unified preprocessing stage, frequency domain smoothing and phase optimization methods are used to initially normalize the original speckle image. This method first performs flat-field correction in the frequency domain to accurately remove the slowly varying background components caused by non-uniform illumination, while retaining high-frequency speckle signals; then, a Gaussian low-pass filter is applied to gently suppress high-frequency electronic noise from the imaging system, thereby outputting a preprocessed image with uniform background and high phase consistency, providing a high-quality reference data stream for subsequent parallel processing.

[0017] In the dual-input parallel processing stage, the preprocessed image is simultaneously fed into two complementary processing channels, which is the innovation of this application's architecture. The first high-efficiency channel pursues extreme processing speed, directly feeding the preprocessed image into the subsequent phase retrieval module to ensure the system has the lowest initial response latency. The second high-quality channel pursues optimal input quality, first sending the preprocessed image to the median-desharpening mask adaptive enhancement module. This module is the core of this application's adaptive approach, integrating median filtering and desharpening mask operators, and innovatively introducing an adaptive enhancement factor feedback unit. This unit analyzes the input image's signal-to-noise ratio, average edge gradient, and noise ratio in real time, dynamically calculating and adjusting the median filter window size and the desharpening mask's sharpening intensity accordingly, achieving a precise and adaptive balance between noise suppression and detail enhancement. The enhanced image optimized by this module is then sent to the phase retrieval module.

[0018] In the intelligent fusion output stage, the core is the phase alignment and coherent summation module. This module receives image data from the two channels mentioned above and initiates an independent iterative reconstruction thread for them. Internally, it employs an innovative dynamic reference update mechanism: in each iteration, it estimates and compensates for random shifts by calculating the cross-correlation function between the sub-image block and the current reference block, achieving precise alignment of the sub-blocks; it then coherently sums the spectral phases of all aligned sub-blocks to enhance the signal and suppress noise, and synthesizes this sum with a fixed amplitude obtained from the average power spectrum of all sub-blocks. After inverse transformation, the reconstruction result is updated; this result then serves as the new reference block for the next iteration, and this process continues until convergence. This mechanism ensures high accuracy and stability in phase recovery. Finally, the module adaptively weights and fuses the two preliminary reconstruction results from the high-efficiency channel and the high-quality channel based on their local quality features, such as sharpness, integrating the contour fidelity of the high-efficiency channel with the detail richness of the high-quality channel, outputting a single optimal reconstructed image.

[0019] Optionally, this application also provides an apparatus for implementing the above method, the apparatus comprising an image acquisition unit, and a preprocessing unit, an adaptive enhancement unit, a phase recovery unit, and a fusion output unit connected in sequence and electrically, wherein the output of the preprocessing unit is connected in a dual-path configuration to the input of the adaptive enhancement unit and the first input of the phase recovery unit, and the output of the adaptive enhancement unit is connected to the second input of the phase recovery unit, thereby implementing the dual-input parallel architecture at the hardware level.

[0020] Compared with the prior art, this application can produce the following beneficial effects: First, it achieves a breakthrough in system-level performance, perfectly balancing efficiency and accuracy. The innovative dual-input synchronous transmission architecture of this application fundamentally overturns the traditional serial processing mode. By decoupling rapid response and fine-grained processing tasks into parallel channels and intelligently fusing them at the back end, the system can simultaneously advance high-speed reconstruction and high-quality reconstruction within the same processing flow. This resolves the core contradiction in dynamic scattering imaging where fast results in unclear images and clear images in slow ones, making it possible to obtain high-precision reconstruction results in dynamic scenarios with high real-time requirements, such as monitoring and detection.

[0021] Second, it possesses feedforward feedback adaptive capability, greatly enhancing environmental robustness. Addressing the challenge of unstable image features caused by dynamic scattering, this application designs a closed-loop feedback unit in the adaptive enhancement module. This unit, based on a feature-aware parameter adaptive mechanism, overcomes the shortcomings of traditional fixed-parameter enhancement methods that suffer from drastic performance drops in dynamic environments, enabling the system to exhibit strong adaptability and stability to interference such as illumination changes and noise fluctuations.

