A low-light image enhancement method, system, device and medium

Abstract

The application discloses a low-illumination image enhancement method, system, device and medium, relates to the field of image processing, and comprises the following steps: performing color space conversion and channel decoupling on a low-illumination RGB image to obtain a luminance component; performing current mapping on the luminance component to obtain a driving current; utilizing the resistance state flipping probability of a magnetic tunnel junction device and the Sigmoid relationship of the applied current, performing nonlinear mapping enhancement on the luminance component based on the driving current to obtain an enhanced luminance component; and performing image reconstruction based on the enhanced luminance component to obtain an enhanced image. The application can realize efficient enhancement of the luminance of a low-illumination image and solve the problems of slow response, high power consumption, color distortion and poor edge adaptability of a traditional scheme.

Description

Technical Field

[0001] This application relates to the field of image processing, and in particular to a method, system, device and medium for low-light image enhancement. Background Technology

[0002] With the widespread deployment of imaging equipment, computer vision technology has been applied on a large scale in key areas such as security monitoring, aerial photography, and smart terminals, providing core support for scene information perception and intelligent decision-making. However, in low-light environments such as dusk and night, acquired images generally suffer from insufficient brightness, blurred details, low contrast, and color distortion, directly leading to a significant degradation in the performance of subsequent tasks such as target detection and image recognition, severely restricting the practical application effect of imaging equipment. Therefore, low-light image enhancement has become a core research hotspot in the field of computer vision, urgently requiring efficient, low-power, and highly adaptable solutions.

[0003] Existing low-light image enhancement technologies are mainly divided into two categories: one is traditional algorithm enhancement based on software, and the other is deep learning enhancement based on hardware acceleration.

[0004] In traditional software-based algorithm enhancement, the brightness adjustment method based on the Sigmoid function is widely used. This method improves the brightness of dark pixels through nonlinear mapping, but it relies entirely on iterative calculations by the central processing unit (CPU) / graphics processing unit (GPU), which has limitations such as large response latency, high computing power consumption, and poor hardware compatibility, making it difficult to meet the real-time and low-power requirements of edge terminals.

[0005] Hardware-accelerated deep learning enhancement often employs oxide memristors and two-dimensional semiconductor devices to construct synapses and neurons in convolutional neural networks (CNNs) to accelerate image enhancement tasks. However, this approach has significant drawbacks, such as: First, neural networks are black-box models, and their parameter optimization and decision-making processes are uninterpretable, failing to meet the high reliability requirements of scenarios like security monitoring; second, multi-layer convolution operations require numerous logic gates and complex control circuits, leading to high hardware integration difficulty, high system power consumption, and difficulty in adapting to miniaturized, low-cost edge terminal devices; third, coupled processing of color and luminance channels can easily cause color casts, overexposure, and other distortion problems, resulting in suboptimal visual quality of the enhanced image.

[0006] Magnetic tunnel junctions (MTJs), as a new generation of spintronic devices, possess advantages such as low power consumption, fast response, high integration density, and controllable resistance state switching. Their switching probability exhibits a natural sigmoid-like nonlinear relationship with the applied current, allowing for direct sigmoid function fitting at the hardware level without complex software iterations. Currently, existing technologies have not combined the sigmoid-like physical characteristics of MTJs with low-light image brightness decoupling enhancement, thus failing to fully leverage the hardware acceleration advantages of spintronic devices in edge image processing.

[0007] Therefore, existing low-light image enhancement technologies suffer from problems such as slow response, high power consumption, color distortion, and poor edge adaptation. There is an urgent need for a hardware-friendly, computationally efficient, and visually high-fidelity low-light image enhancement solution. Summary of the Invention

[0008] The purpose of this application is to provide a low-light image enhancement method, system, device and medium that can achieve efficient enhancement of the brightness of low-light images and solve the problems of slow response, high power consumption, color distortion and poor edge adaptation of traditional solutions.

