Data processing system and method for realizing low-light image enhancement

Abstract

The invention discloses a data processing system and method for realizing low-light image enhancement, and relates to the technical field of image processing, and the method comprises the steps: evaluating the scene approximation degree between a scene and a scene type in scene type data, and obtaining an approximate scene type; acquiring shooting reference data corresponding to the approximate scene type from the low-light platform, shooting in the scene by using the adjusted low-light equipment, and fusing the low-light images in the low-light image set to obtain a fused low-light image; removing noise of the fused low-light-level image, performing image enhancement on the fused low-light-level image after noise removal, and performing dynamic range compression on the fused low-light-level image to obtain an enhanced low-light-level image; and obtaining an enhanced low-light image of the low-light equipment in the scene, evaluating the image quality of the enhanced low-light image, obtaining a target enhanced low-light image, and sending the target enhanced low-light image to a user of the low-light equipment through the low-light platform.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically a data processing system and method for enhancing low-light images. Background Technology

[0002] Low-light devices are imaging devices that can capture visible or near-infrared light under low-light conditions (such as weak ambient light like moonlight or starlight) and generate visible images. Their core principle is to amplify weak light signals using highly sensitive sensors (such as photomultiplier tubes and image intensifiers), transforming dimly lit scenes that are difficult for the human eye to perceive into clear images. However, images captured by low-light devices in low-light conditions have the following drawbacks, including but not limited to: 1. Significant image noise: Due to the low number of photons received by the sensor under low light, the quantum efficiency is low, resulting in a weak image signal and images filled with granular noise points, causing a "snow screen" effect in dark areas; 2. Lack of image depth, making it difficult to distinguish key targets from the background: Due to insufficient light during low-light image capture, the difference in reflected light between objects and the background is small, and the brightness distribution is concentrated in a narrow range. The sensor cannot effectively distinguish brightness levels under low light, resulting in an overall dark image; 3. Color distortion: Low light can cause the automatic white balance algorithm to misjudge the type of light source, resulting in low color saturation and object colors deviating from the true scene.

[0003] Therefore, after low-light images are captured by low-light devices, image enhancement is required to solve problems such as obvious image noise, lack of image depth, and color distortion in low-light images. However, traditional low-light image enhancement can easily amplify noise in the image and has weak adaptability to complex scenes. It not only causes the loss of image details in low-light images but may also lead to a decrease in image realism and interfere with normal observation. Summary of the Invention

[0004] The purpose of this invention is to provide a data processing system and method for enhancing low-light images, in order to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a data processing method for enhancing low-light images, the method comprising:

[0006] Step S100: Obtain scene climate data of the scene where the low-light device is located, obtain scene type data of the low-light platform where the low-light device is located, evaluate the scene similarity between the scene and the scene type in the scene type data, and obtain the approximate scene type;

[0007] Step S200: Obtain shooting reference data corresponding to similar scene types from the low-light platform, adjust the shooting parameters of the low-light device, use the adjusted low-light device to shoot in the scene, obtain a low-light image set, and fuse the low-light images in the low-light image set according to the different exposure times of the low-light images in the low-light image set to obtain a fused low-light image.

[0008] Step S300: Remove noise from the fused low-light image, enhance the fused low-light image after noise removal, and compress the dynamic range of the fused low-light image to obtain the enhanced low-light image;

[0009] Step S400: Acquire the enhanced low-light image of the low-light device in the scene, evaluate the image quality of the enhanced low-light image, obtain the target enhanced low-light image, and send the target enhanced low-light image to the user of the low-light device through the low-light platform.

[0010] Furthermore, step S100 includes:

[0011] Step S101: Obtain the scene where the target object is located when the low-light device is shooting, and obtain the scene climate data of the scene, including the values ​​of various climate parameters.

[0012] Step S102: Obtain scene type data from the low-light platform. The scene type data includes the range of various climate parameters corresponding to each preset scene type. Evaluate the scene similarity between the scene and each scene type. The specific evaluation process is as follows:

[0013] When the value of a certain climate parameter in a scene falls within the range corresponding to a certain climate parameter in a certain scene type, that climate parameter in the scene is marked, and the total number A′ of the marked climate parameters in the scene is obtained. sum ;

[0014] Calculate the scene approximation value B = A′ between the scene and a certain scene type. sum / A sum , where A sum This represents the total number of climate parameters.