[0022] Third, it improves the accuracy and convergence of phase retrieval. The phase retrieval module integrates three techniques: dynamic reference update, sub-block phase alignment, and fixed amplitude constraint. Dynamic reference update allows the iterative process to use the current best estimate as a benchmark, accelerating convergence and avoiding getting trapped in local optima; sub-block phase alignment effectively corrects random pixel shifts caused by dynamic scattering, ensuring the effectiveness of coherent summation; and fixed amplitude constraint provides robust amplitude priors for iteration and suppresses noise amplification. These three techniques work synergistically to significantly improve the signal-to-noise ratio and structural similarity of the final reconstructed image.

[0023] Fourth, it endows the system with modularity and high generalization engineering value: the core modules of this application have clear functional boundaries, well-defined interfaces, and adopt a loosely coupled design. This modularity makes the system highly flexible and scalable. For example, the adaptive enhancement module can be used independently to improve the front-end quality of other imaging systems; the entire dual-input frame can also be embedded as an enhancement unit into existing imaging platforms. Therefore, this application is not only a complete imaging solution, but also a high-performance processing framework that can be widely used in cutting-edge fields such as dynamic smoke imaging, optical monitoring of living tissue, and visual enhancement under adverse weather conditions, possessing enormous industrial application potential.

[0024] Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of a speckle image reconstruction method based on dual-path feedback fusion. Figure 1 As shown, a speckle image reconstruction method based on dual-path feedback fusion in one embodiment of this application includes the following steps: Step S1: Preprocess the original speckle image.

[0025] Step S2: Perform phase recovery processing directly on the preprocessed image to obtain the first image; Step S3: Perform adaptive image enhancement processing on the preprocessed image to obtain an enhanced image, and perform phase recovery processing on the enhanced image to obtain a second image.

[0026] Step S4: Fuse the first image and the second image to output the target reconstructed image.

[0027] This application's speckle image reconstruction method based on dual-path feedback fusion fundamentally overturns the traditional serial processing mode through a dual-input synchronous transmission architecture. By decoupling fast response and fine processing tasks into parallel channels and intelligently fusing them at the back end, high-speed reconstruction and high-quality reconstruction can be carried out in parallel within the same processing flow, achieving a balance between efficiency and quality.

[0028] The following describes the speckle image reconstruction method based on dual-path feedback fusion according to an embodiment of this application, using a specific application as an example.

[0029] In this embodiment, the adaptive image enhancement processing includes: calculating the feature parameters of the preprocessed image in real time and determining the adaptive enhancement parameters; performing filtering and desharpening mask processing on the image based on the adaptive enhancement parameters; and outputting an enhanced image. For example, the signal-to-noise ratio, average edge gradient, and noise pixel ratio of the image are calculated in real time, and the window size of the median filter and the sharpening intensity of the desharpening mask are dynamically adjusted based on the signal-to-noise ratio, average edge gradient, and noise pixel ratio.

[0030] Phase restoration processing includes performing the following steps on the preprocessed image and the enhanced image respectively: Step S21: Divide the input image into multiple sub-image blocks and select an initial reference block. Here, the input image can be a preprocessed image or an enhanced image.

[0031] Step S22: Calculate the cross-correlation function between each sub-image block and the current reference block, estimate the translation amount of each sub-image block relative to the current reference block by locating the cross-correlation peak, and align the sub-image blocks according to the translation amount.

[0032] Step S23: Coherently sum the spectral phases of all aligned sub-image blocks to obtain the average phase.

[0033] Step S24: Average the power spectra of all sub-image blocks to obtain a fixed amplitude.

[0034] Step S25: Combine the average phase and fixed amplitude to form a frequency domain signal, and then perform an inverse transformation to obtain the current iteration reconstruction result.

[0035] Step S26: Use the reconstruction result of the current iteration as the new reference block for the next iteration.