[0009] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a low-light image enhancement method, including: Color space conversion and channel decoupling are performed on low-light RGB images to obtain the luminance component; low-light RGB images refer to RGB images acquired at dusk or night. The brightness component is current-mapped to obtain the driving current; By utilizing the sigmoid-like relationship between the resistance state reversal probability of a magnetic tunnel junction device and the applied current, a nonlinear mapping enhancement of the luminance component is performed based on the driving current to obtain the enhanced luminance component; wherein, the sigmoid-like relationship between the resistance state reversal probability of the magnetic tunnel junction device and the applied current is expressed as: ; In the formula, Let be the resistive state reversal probability of a magnetic tunnel junction device. To apply current, These are the fitting coefficients. Here are the parameters used to describe the curve offset, and exp() is an exponential function; An enhanced image is obtained by reconstructing the image based on the enhanced luminance component.

[0010] Secondly, this application provides a low-light image enhancement system, comprising: The conversion and decoupling module is used to perform color space conversion and channel decoupling on low-light RGB images to obtain the luminance component; low-light RGB images refer to RGB images acquired at dusk or night. A current mapping module is used to perform current mapping on the brightness component to obtain the driving current; A brightness enhancement module is used to perform nonlinear mapping enhancement of the brightness component based on the driving current by utilizing the resistance state reversal probability of the magnetic tunnel junction device and the sigmoid-like relationship of the applied current, thereby obtaining the enhanced brightness component; wherein, the sigmoid-like relationship between the resistance state reversal probability of the magnetic tunnel junction device and the applied current is expressed as: ; In the formula, Let be the resistive state reversal probability of a magnetic tunnel junction device. To apply current, These are the fitting coefficients. Here are the parameters used to describe the curve offset, and exp() is an exponential function; An image enhancement module is used to reconstruct the image based on the enhanced luminance components to obtain an enhanced image.

[0011] Thirdly, this application provides a computer 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 steps of the low-light image enhancement method provided above.

[0012] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the low-light image enhancement method described above.

[0013] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a low-light image enhancement method, system, device, and medium. By performing color space conversion and channel decoupling on low-light RGB images, it addresses the pain points of poor interpretability, complex integration, and high power consumption in deep learning hardware acceleration solutions. It achieves decoupled control of luminance and chrominance channels, avoiding color shift and detail loss during enhancement. By current mapping of the luminance component and utilizing the sigmoid-like relationship between the resistance state reversal probability of a magnetic tunnel junction device and the applied current, nonlinear fitting of the sigmoid function is performed at the hardware level. Based on the driving current, nonlinear mapping enhancement of the luminance component is achieved, resulting in an enhanced luminance component. This eliminates software iteration and complex neural network calculations, overcoming the shortcomings of traditional software sigmoid luminance adjustment methods, such as large computational latency and poor hardware compatibility. Simultaneously, image reconstruction based on the enhanced luminance component yields a high-brightness, high-detail, and high-color-fidelity enhanced image, achieving efficient luminance enhancement for low-light images and solving the problems of slow response, high power consumption, color distortion, and poor edge adaptation in traditional solutions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A schematic flowchart illustrating a low-light image enhancement method according to an embodiment of this application; Figure 2 This is a schematic diagram of the RGB color space provided in an embodiment of this application; Figure 3 This is a schematic diagram of the HSV color space provided in an embodiment of this application; Figure 4 A schematic diagram of the overall implementation architecture of a low-light image enhancement method provided in an embodiment of this application; Figure 5 A schematic diagram of a random characteristic test circuit for a magnetic tunnel junction device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the magnetoresistive loop test results provided in an embodiment of this application; Figure 7 This is a schematic diagram of the energy barrier of a magnetic tunnel junction device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the random flipping characteristic curve provided in an embodiment of this application; Figure 9This is a schematic diagram illustrating the statistical characteristics of random flipping time provided in an embodiment of this application; Figure 10 A schematic diagram of the fitting curve of the sigmoid-like relationship between the resistance state reversal probability and the current provided in an embodiment of this application; Figure 11 A schematic diagram illustrating the overall implementation flow of a low-light image enhancement method provided in an embodiment of this application; Figure 12 An array of enhancement results for low-light RGB images acquired by a camera in various low-light scenes, provided by different methods in an embodiment of this application; Figure 13 An array of enhancement results for low-light RGB images acquired by a drone in various low-light scenarios, provided by different methods in an embodiment of this application; Figure 14 A schematic diagram of the functional modules of a low-light image enhancement system provided in an embodiment of this application; Figure 15 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] This application provides a low-light image enhancement method, system, device, and medium based on magnetic tunnel junctions. By using MTJ hardware fitting of a sigmoid function and combining color space decoupling processing, it achieves efficient brightness enhancement of low-light images while solving the problems of slow response, high power consumption, color distortion, and poor edge adaptation in traditional solutions.

