WO2023092386A1 - 一种图像处理方法、终端设备及计算机可读存储介质 - Google Patents

一种图像处理方法、终端设备及计算机可读存储介质 Download PDF

Info

Publication number
WO2023092386A1
WO2023092386A1 PCT/CN2021/133130 CN2021133130W WO2023092386A1 WO 2023092386 A1 WO2023092386 A1 WO 2023092386A1 CN 2021133130 W CN2021133130 W CN 2021133130W WO 2023092386 A1 WO2023092386 A1 WO 2023092386A1
Authority
WO
WIPO (PCT)
Prior art keywords
feature
network
image
trained
color enhancement
Prior art date
Application number
PCT/CN2021/133130
Other languages
English (en)
French (fr)
Inventor
张锲石
程俊
欧阳祖薇
任子良
高向阳
康宇航
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Priority to PCT/CN2021/133130 priority Critical patent/WO2023092386A1/zh
Publication of WO2023092386A1 publication Critical patent/WO2023092386A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting

Definitions

  • the present application belongs to the technical field of image processing, and in particular relates to an image processing method, a terminal device, and a computer-readable storage medium.
  • the method of brightening images captured under low-light conditions is usually based on retinal theory or generative adversarial networks to achieve brightening.
  • the method based on retinal theory is to use two networks, one of which combines low-light The image is decomposed into illumination and reflection, and another network is used as an enhancer to improve the illumination map of the low-light image.
  • the method based on generating an adversarial network is through Finding the mapping between low-light and normal-light domains, treating low-light enhancement as a domain-transfer learning task, using the generator to estimate normal-light images from low-light images, using the discriminator to estimate the visual quality, however this generator and discriminator
  • the machine needs to be trained at the same time and the expectations are opposite, so it is difficult to get the desired output, and it will also lead to the problem of poor processing effect.
  • the embodiment of the present application provides an image processing method, a terminal device, and a computer-readable storage medium to solve the problem of poor image processing effect in the current method of brightening images captured under low-light conditions. .
  • the embodiment of the present application provides an image processing method, including:
  • the feature images of different scales are input into the trained color enhancement network for processing to obtain the target image.
  • the image processing method also includes:
  • the continuous update network and the color enhancement network are trained based on the sample data set to obtain the trained continuous update network and the trained color enhancement network.
  • the continuous update network includes N U-shaped network structures and feature fusion structures, where N is a positive integer greater than or equal to 2.
  • the U-shaped network structure includes a downsampling structure and an upsampling structure
  • the downsampling structure includes four feature blocks of different scales
  • the upsampling structure includes four feature blocks of different scales
  • the different feature blocks are connected using skip connections.
  • the feature fusion structure includes a convolutional layer, an average pooling layer and a fully connected layer.
  • the color enhancement network includes a residual acquisition module, a convolution module and a weight determination module.
  • the feature images of different scales include a first scale feature image, a second scale feature image, and a third feature scale image.
  • the embodiment of the present application provides a terminal device, including:
  • the feature images of different scales include a first scale feature image, a second scale feature image and a third feature scale image.
  • an embodiment of the present application provides a terminal device, the terminal device includes a processor, a memory, and a computer program stored in the memory and operable on the processor, and the processor executes the The computer program implements the method described in the first aspect or any optional manner of the first aspect.
  • an embodiment of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any operable solution in the first aspect or the first aspect can be implemented. method as described in the selection method.
  • an embodiment of the present application provides a computer program product, which, when the computer program product is run on a terminal device, causes the terminal device to execute the method described in the first aspect or any optional manner of the first aspect.
  • the feature images of different scales are extracted through the continuous update network, and the global information and local information are extracted, which can ensure that the restored normal illumination image will not have missing details, and then enhance the image color and color of the restored image through the color enhancement network. Texture ensures image quality, and solves the problem of poor image processing effect in the current method of brightening images captured under low-light conditions.
  • FIG. 1 is a schematic flow chart of an image processing method provided in an embodiment of the present application
  • Fig. 2 is a schematic structural diagram of the U-shaped network structure of the continuous update network in the embodiment of the present application;
  • Fig. 3 is a schematic flow chart of the feature fusion process provided by the embodiment of the present application.
  • Fig. 4 is a schematic diagram of the process of processing feature images of different scales by the color enhancement network provided by the embodiment of the present application;
  • FIG. 5 is a schematic structural diagram of a terminal device provided in an embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of a terminal device provided by another embodiment of the present application.