[0036] Step S27: Determine whether the difference between the reconstruction results of two adjacent iterations is less than a preset threshold; if yes, confirm convergence and output the first image or the second image. Otherwise, return to step S22 and continue iterating.

[0037] In one embodiment of this application, fusing a first image and a second image to output a target reconstructed image includes: calculating the local quality of the first image and the second image; and weightedly fusing the first image and the second image based on the local quality to obtain and output the target reconstructed image.

[0038] In one embodiment of this application, image preprocessing includes: performing frequency domain flat-field correction on the original speckle image to remove non-uniform illumination background; and performing Gaussian low-pass filtering on the corrected image to suppress high-frequency noise of the imaging system.

[0039] Figure 2This is an exemplary embodiment of the speckle image reconstruction method based on dual-path feedback fusion, illustrating the overall flowchart of the speckle image processing method. Figure 3 This is a schematic diagram of the phase recovery process in a speckle image processing method based on dual-path feedback fusion, as illustrated in an exemplary embodiment of this application; combined with Figure 2 and Figure 3 The present application describes the speckle image reconstruction method based on dual-path feedback fusion according to the embodiments of this application.

[0040] Step S100: Input the original speckle image. Specifically, data acquisition can be performed using... A complementary metal-oxide semiconductor (CMOS) camera captures the target light field through a dynamic scattering medium, such as a turbulent smoke box, to obtain a raw speckle image carrying target information but severely degraded. .

[0041] Step S200, Frequency Domain Smoothing and Phase Optimization Preprocessing. This step performs unified preprocessing, which can be achieved using existing technologies. The module is complete and aims to provide standardized data for subsequent processing. First, an image degradation model is established. ,in For ideal speckle patterns, For non-uniform backgrounds, Additive noise is used. Background noise is estimated by convolution with a large Gaussian kernel, such as 31×31. Perform frequency domain flat field correction: To eliminate Subsequently, a small Gaussian kernel, such as 5×5, was used, with a standard deviation of... right Perform low-pass filtering to suppress high-frequency noise. To obtain a preprocessed image This step significantly improves the uniformity and phase consistency of the image.

[0042] Step S300: Dual-input parallel branch processing.

[0043] Step S310, branch Efficient path: preprocessing images Directly used as input The image is then transmitted to the PACS (Picture Archiving and Communication System) module, which features phase restoration. This path has extremely low latency and is used to quickly generate a reconstructed baseline.

[0044] Step S311: Input the preprocessed image directly into the phase recovery module; Step S312: Start the fast reconstruction thread; Step S313: Obtain the fast reconstruction result.

[0045] Step S320, branch High-quality path: preprocessing the image Enter the MUSE (Muse Public ArtStudio) module for adaptive enhancement.

[0046] Specifically, in step S321, the adaptive enhancement module is entered. This module operates according to the following sub-steps: Step S322: Real-time calculation of signal-to-noise ratio, edge gradient, and noise ratio. Feature extraction, real-time calculation of image... of: Signal-to-noise ratio ,in The mean is represented by the subscript. Represents background noise; Average marginal gradient ,pass Operator computation; Noise pixel ratio .

[0047] Step S323: Dynamically adjust the median filter window and sharpening intensity. Dynamic parameter feedback: Dynamically calculate enhancement parameters based on features: Median filter window size: The constraint is an odd number, such as 3, 5, or 7. Desharpening mask intensity: The constraint is within the interval [0.1, 1.2]. For preset coefficients, These are the normalized parameters.

[0048] Step S324 involves performing median filtering and desharpening mask fusion enhancement. Specifically, sequence processing and fusion are first performed using a dynamic window... Preprocessed images Median filtering is performed to obtain the denoised image. Then, based on dynamic intensity For denoised images Perform unsharpening mask: ,in This is the result of Gaussian blurring.

[0049] Step S325: The enhanced image is input into the phase recovery module. With preprocessed images Linear fusion and normalization are performed to output an enhanced image. , as input It is then sent to the phase recovery PACS module.

[0050] Step S326: Start the high-precision reconstruction thread.