[0018] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] In one exemplary embodiment, this application provides a low-light image enhancement method. This method is executed by a computer device, specifically by a terminal or server, or by both. In this embodiment, the method is described using a server as an example. Figure 1 As shown, the method includes: Step 100: Perform color space conversion and channel decoupling on the low-light RGB image to obtain the luminance component; the low-light RGB image refers to the RGB image acquired at dusk or night. Step 101: Perform current mapping on the luminance component to obtain the driving current; Step 102: Utilizing the sigmoid-like relationship between the resistive state reversal probability of the magnetic tunnel junction device and the applied current, nonlinear mapping enhancement of the luminance component is performed based on the driving current to obtain the enhanced luminance component; this step is achieved through the intrinsic physical response of the magnetic tunnel junction device. Step 103: Reconstruct the image based on the enhanced luminance component to obtain the enhanced image.

[0020] By implementing steps 100-103 above, this application achieves efficient enhancement of low-light image brightness through nonlinear mapping based on physical mechanisms, with a clear and traceable decision-making process. This meets the requirements of high-reliability scenarios such as security monitoring and solves the problems of slow response, high power consumption, color distortion, and poor edge adaptation in traditional solutions.

[0021] In an exemplary embodiment of this application, in order to perfectly adapt to resource-constrained edge terminals, micro-monitoring equipment, aerial photography equipment, etc., and to be compatible with real-time processing of multi-resolution images, the magnetic tunnel junction device used in this application can be an in-plane magnetic tunnel junction with weakly pinned reference layers of a film stack structure; wherein, the film stack structure is, in sequence: Ta with a thirteenth predetermined thickness (e.g., 3 nm), CuN with a third predetermined thickness (e.g., 10 nm), Ta with a fourth predetermined thickness (e.g., 5 nm), Ru with a fifth predetermined thickness (e.g., 5 nm), IrMn with a sixth predetermined thickness (e.g., 8 nm), CoFe with a seventh predetermined thickness (e.g., 2.3 nm), Ru with an eighth predetermined thickness (e.g., 0.7 nm), and Co with a ninth predetermined thickness (e.g., 2.5 nm). 40 Fe 40 B 20 MgO with a first set thickness (e.g., 0.85 nm) and Co with a tenth set thickness (e.g., 2.0 nm). 40 Fe 40 B 20 The second set thickness (e.g., 0.4 nm) of MgO, the eleventh set thickness (e.g., 8 nm) of Ta, and the twelfth set thickness (e.g., 7 nm) of Ru.