[0015] Obtain the maximum value of the scene approximation value between the scene and each scene type, determine the scene approximation between the scene type corresponding to the maximum value of the scene approximation value and the scene type, and record the scene type as the approximate scene type of the scene.

[0016] Furthermore, step S200 includes:

[0017] Step S201: Obtain the approximate scene type of the scene, and obtain the shooting reference data of the low-light device under the approximate scene type from the low-light platform. The shooting reference data includes the reference values ​​corresponding to various shooting parameters of the low-light device.

[0018] Step S202: Based on the shooting reference data, adjust the shooting parameters of the low-light device. Specifically, adjust each shooting parameter in the low-light device to the reference value corresponding to each shooting parameter in the shooting reference data.

[0019] Using a low-light device with adjusted shooting parameters, and taking pictures of the target object in the scene according to different preset exposure times, a number of low-light images are obtained. The low-light images are then collected to obtain a low-light image set, in which the low-light images are all aligned low-light images.

[0020] Step S203: Obtain a specific low-light image from the low-light image set and construct a two-dimensional coordinate system to assign a coordinate to each pixel in the specific low-light image;

[0021] Obtain the pixel weight value W(x,y) of the pixel with coordinates (x,y) in a low-light image;

[0022] Obtain the pixel weight values ​​of each pixel in several low-light images, and obtain the radiation intensity E(x,y) of the pixel with coordinates (x,y) in a certain low-light image;

[0023] Several low-light images in the low-light image set are fused to obtain a fused low-light image. The specific image fusion process is as follows: calculate the brightness contribution value F(x,y) = W(x,y) × E(x,y) of the pixel with coordinates (x,y) in the fused low-light image.

[0024] Calculate the brightness value I of the pixel with coordinates (x, y) in the fused low-light image. hdr (x,y) is used to obtain the brightness values ​​of each pixel in the fused low-light image.

[0025] Furthermore, step S300 includes:

[0026] Step S301: Separate the noise from the fused low-light image and remove the noise from the fused low-light image;

[0027] Step S302: Perform image enhancement on the fused low-light image after noise removal. Specifically, for the pixel with coordinates (x, y) in the fused low-light image, the processing procedure is as follows: For the brightness value I in the fused low-light image... hdr Performing a logarithmic transformation on (x,y) yields the logarithmic value L(x,y) = log(I) of the pixel at coordinates (x,y) in the fused low-light image.hdr (x,y)+ε), where ε is a pre-defined local constant;

[0028] The logarithmic value L(x,y) is convolved using Gaussian kernels of k different scales to calculate the illumination components at k different scales. Specifically, at the k-th scale, the illumination component Lk is the illumination component of the pixel with coordinates (x,y) in the fused low-light image. k (x, y), where k is a preset value, k > 0;

[0029] Calculate the detail component R of the pixel with coordinates (x, y) in the fused low-light image at the k-th scale. k (x,y)=L(x,y)-L k (x,y);

[0030] The detail components of the pixel with coordinates (x,y) in the fused low-light image at k scales are obtained and weighted and fused to obtain the detail fusion value R(x,y). The detail fusion value R(x,y) is then subjected to an exponential transformation to obtain the target brightness I′(x,y)=exp(R(x,y)).

[0031] Step S303: Obtain the fused low-light image after image enhancement, and obtain the maximum value I′ of the target brightness of each pixel in the fused low-light image. max The enhanced low-light image is obtained by dynamically compressing the fused low-light image at coordinates (x,y). Specifically, the compressed target brightness I′(x,y) of the pixel at coordinates (x,y) in the enhanced low-light image is obtained after dynamic range compression of the target brightness I′(x,y). LDR (x,y), I′ LDR (x,y)∈[0,1];

[0032] The noise separation in the above steps of the fused low-light image is to suppress noise in the low-light region and avoid noise in subsequent enhancement and amplification. Furthermore, image processing can effectively enhance the image details in the fused low-light image, making it suitable for image enhancement under complex lighting conditions. Finally, dynamic range compression of the fused low-light image allows the enhanced low-light image to be adapted to the display device, resulting in better display effects.