[0051] Step S327: Obtain high-precision reconstruction results.

[0052] Here we combine Figure 3 This step involves performing phase-aligned coherent summation and iterative reconstruction. The PACS module is the input. and Create independent but structurally identical processing threads. For any input, such as input... Image Perform the following iterations: Step S21: Segment the image into sub-image blocks and select an initial reference block. This step is for initialization: the enhanced image is segmented into m×n sub-image blocks, and one sub-block is randomly selected. As the initial reference block.

[0053] Step S22: Calculate the cross-correlation function between the sub-image patch and the reference patch, locate the peak value, estimate the translation amount of the sub-image patch, and cyclically align all sub-image patches. The translation estimation and alignment are: for the... Sub-block Calculate its relationship with the current reference block. Normalized cross-correlation function ,position Peak coordinates This is the translation amount. Cyclic displacement alignment is performed on all sub-image blocks to obtain the aligned sub-block set. .

[0054] Step S23 involves coherently summing the phases of the aligned sub-image blocks. Specific steps include coherent summation and fixed-amplitude synthesis.

[0055] Step S24: Calculate the average power value of all sub-image blocks as a fixed amplitude. Calculate the average power spectrum of all aligned sub-blocks. As a fixed amplitude, calculate the average phase. .

[0056] Step S25: Synthesize the frequency domain signal and perform an inverse transform to obtain the current reconstruction result. For example, the average power spectrum... Synthetic frequency domain signal .

[0057] Step S26: Set the current result as the new reference block. Inverse transform and reference update: For Perform an inverse Fourier transform to obtain the reconstruction result of the current iteration. This reconstruction result After phase alignment and coherent summation, the input image, such as the preprocessed image, is processed. Or enhance the image The current best estimate. In the next iteration, the current reconstruction result will be... Divide the image into m×n sub-image blocks, and randomly select one sub-block as the new reference block. Used to recalculate the reference block With the original input sub-image patch The cross-correlation function is used to perform more accurate displacement estimation and alignment.

[0058] Step S27: Check if the difference between adjacent iteration results is less than a preset threshold. Convergence determination: Calculate the difference between two adjacent reconstruction results using the following formula: , in, This is the result of the current iteration of reconstruction. This is the result of the previous iteration reconstruction.

[0059] If the difference is less than the preset threshold, that is ( For a preset threshold, such as If convergence is achieved, the preliminary reconstruction result of that channel is output. Otherwise, return to step S22 and continue iterating.

[0060] Input image for the other branch Repeat the same process to obtain the preliminary reconstruction results of the channel. .

[0061] Step S400: Adaptive fusion and output.

[0062] Step S500: Obtain preliminary reconstruction results and Then, the local quality of the two results is calculated and the fusion weights are determined. As a specific example, local gradient magnitude maps of the two images can be calculated to characterize their sharpness. For each pixel location... Calculate separately and gradient magnitude at this point and For example, using Operator. Then, the fusion weights are calculated: , ,in It is a very small positive number to prevent division by zero.

[0063] Step S600: Pixel-level weighted fusion is performed to generate the final reconstructed image, outputting the final high-quality target image. For example, for each pixel location (x, y), weighted fusion is performed according to its quality weight. and satisfy Perform pixel-level weighted fusion: .

[0064] This is the final high-resolution, high-contrast target reconstruction image output by the system.

[0065] To verify the technical effects of this invention, an imaging system as described in step S100 was constructed. The method of this application was used to process speckle images of a dynamic scattering medium acquired under narrowband ambient light interference with a signal-to-noise ratio (SNR) as low as -8.7 dB, and under broadband illumination of different bandwidths, such as 100 nm, 200 nm, and 280 nm. Experimental results show that, compared to traditional coherent averaging methods, single PACS, or MUSE processing paths, the speckle image reconstruction scheme provided in this application, based on a dual-input feedback framework, achieves superior reconstruction results. Specifically, on multiple test targets, the peak signal-to-noise ratio (PSNR) of the reconstructed images from this application is improved by an average of over 3 dB, and the structural similarity (SSIM) is significantly improved. Visually, the reconstructed images have clear edges, rich details, and effectively suppress background noise and scattering noise. Furthermore, by combining the processing flow of this invention with the classic bilateral filtering algorithm and applying it to the defogging enhancement of real foggy traffic scene images, the processed image shows significant improvements in contrast, detail visibility, and color fidelity compared to the original fog image, fully demonstrating the strong robustness and good generalization ability of the solution proposed in this application.

