An adaptive low-light image enhancement method, system, terminal device, and medium

Abstract

This invention discloses an adaptive low-light image enhancement method, system, terminal device, and medium, relating to the field of image processing technology. The method includes: acquiring an image to be enhanced and determining the brightness value of each pixel; processing the brightness values ​​using a parameterized processing function to generate complementary information; constructing an intuitive fuzzy image based on the complementary information and brightness values ​​to express the uncertainty of pixel values; enhancing the contrast of the intuitive fuzzy image based on the global statistical features of the image to obtain a preliminary enhanced image; iteratively optimizing the adjustable parameters of the parameterized processing function with the optimization objective of maximizing the information content or perceptual quality of the enhanced image to obtain the optimal parameter enhanced image; and converting it from the fuzzy domain back to the spatial domain to obtain the final enhanced image. This invention improves the contrast of low-light images while preserving structural details and enhancing visual quality. It also exhibits strong adaptability and high computational efficiency, making it widely applicable to imaging systems under complex lighting conditions.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to an adaptive low-light image enhancement method, system, terminal device, and medium. Background Technology

[0002] With the rapid development of digital imaging technology, low-light image enhancement has become a preprocessing step in fields such as surveillance systems, medical imaging, autonomous navigation, remote sensing, and multimedia processing. Images captured under insufficient lighting conditions generally suffer from poor sharpness, low contrast, loss of texture information, and severe noise interference, affecting the performance of subsequent tasks such as target detection, recognition, and scene analysis.

[0003] To address the aforementioned issues, existing technologies primarily employ the following methods: First, histogram equalization and its improved algorithms, which enhance global contrast by redistributing image grayscale, but are prone to over-enhancement and amplification of noise in dark areas. Second, methods based on Retinex theory, a computational theory based on the human visual system's color constancy perception, which enhances the visibility of dark areas by separating illumination and reflection components; however, methods based on this theory may produce color distortion and have complex parameter tuning requirements. Third, deep learning methods, which learn illumination correction mappings from large-scale data, but require substantial training data and computational resources. Fourth, methods based on fuzzy set theory, including intuitionistic fuzzy methods; however, these methods rely on predefined complementary functions or fixed parameter configurations, have limited adaptive capabilities, and typically require post-processing operations such as histogram equalization, increasing system complexity.

[0004] Therefore, there is an urgent need for a low-light image enhancement method that is more adaptive, more computationally efficient, and does not rely on histogram post-processing, to fill the gap in existing technologies. Summary of the Invention

[0005] The technical problem this invention aims to solve is that, in the field of low-light image enhancement, existing methods suffer from several drawbacks, including reliance on histogram post-processing, fixed parameter configurations that make it difficult to adapt to different lighting conditions, and high computational complexity that hinders real-time deployment. These issues make it difficult to simultaneously achieve enhancement effects, adaptability, and computational efficiency. Therefore, an effective solution is urgently needed to address these technical problems.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides an adaptive low-light image enhancement method, the method comprising: Acquire the image to be enhanced and determine the brightness value of each pixel in the image to be enhanced; The brightness value is processed by a parameterized processing function to generate complementary information for each pixel; Based on the complementary information and the brightness value, an intuitive blurred image is constructed, wherein the intuitive blurred image is used to express the uncertainty of pixel values; Based on the global statistical features of the image to be enhanced, contrast enhancement is performed on the intuitively blurred image to obtain a preliminary enhanced image; With the goal of maximizing the information content or perceptual quality of the enhanced image, the adjustable parameters in the parameterized processing function are iteratively optimized to obtain the optimal parameter-enhanced image. The image enhanced by the optimal parameters is transformed from the fuzzy domain back to the spatial domain to obtain the final enhanced image; The parameterization function is a generalized fuzzy complement function, for input values ​​whose domain is within the target numerical range. The generalized fuzzy complement function Defined as:

[0007] in, , It is an adjustable parameter used to control the shape of the fuzzy complement transform; The complementary information includes membership degree, non-membership degree, and hesitation degree. The step of processing the brightness values ​​using a parameterized processing function to generate complementary information for each pixel includes: Using the brightness value as input to the generalized fuzzy complement function, the membership degree of each pixel is calculated, expressed as:

[0008] in, Indicates the location of the image to be enhanced. The brightness value of that pixel. Indicates the location of the image to be enhanced. The membership degree of a pixel; Using the membership degree as input to the generalized fuzzy complement function, the non-membership degree of each pixel is calculated, expressed as:

[0009] in, Indicates the location of the image to be enhanced. The non-membership degree of a pixel; Based on the membership degree and the non-membership degree, the hesitation degree of each pixel is calculated and expressed as:

[0010] in, Indicates the location of the image to be enhanced. Hesitation at each pixel; The construction of an intuitive blurred image based on the complementary information and the brightness value includes: The sum of the membership degree and the hesitation degree of each pixel in the image to be enhanced is used as the pixel value of the intuitive blur image to construct the intuitive blur image, wherein the pixel value of the intuitive blur image is represented as:

[0011] in, Indicates the location of the intuitive blurred image. The pixel value of the pixel at that location.

[0012] In one implementation, acquiring the image to be enhanced and determining the brightness value of each pixel in the image to be enhanced includes: Acquire a low-light image as the image to be enhanced; The grayscale values ​​of each pixel in the image to be enhanced are obtained, and the grayscale values ​​are normalized to the target value range by linear stretching to obtain the normalized grayscale values ​​of each pixel as the brightness values.

[0013] In one implementation, the step of performing contrast enhancement on the intuitively blurred image based on the global statistical features of the image to be enhanced to obtain a preliminary enhanced image includes: Calculate the global average gray level and gray level standard deviation of the intuitively blurred image; Calculate the target contrast based on the global average gray level and the gray level standard deviation; Using the target contrast, a contrast enhancement transformation is performed on the intuitively blurred image to obtain the preliminary enhanced image.

[0014] In one implementation, the step of iteratively optimizing the adjustable parameters in the parameterized processing function to obtain the optimal parameter-enhanced image, with the optimization objective of maximizing the information content or perceptual quality of the enhanced image, includes: Calculate the Shannon entropy of the enhanced image; By employing a refined search strategy, the range of values ​​for the adjustable parameters is iteratively narrowed, and the parameter value that maximizes the Shannon entropy is searched as the optimal parameter, thus obtaining the optimal parameter-enhanced image of the image to be enhanced.