[0022] In this film stacking structure, a composite film with a thirteenth predetermined thickness (e.g., 3 nm) of Ta, a third predetermined thickness (e.g., 10 nm) of CuN, a fourth predetermined thickness (e.g., 5 nm) of Ta, and a fifth predetermined thickness (e.g., 5 nm) of Ru is used as the bottom electrode; a film with a sixth predetermined thickness (e.g., 8 nm) of IrMn is used as the antiferromagnetic layer; and a film with a seventh predetermined thickness (e.g., 2.3 nm) of CoFe, an eighth predetermined thickness (e.g., 0.7 nm) of Ru, and a ninth predetermined thickness (e.g., 2.5 nm) of Co is used. 40 Fe 40 B 20 The resulting composite film serves as a synthetic antiferromagnetic reference layer, providing an exchange bias field; the tenth layer contains a Co layer with a predetermined thickness (e.g., 2.0 nm). 40 Fe 40 B 20 The formed film serves as the free layer; a film of MgO with a second predetermined thickness (e.g., 0.4 nm) serves as the capping layer; a composite film of Ta with an eleventh predetermined thickness (e.g., 8 nm) and Ru with a twelfth predetermined thickness (e.g., 7 nm) serves as the top electrode; a tunneling barrier formed by MgO with a first predetermined thickness (e.g., 0.85 nm) separates the free layer and the synthesized antiferromagnetic reference layer. In practical applications, the magnetic tunnel junction device can be fabricated into an elliptical structure of 70 nm × 245 nm using micro-nano technology.

[0023] By setting the thickness of the spacer layer Ru in the synthesized antiferromagnetic reference layer to slightly lower than its commonly used optimized value (i.e., the preset thickness value), the exchange coupling (exchange bias effect) inside the synthesized antiferromagnetic reference layer can be weakened, thereby effectively reducing the pinning field of the synthesized antiferromagnetic reference layer. Combined with the anisotropic shape of the magnetic tunnel junction device (i.e., an elliptical structure of 70nm×245nm), a precise sigmoid-like response of magnetization reversal probability and current can be achieved, ensuring smooth brightness mapping and moderate contrast.

[0024] In an exemplary embodiment of this application, in order to completely avoid color cast and overexposure distortion, independent control of brightness and chromaticity features is achieved to prevent brightness enhancement from damaging color harmony. The implementation process of step 100 above in this application can be described as follows: mapping the RGB image to the HSV color space, and performing channel decoupling to separate the hue components. H saturation component S and brightness component V .

[0025] In this context, the RGB color space in which RGB images reside is a device-dependent additive color model, with the structure as follows: Figure 2As shown, this is a unit cube. This additive color model is built in a three-dimensional Cartesian coordinate system, where the three orthogonal axes represent the three primary color channels: red (R), green (G), and blue (B), respectively, and the intensity of each channel is normalized to the range [0,1]. Figure 2 In the diagram, the 8 vertices of the RGB cube correspond to the following extreme color states: (1) Three primary colors: red (1,0,0), green (0,1,0), blue (0,0,1).

[0026] (2) Three secondary colors: cyan (0,1,1), magenta (1,0,1), yellow (1,1,0).

[0027] (3) Neutral color extremes: black (0,0,0), white (1,1,1).

[0028] Along the main diagonal of the cube extending from the black vertex to the white vertex, all points... R , G , B The component values ​​are all equal, forming a continuous grayscale spectrum, which can directly characterize the pixel intensity changes in a grayscale image.

[0029] This additive color model is highly compatible with the physical principles of image acquisition (such as the response of a camera sensor to the RGB band) and image display (such as pixels emitting RGB light), and therefore has become the native color space of the vast majority of imaging hardware.

[0030] In contrast, the HSV (Hue, Saturation, Lightness) color space is an intuitive model that aligns with human visual perception, such as... Figure 3 The image shown is a right circular cone. This intuitive model separates color information from brightness information, facilitating color manipulation in image processing that more closely aligns with human visual perception. For example... Figure 3 As shown, hue is defined as the angular position around the central axis of the HSV cone, ranging from 0° to 360°, corresponding to a continuous color wheel: 0° for red, 60° for yellow, 120° for green, 180° for cyan, 240° for blue, and 300° for magenta. Saturation is the radial parameter of the cone, ranging from 0 to 1, used to characterize the purity of the color: a saturation of 0 represents a neutral gray distributed along the central axis of the cone; a saturation of 1 corresponds to the most vivid and highly saturated pure color on the circumference of the cone's cross-section. Brightness is the longitudinal parameter along the central axis of the cone, ranging from 0 to 1, used to quantify the lightness or darkness of the color: a brightness of 0 corresponds to black at the apex of the cone, and a brightness of 1 corresponds to the maximum brightness at the base of the cone.