[0066] Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment of this application. Please refer to... Figure 4 At the hardware level, the device includes a processor 402, an internal bus 404, a network interface 406, memory 408, a hardware acceleration device 410, and non-volatile memory 412, and may also include other hardware required for its functions. One or more embodiments of this application can be implemented in software, for example, the processor 402 reads the corresponding computer program from the non-volatile memory 412 into memory 408 and then runs it. Of course, in addition to software implementation, one or more embodiments of this application do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the above processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0067] Figure 5 This is a block diagram illustrating a speckle image reconstruction apparatus based on dual-path feedback fusion, as shown in an exemplary embodiment of this application. This speckle image reconstruction apparatus can be applied to, for example... Figure 4The electronic device shown implements the technical solution of this application. The speckle image reconstruction apparatus of this embodiment includes: a preprocessing unit 510, an adaptive enhancement unit 520, a phase recovery unit 530, and a fusion unit 540, wherein: Preprocessing unit 510 preprocesses the original speckle image; The adaptive enhancement unit 520 performs adaptive image enhancement processing on the preprocessed image to obtain the enhanced image; The phase retrieval unit 530 performs phase retrieval processing directly on the preprocessed image to obtain the first image, and performs phase retrieval processing on the enhanced image to obtain the second image; The fusion unit 540 fuses the first image and the second image to output the target reconstructed image.

[0068] In some embodiments, the adaptive enhancement unit 520 calculates the feature parameters of the preprocessed image in real time and determines the adaptive enhancement parameters, performs filtering and desharpening mask processing on the image according to the adaptive enhancement parameters, and outputs the enhanced image.

[0069] In some embodiments, the phase restoration unit 530 performs the following steps on the preprocessed image and on the enhanced image, respectively: Step S21: Divide the input image into multiple sub-image blocks and select an initial reference block. The input image is the preprocessed image or the enhanced image. Step S22: Calculate the cross-correlation function between each sub-image block and the current reference block, estimate the translation amount of each sub-image block relative to the current reference block by locating the cross-correlation peak, and align the sub-image blocks according to the translation amount; Step S23: Coherently sum the spectral phases of all aligned sub-image blocks to obtain the average phase; Step S24: Average the power spectra of all sub-image blocks to obtain a fixed amplitude; Step S25: Combine the average phase and the fixed amplitude to form a frequency domain signal, and then perform an inverse transformation to obtain the current iteration reconstruction result; Step S26: Use the current iteration reconstruction result as the new reference block for the next iteration.

[0070] In some embodiments, the phase recovery unit 530 further performs the following steps: Step S27: Determine whether the difference between the reconstruction results of two adjacent iterations is less than a preset threshold; If so, determine convergence and output either the first image or the second image; Otherwise, return to step S22 and continue iterating.

[0071] In some embodiments, the fusion unit 540 calculates the local quality of the first image and the second image; and performs weighted fusion of the first image and the second image based on the local quality to obtain the target reconstructed image and output it.

[0072] In some embodiments, the preprocessing unit 510 performs frequency domain flat-field correction on the original speckle image to remove non-uniform illumination background; and performs Gaussian low-pass filtering on the corrected image to suppress high-frequency noise of the imaging system.

[0073] In some embodiments, the adaptive enhancement unit 520 calculates the signal-to-noise ratio, average edge gradient, and noise pixel ratio of the image in real time, and dynamically adjusts the window size of the median filter and the sharpening intensity of the desharpening mask according to the signal-to-noise ratio, average edge gradient, and noise pixel ratio.