[0033] Furthermore, step S400 includes:

[0034] Step S401: Acquire several enhanced low-light images of the scene captured within the current period, and evaluate the image quality of the enhanced low-light images. The specific evaluation process is as follows:

[0035] Acquire an enhanced low-light image of the scene, calculate the gradient of the enhanced low-light image in the horizontal and vertical directions respectively, and calculate the detail evaluation value in the enhanced low-light image;

[0036] The maximum value of detail evaluation values ​​of several enhanced low-light images is obtained, the enhanced low-light image corresponding to the maximum value of detail evaluation values ​​is obtained and recorded as the target enhanced low-light image, and the image quality of the target enhanced low-light image is determined to be the best.

[0037] Step S402: Acquire several target-enhanced low-light images captured by the low-light device in the scene, and send these images to the user of the low-light device through the low-light platform.

[0038] To better implement the above method, a data processing system for low-light image enhancement is also proposed. The system includes a scene approximation evaluation module, an image fusion module, an image enhancement module, and an image transmission module.

[0039] The scene approximation evaluation module is used to assess the degree of scene approximation between the scene and the scene type data to obtain the approximate scene type.

[0040] The image fusion module is used to acquire shooting reference data corresponding to similar scene types, adjust the shooting parameters of the low-light device and take pictures to obtain a low-light image set, and fuse the low-light images in the low-light image set to obtain a fused low-light image.

[0041] The image enhancement module is used to remove noise and enhance the image of the fused low-light image, and to perform dynamic range compression on the fused low-light image to obtain the enhanced low-light image;

[0042] The image sending module is used to evaluate the image quality of the enhanced low-light image of the low-light device in the scene, obtain the target enhanced low-light image, and send the target enhanced low-light image to the user of the low-light device in the low-light platform.

[0043] Furthermore, the scene approximation evaluation module includes a data acquisition unit and a scene approximation evaluation unit;

[0044] The data acquisition unit is used to acquire scene climate data and scene type data from the micro-light platform;

[0045] The scene approximation evaluation unit is used to calculate the scene approximation value between the scene and the scene type in the scene type data, and evaluate the degree of scene approximation between the scene and the scene type based on the scene approximation value to obtain the approximate scene type.

[0046] Furthermore, the image fusion module includes a device adjustment unit and an image fusion unit;

[0047] The equipment adjustment unit is used to acquire shooting reference data for similar scene types and adjust the shooting parameters of the low-light equipment.

[0048] The image fusion unit is used to capture images of a scene using a low-light device with adjusted shooting parameters, obtain a set of low-light images, and fuse several low-light images in the low-light image set to obtain a fused low-light image.

[0049] Furthermore, the image enhancement module includes an image enhancement unit;

[0050] The image enhancement unit is used to remove noise from the fused low-light image and enhance the fused low-light image to obtain an enhanced low-light image.

[0051] Furthermore, the image transmission module includes an image quality assessment unit and an image transmission unit;

[0052] The image quality assessment unit is used to acquire several enhanced low-light images captured by the low-light device in the scene, and to assess the image quality of the several enhanced low-light images to obtain the target enhanced low-light image;

[0053] The image transmission unit is used to acquire several target-enhanced low-light images captured by the low-light device in the scene, and to send them to the user of the low-light device through the low-light platform.

[0054] Compared with the prior art, the beneficial effects of the present invention are: the present invention realizes the image enhancement effect of low-light devices in all weather and complex scenes, improves the environmental adaptability of devices, improves the image quality at the hardware level by adjusting the parameters of low-light devices, and solves the problems of obvious image noise, lack of image depth and color distortion in low-light images through image enhancement methods. At the same time, the output target enhanced low-light image can retain the enhanced details and colors to the greatest extent, while ensuring that the output is compatible with the physical limitations of display devices. Attached Figure Description

[0055] Figure 1 This is a flowchart of a data processing method for enhancing low-light images according to the present invention;

[0056] Figure 2 This is a schematic diagram of a data processing system for enhancing low-light images according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Example: Figures 1-2 As shown, the present invention provides a technical solution, a data processing method for enhancing low-light images, the method comprising:

[0059] Step S100: Obtain scene climate data of the scene where the low-light device is located, obtain scene type data of the low-light platform where the low-light device is located, evaluate the scene similarity between the scene and the scene type in the scene type data, and obtain the approximate scene type;

[0060] Step S100 includes:

[0061] Step S101: Obtain the scene where the target object is located when the low-light device is shooting, and obtain the scene climate data of the scene, including the values ​​of various climate parameters.