[0015] Secondly, embodiments of the present invention also provide an adaptive low-light image enhancement system, the system comprising: The data acquisition module is used to acquire the image to be enhanced and determine the brightness value of each pixel in the image to be enhanced. The complementary information calculation module is used to process the brightness value through a parameterized processing function to generate complementary information for each pixel; An intuitive blurred image construction module is used to construct an intuitive blurred image based on the complementary information and the brightness value, wherein the intuitive blurred image is used to express the uncertainty of pixel values; An image enhancement module is used to perform contrast enhancement on the intuitively blurred image based on the global statistical features of the image to be enhanced, so as to obtain a preliminary enhanced image; The parameter optimization module is used to iteratively optimize the adjustable parameters in the parameterized processing function with the goal of maximizing the information content or perceptual quality of the enhanced image, so as to obtain the optimal parameter-enhanced image. The fuzzy domain conversion module is used to convert the enhanced image with the optimal parameters back from the fuzzy domain to the spatial domain, so as to obtain the final enhanced image.

[0016] Thirdly, embodiments of the present invention also provide a terminal device, the terminal device including a memory, a processor, and an adaptive low-light image enhancement program stored in the memory and executable on the processor, wherein when the processor executes the adaptive low-light image enhancement program, it implements the steps of the adaptive low-light image enhancement method described in any of the above schemes.

[0017] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing an adaptive low-light image enhancement program, wherein when the adaptive low-light image enhancement program is executed by a processor, it implements the steps of the adaptive low-light image enhancement method described in any of the above schemes.

[0018] Beneficial Effects: This invention discloses an adaptive low-light image enhancement method, system, terminal device, and medium, relating to the field of image processing technology. The method first acquires the image to be enhanced and determines the brightness value of each pixel in the image. Then, the brightness values ​​are processed using a parameterized processing function to generate complementary information for each pixel. Based on the complementary information and the brightness values, an intuitive blur image is constructed, whereby the intuitive blur image is used to express the uncertainty of pixel values. Subsequently, based on the global statistical features of the image to be enhanced, contrast enhancement is performed on the intuitive blur image to obtain a preliminary enhanced image. Next, with the optimization objective of maximizing the information content or perceptual quality of the enhanced image, the adjustable parameters in the parameterized processing function are iteratively optimized to obtain the optimal parameter enhanced image. Finally, the optimal parameter enhanced image is transformed from the fuzzy domain back to the spatial domain to obtain the final enhanced image.

[0019] This invention effectively models the uncertainty and illumination variations in low-light images by constructing an intuitive blurred image to express pixel value uncertainty. It exhibits strong adaptability to different lighting conditions and generates complementary information through a parameterized processing function for enhancement, avoiding over-enhancement and noise amplification without relying on histogram equalization post-processing. Furthermore, by iteratively optimizing the adjustable parameters of the parameterized processing function with the goal of maximizing the information content or perceptual quality of the enhanced image, it automatically determines the optimal enhancement parameters without manual tuning. Simultaneously, this invention does not require the large-scale training datasets and extensive computational resources needed for deep learning, resulting in high computational efficiency. This invention improves the contrast of low-light images while preserving structural details and enhancing visual quality, making it widely applicable to various imaging systems under complex lighting conditions. Attached Figure Description

[0020] Figure 1 A flowchart illustrating a specific implementation of the adaptive low-light image enhancement method provided in this embodiment of the invention.

[0021] Figure 2 This is a schematic diagram of the image enhancement framework of the adaptive low-light image enhancement method provided in an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram illustrating the adaptive parameter optimization of the image enhancement framework of the adaptive low-light image enhancement method provided in this embodiment of the invention.

[0023] Figure 4 The input and output diagrams show the example images processed by the adaptive low-light image enhancement method provided in this embodiment of the invention.

[0024] Figure 5 This is a comparison chart showing the effects of different enhancement methods on example image 1.

[0025] Figure 6 This is a comparison chart showing the effects of different enhancement methods on example image 2.

[0026] Figure 7 This is a block diagram illustrating the principle of the adaptive low-light image enhancement device provided in this embodiment of the invention.

[0027] Figure 8 This is a block diagram illustrating the internal structure of the terminal device provided in an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0029] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content, operations, or steps, nor does it require execution in the described order. For example, some operations or steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0030] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0031] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. For example, "first control information" and "second control information" are only used to distinguish different control information and do not limit their order.

[0032] Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or the order of execution, and that the words "first" and "second" do not necessarily imply that they are different.

[0033] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0034] With the rapid development of digital imaging technology, low-light image enhancement has become an essential preprocessing step in many applications such as surveillance systems, medical imaging, autonomous navigation, remote sensing, and multimedia processing. Images captured in low-light environments often suffer from poor sharpness, low contrast, loss of texture information, and severe noise interference. These problems can seriously affect the quality of visual perception and may adversely impact subsequent image analysis tasks such as target detection, recognition, and scene interpretation.

[0035] To address these challenges, various image enhancement methods have been developed. Traditional methods include histogram equalization and its improved algorithms, which aim to redistribute image grayscale to enhance global contrast. Models based on Retinex theory improve the visibility of dark areas by separating the illumination and reflection components of an image. In recent years, deep learning-based methods have been proposed, learning complex illumination correction mapping relationships from large-scale datasets.

[0036] Furthermore, fuzzy set theory and its extended methods have been introduced into image enhancement tasks because they can model the uncertainty of pixel gray levels. Fuzzy theory-based methods typically convert image gray levels into fuzzy membership degrees, adjust brightness and contrast through transformation functions, and then reconstruct the enhanced image. Some studies have further extended this approach to intuitionistic fuzzy representation, simultaneously fusing membership and non-membership information.

[0037] Despite these improvements, existing techniques still have many limitations. Histogram-based methods are prone to over-enhancing results and often amplify noise in dark areas. Retinex-based methods sometimes produce unnatural color distortions and often require complex parameter tuning. Deep learning methods typically require large-scale training datasets and significant computational resources, limiting their application in lightweight systems.

[0038] Furthermore, most existing blur and intuitionistic blur enhancement methods rely on predefined complement functions or fixed parameter configurations. This design limits the flexibility of the enhancement process, making it difficult to adaptively adapt to images captured under different lighting conditions. Additionally, some methods require extra post-processing operations, such as histogram equalization, which increases system complexity.

[0039] Therefore, there is still an urgent need for a more adaptive and computationally efficient image enhancement framework that can effectively improve the quality of low-light images while preserving structural details and natural visual effects.

[0040] To address these shortcomings, this invention proposes solutions to several deficiencies in existing low-light image enhancement techniques, including limited adaptability to different lighting conditions, reliance on histogram-based post-processing operations, and excessively high computational demands of learning-based methods. To overcome these limitations, this invention proposes a novel image enhancement framework based on a generalized fuzzy complement function mechanism combined with adaptive parameter optimization. This method converts the input low-light image into a fuzzy representation, calculates complementary information using a flexible complement function formula, and can adapt to different brightness distributions. Subsequently, an intuitive fuzzy structure is used to represent the uncertainty of pixel grayscale. Simultaneously, an entropy-guided optimization strategy is used to automatically determine the optimal enhancement parameters, thereby improving the brightness distribution and structural sharpness of the processed image. By integrating these modules into a unified enhancement framework, this invention can improve the contrast of low-light images, better preserve structural details, and enhance visual quality. Compared with existing technologies, this method provides a more adaptive and computationally efficient solution that can be widely applied to various imaging systems operating under complex lighting conditions.