[0074] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0075] Accordingly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the methods described in any of the above embodiments.

[0076] Accordingly, embodiments of this application also provide a computer program product configured to perform the methods described in any of the above embodiments.

[0077] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0078] In a typical configuration, a computer includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0079] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0080] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0081] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0082] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0083] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0084] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0085] The above description is only a part of the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A speckle image reconstruction method based on dual-path feedback fusion, characterized in that, Includes the following steps: Step S1: Preprocess the original speckle image; Step S2: Perform phase recovery processing directly on the preprocessed image to obtain the first image; Step S3: Perform adaptive image enhancement processing on the preprocessed image to obtain an enhanced image, and perform phase recovery processing on the enhanced image to obtain a second image; Step S4: Fuse the first image and the second image to output the target reconstructed image.

2. The method according to claim 1, characterized in that, The adaptive image enhancement process performed on the preprocessed image in step S3 includes: The system calculates the feature parameters of the preprocessed image in real time and determines the adaptive enhancement parameters. Based on the adaptive enhancement parameters, the image is filtered and desharpened using a mask, and the enhanced image is output.

3. The method according to claim 2, characterized in that, The phase retrieval processing in steps S2 and S3 includes performing the following steps on the preprocessed image and the enhanced image respectively: Step S21: Divide the input image into multiple sub-image blocks and select an initial reference block. The input image is the preprocessed image or the enhanced image. Step S22: Calculate the cross-correlation function between each sub-image block and the current reference block, estimate the translation amount of each sub-image block relative to the current reference block by locating the cross-correlation peak, and align the sub-image blocks according to the translation amount; Step S23: Coherently sum the spectral phases of all aligned sub-image blocks to obtain the average phase; Step S24: Average the power spectra of all sub-image blocks to obtain a fixed amplitude; Step S25: Combine the average phase and the fixed amplitude to form a frequency domain signal, and then perform an inverse transformation to obtain the current iteration reconstruction result; Step S26: Use the current iteration reconstruction result as the new reference block for the next iteration.

4. The method according to claim 3, characterized in that, The phase recovery process in steps S2 and S3 further includes: Step S27: Determine whether the difference between the reconstruction results of two adjacent iterations is less than a preset threshold; If so, determine convergence and output either the first image or the second image; Otherwise, return to step S22 and continue iterating.

5. The method according to any one of claims 1 to 4, characterized in that, In step S4, fusing the first image and the second image to output the target reconstructed image includes: Calculate the local quality of the first image and the second image; The first image and the second image are weighted and fused based on local quality to obtain the target reconstructed image and output.

6. The method according to claim 1, characterized in that, Step S1 includes: Frequency domain flat-field correction is performed on the original speckle image to remove non-uniform illumination background; Gaussian low-pass filtering is applied to the corrected image to suppress high-frequency noise in the imaging system.

7. The method according to claim 2, characterized in that, The real-time calculation of feature parameters of the preprocessed image and determination of adaptive enhancement parameters, followed by filtering and desharpening mask processing of the image based on the adaptive enhancement parameters, includes: The system calculates the signal-to-noise ratio, average edge gradient, and noise pixel ratio of the image in real time, and dynamically adjusts the window size of the median filter and the sharpening intensity of the desharpening mask based on the signal-to-noise ratio, average edge gradient, and noise pixel ratio.

8. A speckle image reconstruction device based on dual-path feedback fusion, characterized in that, include: The preprocessing unit preprocesses the original speckle image; The adaptive enhancement unit performs adaptive image enhancement processing on the preprocessed image to obtain the enhanced image; The phase retrieval unit performs phase retrieval processing directly on the preprocessed image to obtain the first image, and performs phase retrieval processing on the enhanced image to obtain the second image; The fusion unit fuses the first image and the second image to output the target reconstructed image.

9. An electronic device, characterized in that, include: processor; A computer-readable storage medium storing computer program instructions that, when executed by the processor, cause the processor to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is executed by a processor according to any one of claims 1 to 7.

Images