[0031] This HSV color model based on a conical structure (i.e.) Figure 3 The core advantage of the intuitive model shown is: color components ( H , S) and brightness component ( V Completely decoupled, it can independently adjust the image brightness without changing the original color of the scene, thus having important application value in the task of brightness enhancement in low-light imaging.

[0032] During the conversion from RGB to HSV color space, the hue component... H saturation component S and brightness component V It can be described by equations (1)-(3), we have: (1) (2) (3) In the formula, This represents the modulo operation. Indicates "if". , , These are three channel components: red, green, and blue. Δ The width of the image pixels is the dynamic change. , The maximum value among the R, G, and B channels. This represents the minimum value among the R, G, and B channels. "Otherwise" indicates other cases.

[0033] The mathematical calculation process from HSV to RGB color space is shown in equations (4) to (8).

[0034] (4) (5) (6) (7) (8) In the formula, To be based on hue components H The RGB intermediate normalized components generated within the 60° segment interval. This is the brightness compensation value, used to map the intermediate components to the final RGB brightness range. For color purity, This is an intermediate transition value.

[0035] Based on the above description, the intermediate component is superimposed with a brightness compensation value using equation (4). m The final standard RGB color space value is obtained. R , G , BThis completes the color space conversion from HSV to RGB.

[0036] In one exemplary embodiment of this application, in order to further improve the processing speed, the implementation process of step 101 provided above in this application may include: Step 1: Normalize the pixel values ​​in the luminance component that are lower than the set value to obtain the normalized result; the set value can be set according to the specific use case, and no specific limitation is made here.

[0037] Step 2: Map the normalization result to a drive current with an amplitude corresponding to the pixel value, so as to be delivered to the magnetic tunnel junction device.

[0038] In an exemplary embodiment of this application, based on the description of step 102 above, the Sigmoid function is directly fitted using the physical characteristics of the MTJ, which eliminates the need for software iteration and complex neural network operations, resulting in a processing speed significantly superior to traditional and lightweight deep learning schemes. The Sigmoid-like relationship between the resistive state reversal probability of the magnetic tunnel junction device and the applied current used in step 102 above can be expressed as: (9) In the formula, Let be the resistive state reversal probability of a magnetic tunnel junction device. To apply current (i.e., the applied bias current); The fitting coefficients reflect the steepness of the change in the resistance state reversal probability with current. These are parameters used to describe the curve offset.

[0039] By utilizing the intrinsic physical response of a magnetic tunnel junction device, nonlinear mapping enhancement of the brightness component can be achieved, resulting in the enhanced brightness component. V Based on this, the implementation process of step 103 provided above in this application includes: Step 1: The enhanced luminance, hue, and saturation components are spliced ​​and blended to obtain the blended result; Step 2: Remap the fusion result to the RGB color space to obtain the enhanced image (i.e., the low-light enhanced image).

[0040] In one exemplary embodiment of this application, the process of fabricating and testing the performance of the magnetic tunnel junction device may include: Step 1. Fabricate the magnetic tunnel junction device.

[0041] A film stack structure was grown on a thermally heated silicon oxide substrate using magnetron sputtering technology. The sequence of deposition included a bottom electrode, an antiferromagnetic layer, a synthetic antiferromagnetic reference layer, a tunneling barrier, a free layer, a protective layer, and a top electrode. The initial film stack structure was annealed for 2 hours at 300°C under a 1T in-plane magnetic field, with the magnetic field applied along the short axis of the initial stack structure. Junction regions were patterned using electron beam lithography and ion etching. A silicon dioxide insulating layer was prepared by inductively coupled plasma chemical vapor deposition (ICPCVD), and the top electrode was fabricated by magnetron sputtering, ultimately yielding a 70nm × 245nm elliptical structure (i.e., a magnetic tunnel junction device, hereinafter referred to as the device). The energy barrier of the magnetic tunnel junction device is as follows: Figure 7 As shown.

[0042] Step 2. Performance testing.