[0041] This embodiment provides an adaptive low-light image enhancement method, such as... Figure 1 As shown, the specific steps include the following: Step S100: Obtain the image to be enhanced and determine the brightness value of each pixel in the image to be enhanced; In this embodiment, the brightness value is a numerical representation reflecting the brightness of each pixel in the image. For grayscale images, the brightness value is the grayscale value of that pixel, typically ranging from 0 to 255, where 0 corresponds to pure black and 255 corresponds to pure white. For color images, the brightness value can be the brightness component extracted after color space conversion, such as the brightness component extracted after converting from RGB color space to YCbCr, HSV, or Lab color space. The appropriate color space is selected for brightness component extraction based on the actual application scenario. Before performing subsequent fuzzy mathematical operations, since fuzzy operations are usually performed in the [0,1] interval, the obtained brightness values ​​are normalized to map their value range to this interval. Normalization can be achieved using a linear stretching method, that is, using the minimum and maximum brightness values ​​in the image to proportionally map the original brightness values ​​to the target interval. For example, in a surveillance image of a nighttime street scene in a low-light environment, the pixel brightness values ​​are generally low and the dynamic range is narrow. After normalization, the brightness values ​​of each pixel are stretched to the [0,1] range, so that the blurring process can better utilize the numerical space to express information.

[0042] In one implementation, the step of acquiring the image to be enhanced and determining the brightness value of each pixel in the image to be enhanced specifically includes the following steps: Step S110: Obtain a low-light image as the image to be enhanced; Step S120: Obtain the grayscale value of each pixel in the image to be enhanced, and normalize the grayscale value to the target value range by linear stretching to obtain the normalized grayscale value of each pixel as the brightness value.

[0043] In this embodiment, a low-light image refers to an image captured under insufficient lighting conditions, resulting in an overall low brightness level and difficulty in discerning details in dark areas. Such images are widely used in scenarios such as nighttime surveillance, underground space photography, and indoor natural light photography. The main problems they face include a narrow dynamic range of pixel values, compressed textures in dark areas, and relatively strong noise signals.

[0044] Grayscale values ​​are the intensity values ​​of each pixel in a grayscale image. Grayscale images contain only brightness information and not color information; each pixel is represented by a single grayscale value. For color input images, the color image can be converted to grayscale or the brightness channel can be extracted before using the grayscale values ​​as brightness values ​​as input. In this embodiment, the low-light image is represented by a grayscale matrix. It means that, among them For position The grayscale value of the pixel. and These represent the pixel dimensions of the low-light image.

[0045] Normalization is a data processing operation that scales numerical values ​​to a specific range. In this embodiment, normalization uses a linear stretching method. Specifically, the minimum and maximum gray values ​​of each pixel in the image to be enhanced are calculated, and then the gray value of each pixel is mapped to the target numerical range according to a linear relationship. In this embodiment, the target numerical range is set to the [0,1] range. The expression for the normalized gray value is:

[0046] in, This represents the normalized grayscale value. This represents the minimum gray value in a low-light image. This represents the maximum grayscale value in a low-light image.

[0047] After normalization, the relative brightness relationships between pixels are preserved, providing a unified numerical basis for using fuzzy mathematics tools.

[0048] Figure 2 The image normalization process in this embodiment is illustrated. Specifically, the image to be enhanced first enters the normalization process stage, where the grayscale values ​​of each pixel in the image to be enhanced are extracted as brightness values, and the grayscale values ​​are mapped to the target value range through linear stretching to obtain the normalized brightness values.

[0049] Step S200: Process the brightness value using a parameterized processing function to generate complementary information for each pixel; In this embodiment, a parameterized processing function refers to a mathematical function with one or more adjustable parameters, whose output characteristics can be changed according to different parameter values. For a given input value, the parameterized processing function can generate an output value that has a specific correspondence with that input value; this output value is referred to as complementary information in this embodiment. Complementary information provides another numerical perspective starting from the original brightness value, reflecting the inverse or auxiliary characteristics of the brightness value in a fuzzy sense. When the parameters change, the complementary information generated by the same brightness value will also change accordingly, providing a basis and degree of freedom for optimizing the image enhancement effect by adjusting the parameters.

[0050] When the parameterization function employs a generalized fuzzy complement function, it can adapt to the characteristics of different images. As a type of complement function with adjustable parameters, the generalized fuzzy complement function changes its shape by adjusting the parameters, thus adapting to differences in image brightness distribution under different lighting conditions. Its function is to map the input value to an output value with complementary features; the nonlinearity of this mapping is controlled by the adjustable parameters. When the parameter values ​​differ, the complementary information corresponding to the same brightness value changes accordingly. Therefore, the intuitively blurred image constructed based on the complementary information also exhibits different characteristics, providing a foundation for the adaptive optimization of the entire enhancement method.

[0051] In one implementation, the parameterization function is a generalized fuzzy complement function, which is used for input values ​​whose domain is within the target numerical range. The generalized fuzzy complement function Defined as:

[0052] in, , This is an adjustable parameter used to control the shape of the fuzzy complement transform.

[0053] In this embodiment, a generalized fuzzy complement function is used as the specific mathematical form of the parameterized processing function. A generalized fuzzy complement function is a type of parameterized function whose domain is in the interval [0,1] and whose range is also in the interval [0,1]. When the input value is 0, the output value is 1, and when the input value is 1, the output value is 0. The function is monotonically decreasing in its domain.

[0054] Generalized fuzzy complement function In the definition, the independent variable For input values ​​whose domain is in [0,1], and These are two independently adjustable parameters. It mainly controls the overall shape and nonlinearity of the complementary function curve, when When the value is large, the curve changes more steeply in the middle region. When the value is small, the curve tends to flatten out. Parameter The main function controlled is the bias characteristic of the complementary function curve. Different values ​​of can cause the curve to produce varying degrees of response intensity changes in the low-input or high-input regions. The combined effect of the two parameters provides a rich space for morphological adjustment of the complementary function.

[0055] when and At this point, the generalized fuzzy complement function degenerates into a standard Sugeno-class fuzzy complement function. This is one of the most basic forms of complement operation in fuzzy set theory. When and When other specific values ​​are taken, the function can also be reduced to other known fuzzy complement function forms.