  • Fig. 7 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present application.
  • references to “one embodiment” or “some embodiments” or the like described in the specification of the present application mean that a particular feature, structure or characteristic described in connection with the embodiment is included in one or more embodiments of the present application .
  • appearances of the phrases “in one embodiment,” “in some embodiments,” “in other embodiments,” “in other embodiments,” etc. in various places in this specification are not necessarily All refer to the same embodiment, but mean “one or more but not all embodiments” unless specifically stated otherwise.
  • the terms “including”, “comprising”, “having” and variations thereof mean “including but not limited to”, unless specifically stated otherwise.
  • Low-light environment refers to a specific environment where natural lighting conditions are not good and affects people's vision.
  • an environment with an illuminance below 50 lumens is called a low-light environment.
  • Images captured in low-light environments usually have problems such as low brightness, low contrast, and many noise points.
  • An image processing method provided by the embodiment of the present application extracts feature images of different scales through the continuous update network, extracts global information and local information, and can ensure that the recovered normal illumination image will not have missing details, and then through The color enhancement network is used to enhance the image color and texture of the restored image, ensure the image quality, and solve the problem of poor image processing effect in the current method of brightening the image captured under low-light conditions.
  • FIG. 1 is a schematic flowchart of an image processing method provided by an embodiment of the present application.
  • the execution subject of the image processing method provided in the embodiment of the present application is a terminal device, which can be a mobile terminal such as a smart phone, a tablet computer, or a wearable device, or a computer, a cloud server, an auxiliary computer, etc. in various application scenarios. .
  • the image processing method as shown in Figure 1 may include S11 ⁇ S12, which are described in detail as follows:
  • S11 Input the low-light image into the trained continuous update network to extract local features and global features, and obtain feature images of different scales.
  • the local features and global features in the low-light image are extracted by continuously updating the network, and then the local features and the global features are fused to obtain feature images of different scales.
  • the above-mentioned continuous update network may include N U-shaped network structures, and image features of different scales are extracted through the U-shaped network.
  • the above-mentioned continuous update network also includes a feature fusion structure, and U The image features of different scales extracted by the network-type network are used for feature fusion, and finally feature images of different scales are obtained.
  • N is a positive integer greater than or equal to 2.
  • the above-mentioned continuous update network may include four U-shaped network structures, and the structure of each U-shaped network structure may be referred to in FIG. 2 , which shows the U-shaped continuous update network in the embodiment of the present application.
  • FIG. 2 shows the U-shaped continuous update network in the embodiment of the present application.
  • Schematic diagram of the network structure illustrated with two U-shaped networks).
  • the above-mentioned U-shaped network structure may include a down-sampling structure 21 and an up-sampling structure 22, the above-mentioned down-sampling structure includes four feature blocks of different scales, four feature blocks of different scales (A1 in Figure 2 , A2, A3, A4), the scale ratio from high to low can be 8:4:2:1, that is, the step size of the convolution of the same scale of two adjacent feature blocks is 1, and the two adjacent features and the reduction
  • the scale convolution has a stride of 2.
  • the upsampling structure also includes four feature blocks of different scales, and the scale ratio of the four feature blocks of different scales (such as B1, B2, B3, and B4 in Figure 2) can be 1:2:4:8 from low to high. , that is, the step size of the convolution of the same scale of two adjacent feature blocks is 1, and the step size of the increased deconvolution obtained by the scale of two adjacent feature blocks is 2.
  • Different feature blocks are connected using a skip connection. Specifically, by directly merging the different scale features obtained by the downsampling structure with the same scale features obtained by the upsampling settlement, the different scales obtained in the upsampling stage Features are fused with larger-scale features through upsampling. By skipping connections, the upsampling stage can simultaneously acquire global features and local features.
  • connection method can be to double or triple the features of the four different scales obtained by the two U-shaped networks, and then use the feature fusion structure (FF structure) to obtain the four
  • FF structure feature fusion structure
  • the features of the same scale are fused to obtain a single feature, and the obtained single feature will be retransmitted and continue to operate in the U-shaped network.
  • This connection method enables the features of the second U-shaped network to obtain more global features and local features.
  • the above-mentioned feature fusion structure may include a convolutional layer, an average pooling layer, and a fully connected layer.
  • FIG. 3 shows a schematic flowchart of a feature fusion process provided by an embodiment of the present application.