[0062] For example, various climate parameters include light intensity and humidity;

[0063] Step S102: Obtain scene type data from the low-light platform. The scene type data includes the range of various climate parameters corresponding to each preset scene type. Evaluate the scene similarity between the scene and each scene type. The specific evaluation process is as follows:

[0064] For example, the various scene types include rainy night scenes, cloudy snow scenes, etc.

[0065] When the value of a certain climate parameter in a scene falls within the range corresponding to a certain climate parameter in a certain scene type, that climate parameter in the scene is marked, and the total number A′ of the marked climate parameters in the scene is obtained. sum ;

[0066] Calculate the scene approximation value B = A′ between the scene and a certain scene type. sum / A sum , where A sum This represents the total number of climate parameters.

[0067] Obtain the maximum value of the scene approximation value between the scene and each scene type, determine the scene approximation between the scene type corresponding to the maximum value of the scene approximation value and the scene type, and record the scene type as the approximate scene type of the scene;

[0068] Step S200: Obtain shooting reference data corresponding to similar scene types from the low-light platform, adjust the shooting parameters of the low-light device, use the adjusted low-light device to shoot in the scene, obtain a low-light image set, and fuse the low-light images in the low-light image set according to the different exposure times of the low-light images in the low-light image set to obtain a fused low-light image.

[0069] Step S200 includes:

[0070] Step S201: Obtain the approximate scene type of the scene, and obtain the shooting reference data of the low-light device under the approximate scene type from the low-light platform. The shooting reference data includes the reference values ​​corresponding to various shooting parameters of the low-light device.

[0071] For example, shooting parameters include ISO sensitivity, aperture size, etc.

[0072] Step S202: Based on the shooting reference data, adjust the shooting parameters of the low-light device. Specifically, adjust each shooting parameter in the low-light device to the reference value corresponding to each shooting parameter in the shooting reference data.

[0073] Using a low-light device with adjusted shooting parameters, and taking pictures of the target object in the scene according to different preset exposure times, a number of low-light images are obtained. The low-light images are then collected to obtain a low-light image set, in which the low-light images are all aligned low-light images.

[0074] Step S203: Obtain a specific low-light image from the low-light image set and construct a two-dimensional coordinate system to assign a coordinate to each pixel in the specific low-light image;

[0075] Obtain the pixel weight value W(x,y) of the pixel with coordinates (x,y) in a low-light image;

[0076] For example, the specific formula for the pixel weight value W(x,y) is: W(x,y)=exp(-(I(x,y)-μ) 2 / 2σ 2 ), where I(x,y) is the pixel value of the pixel at coordinate (x,y) in a certain low-light image, σ is the preset decay rate used to control the weight decay rate, and u is the average value of the pixel at coordinate (x,y) in a certain low-light image.

[0077] Obtain the pixel weight values ​​of each pixel in several low-light images, and obtain the radiation intensity E(x,y) of the pixel with coordinates (x,y) in a certain low-light image;

[0078] For example, the specific formula for calculating the radiation intensity E(x,y) is: E(x,y)=f -1 (I(x,y)) / (t′·G), where, f -1 Let t′ be the inverse response function of the low-light device, t′ be the exposure time corresponding to a certain low-light image, and G be the gain of the low-light device.

[0079] For example, the inverse response function f of a low-light device -1 (I):

[0080]

[0081] Among them, E max I represents the maximum radiation intensity of the sensor in the low-light device. max γ is the maximum pixel value; γ is the gamma value, typically 2.2.

[0082] Several low-light images in the low-light image set are fused to obtain a fused low-light image. The specific image fusion process is as follows: calculate the brightness contribution value F(x,y) = W(x,y) × E(x,y) of the pixel with coordinates (x,y) in the fused low-light image.

[0083] Calculate the brightness value I of the pixel with coordinates (x, y) in the fused low-light image. hdr (x,y) is used to obtain the brightness value of each pixel in the fused low-light image;

[0084] For example, the brightness value I of the pixel with coordinates (x,y) in a low-light image is fused. hdr The specific formula for calculating (x, y) is as follows:

[0085]

[0086] Where ε is a preset minimum constant; W i (x,y) represents the pixel weight value of the pixel with coordinates (x,y) in the i-th low-light image of the low-light image set; F i (x,y) represents the brightness contribution value of the pixel with coordinates (x,y) in the i-th low-light image; n represents the total number of low-light images in the low-light image set.