[0056] because and As an adjustable parameter, the nonlinear mapping characteristics of the generalized fuzzy complement function can be adjusted according to the actual brightness distribution of the input image. For low-light images with generally low brightness and narrow dynamic range, appropriate selection is recommended. and The value of can amplify the brightness difference in the complementary information generation process. For images with relatively uniform brightness distribution, alternative parameter combinations can produce a moderate complementary response, avoiding over-enhancement. This adaptive parameter adjustment mechanism makes the enhancement method widely applicable.

[0057] More specifically, let pixel grayscale The corresponding membership degree is then defined as follows:

[0058] in This is an adjustable parameter used to control the form of the complement transformation. It can be understood that this formula provides a flexible representation of the behavior of fuzzy complement functions. Under specific parameter settings, this function can degenerate into the classic complement function model commonly used in fuzzy systems.

[0059] In this embodiment, a generalized fuzzy complement function is used, but other mathematical forms with similar complement properties can also be employed. For example, different parameterized complement functions or other monotonic transformation functions can be used to generate pixel complementary information while retaining the idea of ​​adaptive illumination adjustment.

[0060] In one implementation, the complementary information includes membership degree, non-membership degree, and hesitation degree. The step of processing the brightness value using a parameterized processing function to generate complementary information for each pixel specifically includes the following steps: Step S210: Using the brightness value as input to the generalized fuzzy complement function, calculate the membership degree of each pixel, expressed as:

[0061] in, Indicates the location of the image to be enhanced. The brightness value of that pixel. Indicates the location of the image to be enhanced. The membership degree of a pixel; Step S220: Using the membership degree as input to the generalized fuzzy complement function, calculate the non-membership degree of each pixel, expressed as:

[0062] in, Indicates the location of the image to be enhanced. The non-membership degree of a pixel; Step S230: Based on the membership degree and the non-membership degree, calculate the hesitation degree of each pixel, expressed as:

[0063] in, Indicates the location of the image to be enhanced. Hesitation level at each pixel.

[0064] In this embodiment, within the framework of the generalized fuzzy complement function, membership degree describes the degree to which a pixel belongs to the fuzzy concept of "brightness" in an image. The membership degree of each pixel is calculated using the generalized fuzzy complement function. The membership degree is obtained by considering the normalized image brightness value as the degree of "darkness," and the generalized fuzzy complement function provides the complement of "darkness"—the degree of "non-darkness," which is equivalent to the membership degree of "brightness." Pixels with larger normalized brightness values ​​generally have larger membership degrees, reflecting a higher degree of "brightness" for that pixel. When the membership degree is close to 0, the pixel is very dark, and its membership degree is close to 0; when When the value is close to 1, the pixel is very bright, and the membership degree is... The membership value is close to 1; for pixels with intermediate brightness, the membership value is determined by the parameter. Control its nonlinear transition characteristics.

[0065] Non-membership describes the degree to which a pixel does not belong to the fuzzy concept of "brightness". In this embodiment, membership is used as input to the generalized fuzzy complement function to calculate non-membership. The calculation logic of non-membership shows that pixels with higher membership usually have lower non-membership, and pixels with lower membership usually have higher non-membership. This is consistent with the basic constraint in intuitionistic fuzzy sets that the sum of membership and non-membership does not exceed 1. Physically, if a pixel is highly considered to be "bright", then its degree of non-belonging to "brightness" is naturally low.

[0066] Hesitation describes the degree of uncertainty in assigning a pixel's brightness value, reflecting the hesitation space in determining whether a pixel is "bright" or "not bright." According to the definition above, the sum of membership, non-membership, and hesitation is always equal to 1. When the sum of membership and non-membership is less than 1, the remaining part is hesitation, which represents the magnitude of uncertainty in pixel brightness information. In low-light images, the true brightness of pixels in dark areas is difficult to determine due to noise interference and insufficient illumination, thus often exhibiting greater hesitation. Conversely, pixels in bright areas, due to their higher signal-to-noise ratio, have relatively clear brightness assignments and typically exhibit lower hesitation. Unifying these three fuzzy features as complementary information provides a complete feature representation basis for constructing intuitive fuzzy images.

[0067] By employing three dimensions—membership, non-membership, and hesitation—this approach more comprehensively characterizes the brightness features and uncertainties of each pixel in a low-light image. Compared to traditional blurring image enhancement methods that only use a single membership degree, it adds non-membership (a negative indicator) and hesitation (an undetermined indicator), resulting in a more refined description of pixel states. This multi-dimensional representation can better distinguish between genuine brightness changes and noise-induced uncertainties in low-light images under varying lighting conditions and noise levels, providing a more reliable processing foundation for contrast enhancement steps.

[0068] This embodiment constructs an intuitive fuzzy image using membership, non-membership, and hesitation. In alternative embodiments, the fuzzy representation can be extended to other fuzzy frameworks, such as interval-valued fuzzy structures or type-two fuzzy representations, as long as they can effectively model the uncertainty of pixel grayscale.

[0069] Step S300: Based on the complementary information and the brightness value, construct an intuitive blur image, wherein the intuitive blur image is used to express the uncertainty of pixel values; In this embodiment, the intuitive blurred image is an extended representation of blurred images. Unlike describing pixel value characteristics using only a single membership value, the intuitive blurred image is constructed by combining complementary information and the original brightness value, simultaneously expressing both the positive attribution and uncertainty features of pixel values. The uncertainty expressed by the intuitive blurred image includes at least two dimensions: first, the ambiguity caused by the difficulty in determining the true brightness value of pixels due to low light; and second, the hesitation caused by the unclear attribution of pixel values ​​due to noise or other interference factors. This representation allows for a richer description and representation of texture and detail information that was originally compressed or hidden in low-light images.

[0070] In low-light conditions, the number of photons received by the image sensor decreases, leading to a lower signal-to-noise ratio. Pixel values ​​not only reflect the brightness information of the scene but also contain a significant proportion of random noise. Traditional single membership representations struggle to distinguish between brightness information and noise interference. Intuitive blurring, however, uses multi-dimensional feature representations to describe brightness information and uncertainty separately. This allows enhancement processing to more effectively adjust pixel value distribution, improving brightness in dark areas while suppressing excessive noise amplification. Therefore, intuitive blurring can effectively model uncertainties and illumination variations in low-light images.

[0071] In one implementation, constructing the intuitive blurred image based on the complementary information and the brightness value specifically includes the following steps: Step S310: The sum of the membership degree and the hesitation degree of each pixel in the image to be enhanced is used as the pixel value of the intuitive blur image to construct the intuitive blur image, wherein the pixel value of the intuitive blur image is represented as:

[0072] in, Indicates the location of the intuitive blurred image. The pixel value of the pixel at that location.