  • the features extracted by the U-shaped structure are first input into the feature fusion module, and the convolution layer in the feature fusion module is processed to obtain the initial convolution, and then the initial convolution is passed through the average pool
  • the layer will compress the features, and then get a single feature value, and then determine the correlation between different channels through two fully connected layers to get the weight value, and then combine the obtained weight value with the feature obtained after the initial convolution Values are multiplied to get the first weighted feature, add the first weighted feature to the feature value obtained after the initial convolution, repeat the above operation, you can get the second special weighted feature value, and finally, compress the second weighted feature,
  • the feature image corresponding to the fusion of these four features can be obtained.
  • the features of different scales are operated in this way, and finally the feature images of different scales can be obtained.
  • the aforementioned feature images of different scales include a feature image of a first scale, a feature image of a second scale, and a feature image of a third scale.
  • S12 Input the feature images of different scales into the trained color enhancement network for processing to obtain a target image.
  • the color enhancement network can determine the weight of each feature image based on the residuals of feature images of different scales, and then multiply each feature image by the corresponding weight to combine an image, that is, target image.
  • FIG. 4 is a schematic diagram of a process of processing feature images of different scales by the color enhancement network provided by the embodiment of the present application.
  • the above color enhancement network 400 may include a residual acquisition module 401 , a convolution module 402 and a weight determination module 403 .
  • the first residual Y2-Y1, the second residual Y3-Y2, and the third residual Y3-Y1 are input into the convolution unit for processing, and the first feature D1 and the second feature D2 can be obtained, and the first The feature D1 and the second feature D2 are convolved to obtain the third feature D3 and the fourth feature D4, and then the second feature D2 and the fourth feature D4 are fused through the weight determination module, and the fifth feature D5 is obtained by convolution, and then the The weight of the fifth feature D5 is decomposed by the weight determination module 403 to obtain the first weight G1, the second weight G2 and the third weight G3.
  • the first weight G1 is multiplied by the first scale feature image Y1
  • the second weight G2 is multiplied by the second-scale feature image Y2
  • the third weight G3 is multiplied by the third-scale feature image Y3, and then the multiplied images are fused to obtain the target image.
  • the above image processing method may further include the following steps:
  • the continuous update network and the color enhancement network are trained based on the sample data set to obtain the trained continuous update network and the trained color enhancement network.
  • the above sample data may include historical low-light images and corresponding normal-light images.
  • sample data set In practical application, you can select not less than 1000 sets of sample data to obtain a sample data set. Divide the sample dataset into training, validation, and test sets. In order to meet the training requirements, 50% of the sample data can be used as the training set, and the rest as the verification set and test set.
  • the continuous update network is trained with the training set data, and the verification set is used to quickly adjust the parameters, and then the test set is used to test the continuous update network, and the trained continuous update network is obtained.
  • the historical low-light images in the sample image can be input into the pre-built continuous update network for processing to obtain feature images of different scales, and then the loss function based on the normal illumination image and the obtained feature images of different scales Adjust the network parameters in the continuous update network.
  • the structural similarity loss and Total Variation (TV) loss between the third-scale feature image and the corresponding normal illumination image can be adjusted as a loss function.
  • the loss When the function converges, verify and test the continuously updated network with adjusted network parameters based on the sample data in the verification set and test set. If the verification and test pass, it means that the training of the continuous update network is completed, and the continuous update network can be obtained after the training is completed. Used in S11.
  • SSIM Structural Similarity
  • the feature images of different scales output by the trained continuous update network can be input into the pre-built color enhancement network for processing to obtain the target image, and then based on the normal illumination image and the loss function of the obtained target image.
  • the network parameters in the color enhancement network are adjusted. Specifically, the perception loss (perception loss) and the total variation (Total Variation, TV) loss of the target image and the corresponding normal illumination image can be used as the loss function to adjust the network parameters.
  • the loss function converges, verify and test the color enhancement network after adjusting the network parameters based on the sample data in the verification set and test set. If the verification and test pass, it means that the training of the color enhancement network is completed, and the color enhancement network obtained after the training can be completed. Used in S12.
  • the above perceptual loss includes feature reconstruction loss (feature reconstruction loss) and style reconstruction loss (style reconstruction loss), the above-mentioned feature reconstruction loss can represent the Euclidean distance between features, so as to measure the similarity of features, the style reconstruction loss is mainly to get better image color and texture, the style reconstruction loss is the output image (output The target image) is only the squared Frobenis norm that differs from the Gram matrix of the normally illuminated image, and the image that minimizes the style reconstruction loss retains the style features of the normally illuminated image, but does not preserve its spatial structure.