[0043] A test platform was constructed using a DC source meter and a nanovolt meter, connected to a computer via a General-Purpose Interface Bus (GPIB), and the test process was controlled by Agilent VEE Pro programming to perform the following tests: (1) Magnetoresistive characteristic test: Under a DC bias of 10μA, a magnetic field was applied along the major and minor axes of the device, and the coercive field of the reference layer was measured to be approximately ±160Oe and ±135Oe, respectively, to verify the weak pinning characteristic; among which, the random characteristic test circuit is as follows Figure 5 As shown.

[0044] (2) Random flip-over characteristic test: Under an external magnetic field of 135 Oe (optimal flip-over field strength), DC currents of different amplitudes were applied, and the resistive state flip-over probability was monitored. The sigmoid-like relationship between the resistive state flip-over probability and the current was obtained, and the fitted curve is shown in the figure. Figure 10 As shown in the figure. The resistance switching characteristic curve and the statistical characteristics of random switching time are respectively shown in the figure. Figure 8 and Figure 9 As shown, the lap test results of magnetoresistive resistance are as follows: Figure 6 As shown.

[0045] Test results confirm that the device provided in this application has complete flipping, high contrast, excellent signal-to-noise ratio and reading accuracy, and can accurately realize nonlinear mapping of luminance components.

[0046] In one exemplary embodiment of this application, when the enhanced low-light RGB image is captured by a camera or a drone, the overall implementation flow of the low-light image enhancement method provided in this application is as follows: Figure 4 As shown. Figure 4In this context, a low-light image is a low-illuminance RGB image, low-light features refer to the luminance component obtained through color space conversion and channel decoupling, enhancement features refer to the enhanced luminance component, probability refers to the resistance state reversal probability of a magnetic tunnel junction device, and DC current refers to the applied current.

[0047] Based on the above description, the overall implementation process of the low-light image enhancement method provided in this application is as follows: Figure 11 As shown.

[0048] In one exemplary embodiment of this application, three sets of control algorithms are set up on a low-light image dataset: a traditional sigmoid software scheme, a lightweight deep learning (Illumination Adaptive Transformer, IAT) scheme, and the method provided in this application; the test metrics include Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and processing time. PSNR and SSIM are two commonly used supervised evaluation metrics based on reference images. They quantitatively evaluate the quality of the enhanced image by calculating the differences and correlations between the enhanced image and the reference image.

[0049] PSNR is an objective metric used to evaluate the degree of image distortion. It enhances the degree of distortion of an image relative to a reference image by quantizing the error pixel by pixel, and its value ranges from [0, +∞). Its calculation depends on the mean square error, and the specific formula is as follows: (10) In the formula, MAX I The maximum value of a pixel is represented by , and MSE represents the mean square error. A higher PSNR value indicates less distortion or noise in the enhanced image, a closer resemblance to the reference image, and better performance.

[0050] Unlike PSNR, which focuses on pixel-by-pixel error, SSIM evaluates the degree to which an enhanced image retains structural and texture information relative to a reference image by comparing brightness, contrast, and structural features, with values ​​ranging from [0,1]. For a given output image... and reference image The specific formula for SSIM is: (11) In the formula, and A constant introduced to ensure numerical stability and avoid a denominator of zero. Output image. and reference image mean , Standard deviation , and covariance All values ​​were calculated using Gaussian filtering. A higher SSIM value indicates better preservation of structural information.

[0051] Based on the above description, the performance of the traditional Sigmoid software solution, the IAT solution, and the method provided in this application were evaluated on a computer. The computer's hardware configuration was as follows: Intel(R) Core(TM) i5-9400 CPU @2.90GHz, 16GB of RAM, and an NVIDIA GeForce RTX 2070 graphics processor (GPU). The development environment was PyCharm.