[0073] In this embodiment, the membership degree and hesitation degree of each pixel in the image to be enhanced are added together, and this sum is used as the pixel value of the corresponding pixel in the intuitionistic blurred image. In the constructed intuitionistic blurred image, each pixel value... This combines information from two sources. It indicates the degree to which a pixel is "bright". This reflects the degree of hesitation in pixel attribution. Incorporating hesitation into pixel value calculations allows pixels with initially ambiguous attributes to receive some brightness compensation in intuitively blurred images. For dark area pixels with high hesitation, their pixel values ​​are adjusted accordingly. This results in a relative improvement, providing a more reasonable starting point for subsequent contrast enhancement. For bright pixels with high membership and low hesitation, their pixel values ​​are mainly determined by their own brightness characteristics and will not change significantly due to the introduction of hesitation.

[0074] The properties of intuitively blurred images are influenced by the adjustable parameters in the generalized fuzzy complement function. The parameter values ​​directly determine the calculation of membership and non-membership, thus affecting the degree of hesitation and the pixel values ​​of the final intuitively blurred image. Therefore, under different parameter values, the same input image will generate intuitively blurred images with different feature representations.

[0075] Figure 2This demonstrates the process of constructing an intuitive blurred image. Specifically, it calculates the membership degree for each pixel using a generalized fuzzy complement function. Non-membership degree and hesitation The sum of membership degree and hesitation degree is used as the pixel value. This allows for the construction of intuitive fuzzy images that can express the uncertainty of pixel values.

[0076] Step S400: Based on the global statistical features of the image to be enhanced, perform contrast enhancement on the intuitively blurred image to obtain a preliminary enhanced image; In this embodiment, global statistical features refer to the feature quantities obtained by statistically analyzing the brightness values ​​of all pixels in the entire intuitively blurred image or a specific region therein. These feature quantities reflect the overall brightness level of the image, the concentration of brightness distribution, and the dispersion between pixel values. Contrast enhancement refers to the operation of improving the visual effect of an image by adjusting the distribution of pixel brightness values. Contrast enhancement is calculated using global statistical features as input parameters, without relying on post-processing methods such as histogram equalization or limiting contrast adaptive histogram equalization, thus avoiding the introduced over-enhancement effect and the problem of amplifying noise in dark areas.

[0077] Contrast enhancement using global statistical features automatically adjusts the enhancement level based on the image's brightness distribution characteristics. For low-light images with generally low brightness and a small standard deviation, the statistical features reflect the degree of concentration in the dark areas, applying an appropriate amount of enhancement accordingly. For images with relatively uniform brightness distribution, the statistical features reflect a smaller enhancement requirement, avoiding over-enhancement that could lead to highlight clipping. Regardless of the brightness distribution type of the input low-light image, contrast adjustment can be adaptively performed, improving visibility while maintaining the image's natural appearance.

[0078] Specifically, statistical contrast adjustment methods based on image mean and variance can be used, or other contrast enhancement strategies, such as local contrast adjustment or adaptive grayscale scaling, can be used to improve brightness distribution.

[0079] In one implementation, the step of performing contrast enhancement on the intuitively blurred image based on the global statistical features of the image to be enhanced to obtain a preliminary enhanced image specifically includes the following steps: Step S410: Calculate the global average gray level and gray level standard deviation of the intuitive blurred image; Step S420: Calculate the target contrast based on the global average gray level and the gray level standard deviation; Step S430: Using the target contrast, perform a contrast enhancement transformation on the intuitively blurred image to obtain the preliminary enhanced image.

[0080] In this embodiment, the global average grayscale is... Defined as the arithmetic mean of all pixel values ​​in an intuitively blurred image, it reflects the overall brightness level of the image and measures the grayscale standard deviation. Defined as the square root of the squared mean of the differences between each pixel value and the global average gray level, it reflects the dispersion of the image brightness distribution and contrast information.

[0081] Based on the two statistical characteristics mentioned above, target contrast The calculation expression is:

[0082] Wherein, the coefficient 1.5 is the contrast expansion factor. The operation ensures that the target contrast does not exceed the upper limit of 1 to prevent excessive contrast stretching that could lead to image distortion. When the standard deviation of the intuitively blurred image itself is small, i.e., the image contrast is low, the target contrast is set to 1.5 times the standard deviation to achieve moderate enhancement. When the standard deviation is large, the target contrast is limited to within 1 to avoid over-enhancement.

[0083] After obtaining the target contrast ratio, the calculation expression for the contrast enhancement transformation is:

[0084] in, This represents an intuitively blurred image. Firstly, through... The image is standardized to eliminate differences in both mean and standard deviation, ensuring that pixel values ​​are distributed in a standard space with a mean of 0 and a standard deviation of 1; then the target contrast is... The transformation is applied to the standardized pixel values, then the numerical range is adjusted by a coefficient of 2, and the values ​​are shifted by adding 1. After this transformation, the pixel value distribution range of the initially enhanced image is adjusted and expanded, the dark areas are moderately enhanced while the structural proportions of the bright areas are maintained.

[0085] The contrast enhancement strategy in this embodiment differs from conventional histogram equalization methods. It does not require statistical analysis of the cumulative distribution function of the image's gray levels, nor does it perform pixel value redistribution based on distribution mapping. Instead, it directly performs linear contrast adjustment based on the image's mean and standard deviation—two global statistical features. The advantage of this approach is that it maintains the relative proportions between the original image's pixel values, avoiding the detail loss or noise amplification problems caused by distribution mapping in histogram equalization.

[0086] Figure 2 The process of global contrast enhancement is demonstrated. Specifically, it is based on the global average grayscale of an intuitively blurred image. and grayscale standard deviation Calculate target contrast The image with blurred intuition is then subjected to a contrast enhancement transform to obtain a preliminary enhanced image. .

[0087] Step S500: With maximizing the information content or perceptual quality of the enhanced image as the optimization objective, the adjustable parameters in the parameterized processing function are iteratively optimized to obtain the optimal parameter enhanced image; In this embodiment, the optimization objective refers to the evaluation criteria used to measure the quality of the current enhancement effect during the parameter search process. In this embodiment, the optimization objective is to maximize the information content or perceptual quality of the enhanced image. Information content can be measured using image entropy, a statistical measure that reflects the richness of information in an image. Specifically, when the gray-level distribution of the enhanced image is more uniform and contains more effective gray levels, its entropy value is usually larger, indicating that the image contains richer detail information. When the enhanced image has large areas of overly bright or dark areas, its entropy value is usually smaller, indicating that some detail information has been lost. Perceptual quality can be measured using no-reference image quality evaluation metrics, such as the Naturalness Image Quality Evaluation Metric (NIQE) or the No-Reference Image Spatial Domain Quality Evaluation Metric (BRISQUE). These metrics are based on the statistical prior features of natural images and can quantify the degree to which the enhanced image deviates from natural statistical laws. Iterative optimization refers to a cyclical process of repeatedly adjusting parameters, evaluating the effect, and then readjusting the parameters. Specifically, in each iteration, an enhanced image is generated using the current parameters, and the value of the optimization target is calculated. Then, the direction and step size of parameter adjustment are guided by the changing trend of the target value. After multiple iterations, the parameters gradually converge to the value that maximizes the optimization target. Since the adjustable parameters of the parameterized processing function determine the way complementary information is generated, affecting the feature expression of the intuitively blurred image and the final enhancement effect, iterative optimization of these parameters allows the enhancement method to automatically adapt to different lighting conditions.