[0087] Step S300: Remove noise from the fused low-light image, enhance the fused low-light image after noise removal, and compress the dynamic range of the fused low-light image to obtain the enhanced low-light image;

[0088] Step S300 includes:

[0089] Step S301: Separate the noise from the fused low-light image and remove the noise from the fused low-light image;

[0090] For example, common methods for separating and removing noise from fused low-light images include:

[0091] Median filtering: Replacing the current pixel with the median of its neighborhood is more effective at removing salt-and-pepper noise while preserving edge information.

[0092] Gaussian filtering: Applying a Gaussian kernel to blur an image can effectively remove Gaussian noise.

[0093] Fourier transform: Converts an image to the frequency domain and filters out high-frequency components to remove noise (low-pass filtering).

[0094] Wavelet transform: Decompose the image using wavelet transform, then remove noise in different frequency bands, and finally reconstruct the image;

[0095] Convolutional Neural Networks (CNNs): Using deep learning frameworks to train models to identify and remove noise can achieve better results;

[0096] Non-local means denoising (NLM): Utilizes pixel information from similar regions in an image to smooth noise while preserving details as much as possible;

[0097] Step S302: Perform image enhancement on the fused low-light image after noise removal. Specifically, for the pixel with coordinates (x, y) in the fused low-light image, the processing procedure is as follows: For the brightness value I in the fused low-light image... hdr Performing a logarithmic transformation on (x,y) yields the logarithmic value L(x,y) = log(I) of the pixel at coordinates (x,y) in the fused low-light image. hdr (x,y)+ε), where ε is a pre-defined local constant;

[0098] The logarithmic value L(x,y) is convolved using Gaussian kernels of k different scales to calculate the illumination components at k different scales. Specifically, at the k-th scale, the illumination component Lk is the illumination component of the pixel with coordinates (x,y) in the fused low-light image. k (x, y), where k is a preset value, k > 0;

[0099] For example, k is usually 3, L k (x,y)=G σk (x,y)*L(x,y), where G σk (x,y) represents the Gaussian kernel for the pixel at coordinates (x,y) of the k-th scale. σ The specific formula for (x, y) is:

[0100]

[0101] Where σ represents the standard deviation, which affects the shape and width of the Gaussian function;

[0102] Calculate the detail component R of the pixel with coordinates (x, y) in the fused low-light image at the k-th scale. k (x,y)=L(x,y)-L k (x,y);

[0103] The detail components of the pixel with coordinates (x,y) in the fused low-light image at k scales are obtained and weighted and fused to obtain the detail fusion value R(x,y). The detail fusion value R(x,y) is then subjected to an exponential transformation to obtain the target brightness I′(x,y)=exp(R(x,y)).

[0104] For example, when k is 3, the detail components of the pixels with coordinates (x,y) in the fused low-light image at k scales are obtained and weighted, and the specific formula for the obtained detail fusion value R(x,y) is as follows:

[0105]

[0106] Where w1, w2 and w3 are preset weights, with values ​​of 0.6, 0.3 and 0.1 respectively; R1(x,y), R2(x,y) and R3(x,y) are detail components of the pixels with coordinates (x,y) in the fused low-light images at the 1st, 2nd and 3rd scales respectively.

[0107] Step S303: Obtain the fused low-light image after image enhancement, and obtain the maximum value I′ of the target brightness of each pixel in the fused low-light image. max The enhanced low-light image is obtained by dynamically compressing the fused low-light image at coordinates (x,y). Specifically, the compressed target brightness I′(x,y) of the pixel at coordinates (x,y) in the enhanced low-light image is obtained after dynamic range compression of the target brightness I′(x,y). LDR (x,y), I′ LDR (x,y)∈[0,1];

[0108] For example, I′ LDR The specific formula for (x, y) is:

[0109]

[0110] Where μ is a preset compressive strength parameter (usually taken as 0.1 to 1.0);

[0111] Step S400: Acquire the enhanced low-light image of the low-light device in the scene, evaluate the image quality of the enhanced low-light image, obtain the target enhanced low-light image, and send the target enhanced low-light image to the user of the low-light device through the low-light platform;