[0052] To ensure fairness in the comparison, the training and testing code based on the learning method in this embodiment are both taken from the database corresponding to the original software. For example... Figure 12 and Figure 13 The comparison results of the three sets of algorithms shown in various low-light scenarios reveal that the traditional Sigmoid software scheme disrupts the synergistic relationship between the H, S, and V channels, resulting in significant feature distortion in the output. The IAT scheme, a reconstruction algorithm based on a neural network design, fails to fully consider the inherent color consistency constraints of natural scenes and the nonlinear coupling relationship between color channels, leading to a certain chromaticity shift in its reconstruction results, exhibiting an overall gray-green visual deviation. In contrast, the method proposed in this application effectively preserves the structural details and color fidelity of the original image, and the enhancement results are more consistent with human visual perception in various low-light environments, both indoors and outdoors. Based on this, the test results are as follows: 1. Image quality indicators: Compared with the traditional Sigmoid software solution, this application improves PSNR by 3.82 dB and SSIM by 9.91%; compared with the IAT solution, SSIM is improved by 13.46%, with no obvious color cast and complete detail retention; 2. Time consumption metrics: The time consumption of this application is comparable to that of the traditional Sigmoid software solution. For images with resolutions of 128×128, 256×256, 512×512, and 1024×1024, it reduces the time consumption by 17.92 ms, 33.89 ms, 96.36 ms, and 417.23 ms respectively compared to the IAT solution. 3. Adaptability: This application can stably process multi-resolution images, has low hardware power consumption, and can be directly deployed on edge terminals and embedded imaging devices.

[0053] Figure 12 and Figure 13 In the middle, the first column consists of low-light RGB images ( Figure 12The data is captured by the camera in different scenarios. Figure 13 The first column contains images acquired by drones in different scenarios; the second column contains enhanced images from the traditional Sigmoid software solution; the third column contains enhanced images from the IAT solution; and the fourth column contains enhanced images from the method provided in this application. Figure 12 The fifth column is a reference image.

[0054] Based on the same inventive concept, this application also provides a low-light image enhancement system for implementing the low-light image enhancement method described above. The solution provided by this system is similar to the implementation described in the above method; therefore, the specific limitations in one or more low-light image enhancement system embodiments provided below can be found in the limitations of the low-light image enhancement method described above, and will not be repeated here.

[0055] In one exemplary embodiment, such as Figure 14 As shown, a low-light image enhancement system is provided, including: a conversion decoupling module, a current mapping module, a brightness enhancement module, and an image enhancement module.

[0056] The conversion and decoupling module is used to perform color space conversion and channel decoupling on low-light RGB images to obtain the luminance component; low-light RGB images refer to RGB images acquired at dusk or night. The current mapping module is used to perform current mapping on the luminance components to obtain the driving current; The brightness enhancement module utilizes the sigmoid-like relationship between the resistance state reversal probability of the magnetic tunnel junction device and the applied current to perform nonlinear mapping enhancement of the brightness component based on the driving current, thereby obtaining the enhanced brightness component. The image enhancement module is used to reconstruct the image based on the enhanced luminance component to obtain an enhanced image.

[0057] The system provided in this application can be integrated into a micro embedded platform to directly interface with surveillance cameras, aerial cameras, etc., to achieve real-time low-light image enhancement, thereby meeting the high-definition imaging needs of outdoor, nighttime and other scenarios.

[0058] In summary, this application, based on mature MTJ device fabrication technology and image processing workflow, can achieve mass production and engineering deployment. It is applicable to scenarios such as security monitoring, drone aerial photography, vehicle night vision, and smart terminals, thereby significantly improving imaging quality and processing efficiency in low-light environments. It has strong industrial application value and market promotion prospects.

[0059] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 15As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores low-light image enhancement data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a low-light image enhancement method.

[0060] Those skilled in the art will understand that Figure 15 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment to which the present application is applied. Specific computer equipment may include, for example, [the following is a list of possible additional structures]. Figure 15 The diagram shows more or fewer components, or combinations of certain components, or different component arrangements.