[0088] Specifically, entropy-based optimization strategies and flexible search methods can be used to determine enhancement parameters. In other embodiments, other optimization techniques can be used, such as searching for optimal parameters with or without reference quality metrics, including heuristic search, gradient-based optimization, evolutionary algorithms, or other numerical optimization strategies, all with the goal of maximizing image information content or perceived quality.

[0089] In one implementation, the step of iteratively optimizing the adjustable parameters in the parameterized processing function to obtain the optimal parameter-enhanced image, with the goal of maximizing the information content or perceptual quality of the enhanced image, specifically includes the following steps: Step S510: Calculate the Shannon entropy of the enhanced image; Step S520: Through a refined search strategy, iteratively narrow the range of values ​​for the adjustable parameters, search for the parameter value that maximizes the Shannon entropy as the optimal parameter, and obtain the optimal parameter enhanced image of the image to be enhanced.

[0090] In this embodiment, Shannon entropy is a classic metric in information theory used to measure information content. In the field of image processing, Shannon entropy reflects the richness of information contained in an image by calculating the probability distribution of image gray levels. The Shannon entropy of the enhanced image... The calculation method is as follows:

[0091] in, Indicates the gray levels in the enhanced image The probability of occurrence of a grayscale value in an image, i.e., the probability of grayscale values ​​appearing in the image. The Shannon entropy is the ratio of the number of pixels in a single grayscale region to the total number of pixels in an image. The more uniform the grayscale distribution and the closer the probabilities of each grayscale level, the greater the Shannon entropy, indicating richer information and better detail retention in the image. Conversely, when a large area of ​​a single grayscale region appears in the image, the Shannon entropy decreases, indicating some information loss.

[0092] With maximizing Shannon entropy as the optimization objective, parameter optimization is performed on the adjustable parameters of the generalized fuzzy complement function. The optimization process is carried out on the above. During the optimization process, parameters... It can be set to a fixed value or used as another variable to be optimized, depending on the actual application requirements. The parameter search adopts a fine-grained search strategy, first setting the parameters. The initial search range and step size are determined. Then, several candidate parameter values ​​are selected within the current range. A complete image enhancement process is performed using each candidate parameter value, and the Shannon entropy of the enhanced image after each enhancement is calculated as the optimization evaluation value corresponding to that parameter. Subsequently, the Shannon entropy values ​​under different parameter values ​​are compared to determine the current optimal parameter. Then, the search range is narrowed and the step size is reduced around the current optimal parameter, entering the next round of iterative search. Through multiple iterations, the range of parameter values ​​is gradually narrowed until the search range is reduced to a preset convergence threshold or the change in Shannon entropy is less than a preset tolerance. The iteration terminates, and the parameter value that maximizes the Shannon entropy is taken as the final optimal parameter. ,Right now:

[0093] This optimization strategy requires no manual intervention in parameter selection and can automatically determine the most suitable parameter values ​​for different types of low-light images. For extremely low-light images with strong noise or a very narrow dynamic range, the optimization process tends to select parameter values ​​that make the gray-level distribution of the enhanced image more balanced to improve information content. For images with relatively uniform illumination, the optimization process will select more conservative parameter values ​​to avoid over-enhancement. After obtaining the optimal parameters through the parameter optimization module, the optimal parameters are fed back to the generalized fuzzy complement function calculation stage to re-execute the intuitive fuzzy image construction and contrast enhancement process, and finally generate and output the enhanced image with the optimal parameters.

[0094] Figure 3 Adjustable parameters are shown The optimization process. When optimizing parameters, the parameters are first set. Initial search range, maximum number of iterations and convergence tolerance threshold Before iterative optimization begins, the initial enhanced image is acquired and its Shannon entropy is calculated as the globally optimal entropy value. The initial record.

[0095] Once the iteration loop begins, in each iteration, a set of candidate parameter values ​​is selected within the currently defined parameter search range. For each candidate parameter Perform the complete enhancement process to generate the corresponding enhanced image. Subsequently, the information content of the enhanced image was calculated using the Shannon entropy calculation function, and its Shannon entropy value was obtained. The calculation formula is: At the same time, this entropy value is assigned to the evaluation variable for the current round. ,Right now .

[0096] Then, calculate the current entropy value. With the global optimal entropy value The absolute value of the difference between ,Right now This is used to quantify the improvement brought about by the current candidate parameter. The absolute value of this difference is then determined. Is it less than the preset convergence tolerance threshold? If yes, it indicates that the improvement has converged, and the iteration terminates. If not, then maximize the improvement based on the search results. Find the optimal parameters for this round to achieve the goal. The parameters are updated accordingly, and then the search range is narrowed down around the optimal point. The process continues until the maximum number of iterations is reached. Until the convergence condition is met.

[0097] After the iteration is complete, output the finally found globally optimal parameters. And use these parameters to generate the final optimal parameter-enhanced image.

[0098] Figure 2 The process of parameter optimization is demonstrated. Specifically, Shannon entropy is calculated as a measure of information content, and the adjustable parameters of the generalized fuzzy complement function are iteratively adjusted through a refined search strategy to search for the parameter values ​​that maximize Shannon entropy. To obtain the optimal parameters for image enhancement .

[0099] Step S600: Convert the enhanced image with the optimal parameters from the fuzzy domain back to the spatial domain to obtain the final enhanced image.

[0100] In this embodiment, converting the enhanced image with optimal parameters from the fuzzy domain back to the spatial domain refers to the operation of restoring the image data obtained after the aforementioned steps, which is in the fuzzy representation space, to the standard image grayscale value range through numerical mapping. In the fuzzy domain, image pixel values ​​are typically in the [0,1] interval or other normalized ranges, while image grayscale values ​​in the standard spatial domain are typically in the [0,255] interval, for direct use by display devices or subsequent image analysis tasks. The conversion process can be achieved through inverse normalization. Specifically, using the minimum and maximum brightness values ​​of the image to be enhanced, the pixel values ​​in the fuzzy domain are mapped back to the original grayscale value range proportionally. Converting back to the standard spatial domain ensures that the final output enhanced image can be directly read and processed by standard image processing workflows, guaranteeing compatibility. The final expression for the enhanced image is:

[0101] in, The optimized blurred grayscale value, For the final output image at position The pixel value.