[0112] Step S400 includes:

[0113] Step S401: Acquire several enhanced low-light images of the scene captured within the current period, and evaluate the image quality of the enhanced low-light images. The specific evaluation process is as follows:

[0114] Acquire an enhanced low-light image of the scene, calculate the gradient of the enhanced low-light image in the horizontal and vertical directions respectively, and calculate the detail evaluation value in the enhanced low-light image;

[0115] For example, the process of calculating the detail evaluation value EI in an enhanced low-light image I is as follows:

[0116] Calculate the gradient G in the horizontal and vertical directions of the enhanced low-light image, respectively. x =S x *I, G y =S y *I, where S x and S y They are respectively:

[0117]

[0118] Calculate the gradient magnitude G of the pixel with coordinates (x, y) in the enhanced low-light image I. m (x,y):

[0119]

[0120] The average gradient magnitude of each pixel in the enhanced low-light image I is obtained to obtain the detail evaluation value EI in the enhanced low-light image I.

[0121] The maximum value of detail evaluation values ​​of several enhanced low-light images is obtained, the enhanced low-light image corresponding to the maximum value of detail evaluation values ​​is obtained and recorded as the target enhanced low-light image, and the image quality of the target enhanced low-light image is determined to be the best.

[0122] Step S402: Acquire several target-enhanced low-light images captured by the low-light device in the scene, and send these target-enhanced low-light images to the user of the low-light device through the low-light platform;

[0123] To better implement the above method, a data processing system for low-light image enhancement is also proposed. The system includes a scene approximation evaluation module, an image fusion module, an image enhancement module, and an image transmission module.

[0124] The scene approximation evaluation module is used to assess the degree of scene approximation between the scene and the scene type data to obtain the approximate scene type.

[0125] The image fusion module is used to acquire shooting reference data corresponding to similar scene types, adjust the shooting parameters of the low-light device and take pictures to obtain a low-light image set, and fuse the low-light images in the low-light image set to obtain a fused low-light image.

[0126] The image enhancement module is used to remove noise and enhance the image of the fused low-light image, and to perform dynamic range compression on the fused low-light image to obtain the enhanced low-light image;

[0127] The image transmission module is used to evaluate the image quality of the enhanced low-light image of the low-light device in the scene, obtain the target enhanced low-light image, and send the target enhanced low-light image to the user of the low-light device in the low-light platform;

[0128] The scene approximation evaluation module includes a data acquisition unit and a scene approximation evaluation unit.

[0129] The data acquisition unit is used to acquire scene climate data and scene type data from the micro-light platform;

[0130] The scene approximation evaluation unit is used to calculate the scene approximation value between the scene and the scene type in the scene type data, and evaluate the degree of scene approximation between the scene and the scene type based on the scene approximation value to obtain the approximate scene type;

[0131] The image fusion module includes a device adjustment unit and an image fusion unit.

[0132] The equipment adjustment unit is used to acquire shooting reference data for similar scene types and adjust the shooting parameters of the low-light equipment.

[0133] The image fusion unit is used to capture the scene using a low-light device with adjusted shooting parameters to obtain a low-light image set, and to fuse several low-light images in the low-light image set to obtain a fused low-light image.

[0134] The image enhancement module includes an image enhancement unit;

[0135] The image enhancement unit is used to remove noise from the fused low-light image and enhance the fused low-light image to obtain an enhanced low-light image;

[0136] The image transmission module includes an image quality assessment unit and an image transmission unit.

[0137] The image quality assessment unit is used to acquire several enhanced low-light images captured by the low-light device in the scene, and to assess the image quality of the several enhanced low-light images to obtain the target enhanced low-light image;

[0138] The image transmission unit is used to acquire several target-enhanced low-light images captured by the low-light device in the scene, and to send them to the user of the low-light device through the low-light platform.