[0061] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0062] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0063] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0064] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0065] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (RRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0066] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0067] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0068] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A low-light image enhancement method, characterized in that, include: Color space conversion and channel decoupling are performed on low-light RGB images to obtain the luminance component; Low-light RGB images refer to RGB images acquired at dusk or night. The brightness component is current-mapped to obtain the driving current; By utilizing the sigmoid-like relationship between the resistance state reversal probability of a magnetic tunnel junction device and the applied current, a nonlinear mapping enhancement of the luminance component is performed based on the driving current to obtain the enhanced luminance component; wherein, the sigmoid-like relationship between the resistance state reversal probability of the magnetic tunnel junction device and the applied current is expressed as: ； In the formula, Let be the resistive state reversal probability of a magnetic tunnel junction device. To apply current, These are the fitting coefficients. Here are the parameters used to describe the curve offset, and exp() is an exponential function; An enhanced image is obtained by reconstructing the image based on the enhanced luminance component.

2. The low-light image enhancement method according to claim 1, characterized in that, The magnetic tunnel junction device is an in-plane magnetic tunnel junction with a weakly pinned reference layer; the in-plane magnetic tunnel junction with a weakly pinned reference layer adopts a film stacking structure. In the film stacked structure, a composite film formed of Ta, CuN, Ta, and Ru is used as the bottom electrode; an IrMn film is used as the antiferromagnetic layer; and a CoFe, Ru, and Co film is used as the bottom electrode. 40 Fe 40 B 20 The resulting composite film serves as the synthetic antiferromagnetic reference layer; Co is used. 40 Fe 40 B 20 The formed film serves as the free layer; a film of a second predetermined thickness formed using MgO serves as the capping layer; a composite film formed using Ta and Ru serves as the top electrode; the free layer and the synthetic antiferromagnetic reference layer are separated by a tunneling barrier formed using MgO of a first predetermined thickness.

3. The low-light image enhancement method according to claim 2, characterized in that, In the synthetic antiferromagnetic reference layer, the Ru-formed film serves as a spacer layer; the thickness of the spacer layer is lower than a preset thickness value.

4. The low-light image enhancement method according to claim 1, characterized in that, The magnetic tunnel junction device is an elliptical structure with a diameter of 70 nm × 245 nm.

5. The low-light image enhancement method according to claim 1, characterized in that, The luminance component is current-mapped to obtain the driving current, including: The pixel values ​​in the luminance component that are lower than the set value are normalized to obtain the normalized result; The normalization result is mapped to a driving current with an amplitude corresponding to the pixel value.

6. The low-light image enhancement method according to claim 1, characterized in that, Color space conversion and channel decoupling are performed on low-light RGB images to obtain the luminance component, including: The RGB image is mapped to the HSV color space, and the channels are decoupled to separate the hue component, saturation component, and brightness component.

7. The low-light image enhancement method according to claim 6, characterized in that, Image reconstruction is performed based on the enhanced luminance components to obtain an enhanced image, including: The enhanced luminance component, hue component, and saturation component are spliced ​​and blended to obtain a fusion result; The fusion result is remapped to the RGB color space to obtain the enhanced image.

8. A low-light image enhancement system, characterized in that, include: The conversion and decoupling module is used to perform color space conversion and channel decoupling on low-light RGB images to obtain the luminance component; Low-light RGB images refer to RGB images acquired at dusk or night. A current mapping module is used to perform current mapping on the brightness component to obtain the driving current; A brightness enhancement module is used to perform nonlinear mapping enhancement of the brightness component based on the driving current by utilizing the resistance state reversal probability of the magnetic tunnel junction device and the sigmoid-like relationship of the applied current, thereby obtaining the enhanced brightness component; wherein, the sigmoid-like relationship between the resistance state reversal probability of the magnetic tunnel junction device and the applied current is expressed as: ； In the formula, Let be the resistive state reversal probability of a magnetic tunnel junction device. To apply current, These are the fitting coefficients. Here are the parameters used to describe the curve offset, and exp() is an exponential function; An image enhancement module is used to reconstruct the image based on the enhanced luminance components to obtain an enhanced image.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the low-light image enhancement method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the low-light image enhancement method according to any one of claims 1-7.

Images