[0102] Figure 2 The process of image reconstruction output is demonstrated. Specifically, after determining the optimal parameters and generating the enhanced image with the optimal parameters through parameter optimization, the enhanced image with the optimal parameters, which is in the fuzzy domain, is converted back to the original grayscale range through inverse normalization, completing the mapping from the fuzzy domain to the spatial domain, and outputting the final enhanced image. .

[0103] To illustrate the technical effects of this embodiment, Figure 4 The example input and output images of the image enhancement method framework proposed in this embodiment are shown, demonstrating the improvement effect of this embodiment on the brightness and visibility of low-light images. Figure 4In the diagram, (a), (b), (e), and (f) are the overall and partial graphs of the two input examples, respectively. Figure 4 In the diagram, (c), (d), (g), and (h) represent the overall and partial images output after executing the image enhancement method of this embodiment, respectively, for the two input examples. It can be seen that the image enhancement method of this embodiment can obtain a better enhanced image for low-light image processing.

[0104] To further demonstrate the technical effectiveness of this embodiment, quantitative performance comparisons were performed using a general benchmark dataset. Tables 1 and 2 present the quantitative comparison results between the enhancement method of this embodiment and some existing low-light image enhancement models based on blur complement functions on the LOLv1 and LOLv2-real datasets, respectively. Figure 5 and Figure 6 The image effects of the enhancement method described in this embodiment and the output images of some existing low-light image enhancement models based on blur complement functions are shown in the examples. Figure 5 and Figure 6 The upper-middle image is the overall view, and the lower-middle image is a partial view.

[0105] Table 1

[0106] Table 2

[0107] In Tables 1 and 2, LOLv1 and LOLv2-real are two low-light image datasets. IFGH (Intuitionistic Fuzzy Generator with Histogram Equalization), IVIFGH (Interval-Valued Intuitionistic Fuzzy Generator with Histogram Equalization), IVIFGC (Interval-Valued Intuitionistic Fuzzy Generator with Contrast-Limited Adaptive Histogram Equalization), SIFGC (Selvam's Intuitionistic Fuzzy Generator with Contrast-Limited Adaptive Histogram Equalization), CIFGH (Chinnappan's Intuitionistic Fuzzy Generator with Histogram Equalization), IFGHER (Intuitionistic Fuzzy Generator with Histogram Equalization) Retinex (based on histogram equalization) and Retinex's intuitive blur generator are existing low-light image enhancement models based on fuzzy complement functions. In Tables 1 and 2, the up arrows indicate that the enhancement method of this embodiment has a larger value in this metric compared to other existing methods, while the down arrows indicate that the enhancement method of this embodiment has a smaller value in this metric compared to other existing methods.

[0108] The results in Tables 1 and 2 demonstrate that, compared to various existing enhancement methods based on fuzzy complement functions, the image enhancement method of this embodiment achieves superior performance across multiple image quality evaluation metrics. Specifically, this method achieves higher PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index Measure) values, while lower RMSE (Root Mean Square Error) and MAE (Mean Absolute Error), indicating better structure preservation and smaller reconstruction errors. Furthermore, lower BRISQUE (Blind / Referenceless Image Spatial Quality Evaluator) and NIQE (Naturalness Image Quality Evaluator) scores indicate that the enhanced image possesses superior subjective perceived quality and a more natural appearance. These results confirm that the generalized fuzzy complement function framework proposed in this embodiment can effectively enhance low-light images while preserving important structural details.

[0109] In summary, the low-light image enhancement method proposed in this embodiment processes low-light images using a generalized fuzzy complement function to construct an intuitive fuzzy image representation that includes membership, non-membership, and hesitation. This effectively models the uncertainty of pixel grayscale and illumination variations in low-light images, providing richer image feature information for image enhancement. Through a contrast enhancement strategy based on global image statistical features, brightness distribution improvement and contrast enhancement are achieved without relying on post-processing operations such as histogram equalization, avoiding over-enhancement and noise amplification. A parameter optimization strategy aiming to maximize the information content or perceptual quality of the enhanced image employs a refined search to automatically determine the optimal parameters, eliminating the need for manual parameter tuning and enhancing the method's adaptability to different lighting conditions. The entire enhancement method uses a unified pipeline design, sequentially executing image normalization, fuzzy complement function calculation, intuitive fuzzy image construction, contrast enhancement, parameter optimization, and image reconstruction. These modules work collaboratively to effectively preserve structural details and natural visual perception while improving the contrast and visibility of low-light images. Meanwhile, this method does not require the large-scale training dataset and extensive computing resources needed for deep learning, has high computational efficiency, can be deployed in real-time imaging systems, and is suitable for various imaging application scenarios under complex lighting conditions, such as monitoring systems, medical imaging, autonomous navigation, and multimedia processing.

[0110] The generalized fuzzy complement function framework proposed in this invention provides a flexible mathematical foundation for modeling image grayscale uncertainty. Although this application primarily focuses on low-light image enhancement, this theoretical framework can also be extended to other image processing tasks, such as image fusion, image segmentation, edge detection, and image restoration. The ability of the generalized complement function formula to adaptively represent brightness changes makes it applicable to a wide range of visual information processing problems. The proposed intuitive fuzzy representation can serve as a general framework for processing fuzzy or uncertain image information. Compared with traditional fuzzy methods, this model provides a richer description of pixel features by fusing membership, non-membership, and hesitation components. This approach can be extended to other fuzzy-based decision systems, computer vision tasks, and intelligent image analysis applications. The entropy-guided parameter optimization strategy described in this invention can be further extended to other adaptive image enhancement systems. The principle of maximizing information content can be used to optimize the parameters of various image processing algorithms, thereby improving the quality of enhanced images without the need for manual parameter selection. Furthermore, the proposed enhancement method framework can be integrated with modern intelligent imaging systems. For example, it can be embedded in image acquisition devices such as digital cameras, surveillance equipment, drones, and mobile terminals to provide real-time low-light image enhancement functionality. The design concept of combining fuzzy mathematical modeling and adaptive optimization proposed in this invention can inspire the development of more intelligent image processing technologies that balance computational efficiency and high visual quality. This approach demonstrates the broader applicability and potential development direction of this invention.

[0111] like Figure 7 As shown in the figure, this embodiment of the invention provides an adaptive low-light image enhancement system, which includes: a data acquisition module 10, a complementary information calculation module 20, an intuitive fuzzy image construction module 30, an image enhancement module 40, a parameter optimization module 50, and a fuzzy domain conversion module 60.