[0139] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A data processing method for enhancing low-light images, characterized in that, The method includes: Step S100: Obtain scene climate data of the scene where the low-light device is located, obtain scene type data in the low-light platform where the low-light device is located, evaluate the scene similarity between the scene and the scene type in the scene type data, and obtain the approximate scene type; Step S200: Obtain the shooting reference data corresponding to the approximate scene type from the low-light platform, adjust the shooting parameters of the low-light device, use the adjusted low-light device to shoot in the scene, obtain a low-light image set, and fuse the low-light images in the low-light image set according to the different exposure times of the low-light images in the low-light image set to obtain a fused low-light image. Step S300: Remove noise from the fused low-light image, enhance the fused low-light image after noise removal, and compress the fused low-light image dynamically to obtain an enhanced low-light image; Step S400: Acquire an enhanced low-light image of the low-light device in the scene, evaluate the image quality of the enhanced low-light image, obtain a target enhanced low-light image, and send the target enhanced low-light image to the user of the low-light device through the low-light platform.

2. The data processing method for low-light image enhancement according to claim 1, characterized in that, Step S100 includes: Step S101: Obtain the scene where the target object is located when the low-light device is shooting, and obtain the scene climate data of the scene, which includes the values ​​of various climate parameters. Step S102: Obtain scene type data from the low-light platform. The scene type data includes the range of each climate parameter in each preset scene type. Evaluate the scene similarity between the scene and each scene type. The specific evaluation process is as follows: When the value of a certain climate parameter in the scenario falls within the range corresponding to that climate parameter in a certain scenario type, the climate parameter in the scenario is marked, and the total number A′ of the marked climate parameters in the scenario is obtained. sum ; Calculate the scene approximation value B = A′ between the scene and a certain scene type. sum / A sum , where A sum This represents the total number of the aforementioned climate parameters; Obtain the maximum value of the scene approximation value between the scene and each of the scene types, determine the scene type corresponding to the maximum value of the scene approximation value and the scene approximation between the scene and the scene, and record the scene type as the approximate scene type of the scene.

3. The data processing method for low-light image enhancement according to claim 2, characterized in that, Step S200 includes: Step S201: Obtain the approximate scene type of the scene, and obtain the shooting reference data of the low-light device under the approximate scene type from the low-light platform. The shooting reference data includes the reference values ​​corresponding to each shooting parameter of the low-light device. Step S202: Based on the shooting reference data, adjust the shooting parameters of the low-light device, specifically, adjust each shooting parameter in the low-light device to the reference value corresponding to each shooting parameter in the shooting reference data; Using the low-light equipment with the shooting parameters adjusted as described above, and according to different preset exposure times, the target object is photographed in the scene to obtain several low-light images. The several low-light images are then collected to obtain a low-light image set, wherein the several low-light images are all aligned low-light images. Step S203: Obtain a certain low-light image from the set of low-light images, and construct a two-dimensional coordinate system to set a coordinate for each pixel in the certain low-light image; Obtain the pixel weight value W(x,y) of the pixel with coordinates (x,y) in a certain low-light image; Obtain the pixel weight value of each pixel in several low-light images in the low-light image, and obtain the radiation intensity E(x,y) of the pixel with coordinates (x,y) in a certain low-light image; Several low-light images in the low-light image set are fused to obtain a fused low-light image. The specific image fusion process is as follows: calculate the brightness contribution value F(x,y) = W(x,y) × E(x,y) of the pixel with coordinates (x,y) in the fused low-light image. Calculate the brightness value I of the pixel with coordinates (x, y) in the fused low-light image. hdr (x,y) is used to obtain the brightness value of each pixel in the fused low-light image.

4. The data processing method for low-light image enhancement according to claim 3, characterized in that, Step S300 includes: Step S301: Separate the noise from the fused low-light image and remove the noise from the fused low-light image; Step S302: Perform image enhancement on the fused low-light image after noise removal. Specifically, for the pixel with coordinates (x, y) in the fused low-light image, the processing procedure is as follows: For the brightness value I in the fused low-light image... hdr Logarithmic transformation is performed on (x,y) to obtain the logarithmic value L(x,y) = log(I) of the pixel with coordinates (x,y) in the fused low-light image. hdr (x,y)+ε), where ε is a pre-defined local constant; The logarithmic value L(x,y) is convolved using Gaussian kernels of k different scales to calculate the illumination components at the k different scales, where the illumination component L at coordinates (x,y) of the pixel in the fused low-light image at the k-th scale is... k (x, y), where k is a preset value, k > 0; Calculate the detail component R of the pixel with coordinates (x, y) in the fused low-light image at the k-th scale. k (x,y)=L(x,y)-L k (x,y); The detail components of the pixels with coordinates (x,y) in the fused low-light image at the k scales are obtained and weighted and fused to obtain the detail fusion value R(x,y). The detail fusion value R(x,y) is then subjected to an exponential transformation to obtain the target brightness I′(x,y)=exp(R(x,y)). Step S303: Obtain the fused low-light image after image enhancement, and obtain the maximum value I′ of the target brightness of each pixel in the fused low-light image. max The fused low-light image is dynamically compressed to obtain an enhanced low-light image. Specifically, after dynamic range compression of the target brightness I′(x,y), the compressed target brightness I′ of the pixel at coordinates (x,y) in the enhanced low-light image is obtained. LDR (x,y), I′ LDR (x,y)∈[0,1].