[0112] Specifically, the data acquisition module 10 is used to acquire the image to be enhanced and determine the brightness value of each pixel in the image to be enhanced; the complementary information calculation module 20 is used to process the brightness value through a parameterized processing function to generate complementary information for each pixel; the intuitive fuzzy image construction module 30 is used to construct an intuitive fuzzy image based on the complementary information and the brightness value, wherein the intuitive fuzzy image is used to express the uncertainty of pixel values; the image enhancement module 40 is used to perform contrast enhancement on the intuitive fuzzy image based on the global statistical features of the image to be enhanced to obtain a preliminary enhanced image; the parameter optimization module 50 is used to iteratively optimize the adjustable parameters in the parameterized processing function with the optimization goal of maximizing the information content or perceptual quality of the enhanced image to obtain the optimal parameter enhanced image; the fuzzy domain conversion module 60 is used to convert the optimal parameter enhanced image from the fuzzy domain back to the spatial domain to obtain the final enhanced image.

[0113] Based on the above embodiments, the present invention also provides a terminal device, the principle block diagram of which can be as follows: Figure 8 As shown, the terminal device includes a processor, memory, network interface, display screen, and temperature sensor connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements an adaptive low-light image enhancement method. The display screen can be a liquid crystal display (LCD) or an e-ink display. The temperature sensor is pre-installed inside the terminal device to detect the operating temperature of the internal components.

[0114] Those skilled in the art will understand that Figure 8 The schematic diagram shown is only a partial structural diagram related to the present invention and does not constitute a limitation on the terminal device to which the present invention is applied. The specific terminal device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0115] In one embodiment, a terminal device is provided, including a memory and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors, the one or more programs including instructions for performing operations as described in the embodiments of the methods above.

[0116] Those skilled in the art will understand that all or part of the processes in the methods of 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. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0117] 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.

[0118] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An adaptive low-light image enhancement method, characterized in that, The method includes: Acquire the image to be enhanced and determine the brightness value of each pixel in the image to be enhanced; The brightness value is processed by a parameterized processing function to generate complementary information for each pixel; Based on the complementary information and the brightness value, an intuitive blurred image is constructed, wherein the intuitive blurred image is used to express the uncertainty of pixel values; Based on the global statistical features of the image to be enhanced, contrast enhancement is performed on the intuitively blurred image to obtain a preliminary enhanced image; With the goal of maximizing the information content or perceptual quality of the enhanced image, the adjustable parameters in the parameterized processing function are iteratively optimized to obtain the optimal parameter-enhanced image. The image enhanced by the optimal parameters is transformed from the fuzzy domain back to the spatial domain to obtain the final enhanced image; The parameterization function is a generalized fuzzy complement function, for input values ​​whose domain is within the target numerical range. The generalized fuzzy complement function Defined as: in, , It is an adjustable parameter used to control the shape of the fuzzy complement transform; The complementary information includes membership degree, non-membership degree, and hesitation degree. The step of processing the brightness values ​​using a parameterized processing function to generate complementary information for each pixel includes: Using the brightness value as input to the generalized fuzzy complement function, the membership degree of each pixel is calculated, expressed as: in, Indicates the location of the image to be enhanced. The brightness value of that pixel. Indicates the location of the image to be enhanced. The membership degree of a pixel; Using the membership degree as input to the generalized fuzzy complement function, the non-membership degree of each pixel is calculated, expressed as: in, Indicates the location of the image to be enhanced. The non-membership degree of a pixel; Based on the membership degree and the non-membership degree, the hesitation degree of each pixel is calculated and expressed as: in, Indicates the location of the image to be enhanced. Hesitation at each pixel; The construction of an intuitive blurred image based on the complementary information and the brightness value includes: The sum of the membership degree and the hesitation degree of each pixel in the image to be enhanced is used as the pixel value of the intuitive blur image to construct the intuitive blur image, wherein the pixel value of the intuitive blur image is represented as: in, Indicates the location of the intuitive blurred image. The pixel value of the pixel at that location.

2. The adaptive low-light image enhancement method according to claim 1, characterized in that, The step of acquiring the image to be enhanced and determining the brightness value of each pixel in the image to be enhanced includes: Acquire a low-light image as the image to be enhanced; The grayscale values ​​of each pixel in the image to be enhanced are obtained, and the grayscale values ​​are normalized to the target value range by linear stretching to obtain the normalized grayscale values ​​of each pixel as the brightness values.

3. The adaptive low-light image enhancement method according to claim 1, characterized in that, The process of performing contrast enhancement on the intuitively blurred image based on the global statistical features of the image to be enhanced, to obtain a preliminary enhanced image, includes: Calculate the global average gray level and gray level standard deviation of the intuitively blurred image; Calculate the target contrast based on the global average gray level and the gray level standard deviation; Using the target contrast, a contrast enhancement transformation is performed on the intuitively blurred image to obtain the preliminary enhanced image.

4. The adaptive low-light image enhancement method according to claim 1, characterized in that, The step of iteratively optimizing the adjustable parameters in the parameterized processing function to obtain the optimal parameter-enhanced image, with the goal of maximizing the information content or perceptual quality of the enhanced image, includes: Calculate the Shannon entropy of the enhanced image; By employing a refined search strategy, the range of values ​​for the adjustable parameters is iteratively narrowed, and the parameter value that maximizes the Shannon entropy is searched as the optimal parameter, thus obtaining the optimal parameter-enhanced image of the image to be enhanced.

5. An adaptive low-light image enhancement system, characterized in that, The system is applied to implement the adaptive low-light image enhancement method as described in any one of claims 1-4, the system comprising: The data acquisition module is used to acquire the image to be enhanced and determine the brightness value of each pixel in the image to be enhanced. The complementary information calculation module is used to process the brightness value through a parameterized processing function to generate complementary information for each pixel; An intuitive blurred image construction module is used to construct an intuitive blurred image based on the complementary information and the brightness value, wherein the intuitive blurred image is used to express the uncertainty of pixel values; An image enhancement module is used to perform contrast enhancement on the intuitively blurred image based on the global statistical features of the image to be enhanced, so as to obtain a preliminary enhanced image; The parameter optimization module is used to iteratively optimize the adjustable parameters in the parameterized processing function with the goal of maximizing the information content or perceptual quality of the enhanced image, so as to obtain the optimal parameter-enhanced image. The fuzzy domain conversion module is used to convert the enhanced image with the optimal parameters from the fuzzy domain back to the spatial domain to obtain the final enhanced image.

6. A terminal device, characterized in that, The terminal device includes a memory, a processor, and an adaptive low-light image enhancement program stored in the memory and executable on the processor. When the processor executes the adaptive low-light image enhancement program, it implements the steps of the adaptive low-light image enhancement method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an adaptive low-light image enhancement program, which, when executed by a processor, implements the steps of the adaptive low-light image enhancement method as described in any one of claims 1-4.

Images