5. A data processing method for enhancing low-light images according to claim 4, characterized in that, Step S400 includes: Step S401: Acquire several enhanced low-light images of the scene captured within the current period, and evaluate the image quality of the enhanced low-light images. The specific evaluation process is as follows: Acquire an enhanced low-light image of the scene, calculate the gradient of the enhanced low-light image in the horizontal and vertical directions respectively, and calculate the detail evaluation value in the enhanced low-light image; The maximum value of the detail evaluation value of the plurality of enhanced low-light images is obtained, the enhanced low-light image corresponding to the maximum value of the detail evaluation value is obtained and recorded as the target enhanced low-light image, and the image quality of the target enhanced low-light image is determined to be the best. Step S402: Acquire several target-enhanced low-light images captured by the low-light device in the scene, and send the several target-enhanced low-light images to the user of the low-light device through the low-light platform.

6. A data processing system for realizing low-light image enhancement, used to execute the data processing method for realizing low-light image enhancement as described in any one of claims 1-5, characterized in that, The system includes a scene approximation evaluation module, an image fusion module, an image enhancement module, and an image transmission module; The scene approximation evaluation module is used to determine the degree of scene approximation between the scene and the scene type in the scene type data, and to obtain the approximate scene type. The image fusion module is used to acquire shooting reference data corresponding to the similar scene type, adjust the shooting parameters of the low-light device and take a picture to obtain a low-light image set, and fuse the low-light images in the low-light image set to obtain a fused low-light image. The image enhancement module is used to remove noise and enhance the image of the fused low-light image, and to perform dynamic range compression on the fused low-light image to obtain an enhanced low-light image; The image sending module is used to evaluate the image quality of the enhanced low-light image of the low-light device in the scene, obtain the target enhanced low-light image, and send the target enhanced low-light image to the user of the low-light device in the low-light platform.

7. A data processing system for realizing low-light image enhancement according to claim 6, characterized in that, The scene approximation evaluation module includes a data acquisition unit and a scene approximation evaluation unit; The data acquisition unit is used to acquire scene climate data of the scene and scene type data in the low-light platform; The scene approximation evaluation unit is used to calculate the scene approximation value between the scene and the scene type in the scene type data, and evaluate the scene approximation degree between the scene and the scene type based on the scene approximation value to obtain the approximate scene type.

8. A data processing system for realizing low-light image enhancement according to claim 6, characterized in that, The image fusion module includes a device adjustment unit and an image fusion unit; The device adjustment unit is used to acquire shooting reference data under a similar scene type to the scene, and to adjust the shooting parameters of the low-light device; The image fusion unit is used to capture the scene using the low-light device with adjusted shooting parameters to obtain a low-light image set, and to fuse several low-light images in the low-light image set to obtain a fused low-light image.

9. A data processing system for realizing low-light image enhancement according to claim 6, characterized in that, The image enhancement module includes an image enhancement unit; The image enhancement unit is used to remove noise from the fused low-light image and enhance the fused low-light image to obtain an enhanced low-light image.

10. A data processing system for realizing low-light image enhancement according to claim 6, characterized in that, The image transmission module includes an image quality assessment unit and an image transmission unit; The image quality assessment unit is used to acquire several enhanced low-light images captured by the low-light device in the scene, and to perform image quality assessment on the several enhanced low-light images to obtain the target enhanced low-light image. The image sending unit is used to acquire several target-enhanced low-light images captured by the low-light device in the scene, and send them to the user of the low-light device through the low-light platform.

Images