  • Perceptual loss is a loss function that combines feature reconstruction loss and style reconstruction loss. Similarly, in order to minimize the gradient of the entire image, the loss function of the color enhancement network is also constructed based on the total variation loss.
  • the image processing method provided by the embodiment of the present application extracts feature images of different scales through the continuous update network, extracts global information and local information, and can ensure that the recovered normal illumination image will not have the phenomenon of missing details , and then enhance the image color and texture of the restored image through the color enhancement network to ensure the image quality, and solve the problem of poor image processing effect in the current method of brightening the image captured under low-light conditions.
  • the embodiment of the present invention further provides an embodiment of a terminal device implementing the foregoing method embodiment.
  • FIG. 5 is a schematic structural diagram of a terminal device provided in an embodiment of the present application.
  • each unit included in the terminal device is configured to execute each step in the embodiments corresponding to FIG. 1 to FIG. 4 .
  • the terminal device 50 includes: a feature extraction module 51 and a color enhancement module 52 . in:
  • the feature extraction module 51 is used to input the low-light images into the trained continuous update network for local feature and global feature extraction to obtain feature images of different scales.
  • the color enhancement module 52 is used to input the feature images of different scales into the trained color enhancement network for processing to obtain the target image.
  • the terminal device also includes:
  • An acquisition module used to acquire a sample data set
  • the training module is used to train the continuous update network and the color enhancement network based on the sample data set, so as to obtain the trained continuous update network and the trained color enhancement network.
  • the continuous update network includes N U-shaped network structures and feature fusion structures, where N is a positive integer greater than or equal to 2.
  • the U-shaped network structure includes a downsampling structure and an upsampling structure
  • the downsampling structure includes four feature blocks of different scales
  • the upsampling structure includes four feature blocks of different scales
  • the different feature blocks are connected using skip connections.
  • the feature fusion structure includes a convolutional layer, an average pooling layer and a fully connected layer.
  • the color enhancement network includes a residual acquisition module, a convolution module and a weight determination module.
  • the terminal device provided by the embodiment of the present application can also extract feature images of different scales through the continuous update network, extract global information and local information, and can ensure that the restored normal illumination image will not have missing details. Then, the image color and texture of the restored image are enhanced through the color enhancement network to ensure the image quality, and solve the problem of poor image processing effect in the current method of brightening the image captured under low-light conditions.
  • Fig. 6 is a schematic structural diagram of a terminal device provided by another embodiment of the present application.
  • the terminal device 6 provided in this embodiment includes: a processor 60 , a memory 61 , and a computer program 62 stored in the memory 61 and operable on the processor 60 , such as an image processing program.
  • the processor 60 executes the computer program 62, it realizes the steps in the above embodiments of the various image processing methods, such as S11-S12 shown in FIG. 1 .
  • functions of the modules/units in the foregoing terminal device embodiments are implemented, for example, functions of the units 51-52 shown in FIG. 5 .
  • the computer program 62 can be divided into one or more modules/units, and the one or more modules/units are stored in the memory 61 and executed by the processor 60 to complete the application .
  • the one or more modules/units may be a series of computer program instruction segments capable of accomplishing specific functions, and the instruction segments are used to describe the execution process of the computer program 62 in the terminal device 6 .
  • the computer program 62 may be divided into a first acquisition unit and a first processing unit. For specific functions of each unit, please refer to the relevant description in the embodiment corresponding to FIG. 5 , which will not be repeated here.
  • the terminal device may include, but not limited to, a processor 60 and a memory 61 .
  • FIG. 6 is only an example of the terminal device 6, and does not constitute a limitation on the terminal device 6. It may include more or less components than those shown in the figure, or combine certain components, or different components. , for example, the terminal device may also include an input and output device, a network access device, a bus, and the like.
  • the so-called processor 60 can be a central processing unit (Central Processing Unit, CPU), and can also be other general-purpose processors, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the storage 61 may be an internal storage unit of the terminal device 6 , such as a hard disk or memory of the terminal device 6 .
  • the memory 61 may also be an external storage device of the terminal device 6, such as a plug-in hard disk equipped on the terminal device 6, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (Flash Card), etc. Further, the memory 61 may also include both an internal storage unit of the terminal device 6 and an external storage device.
  • the memory 61 is used to store the computer program and other programs and data required by the terminal device.
  • the memory 61 can also be used to temporarily store data that has been output or will be output.
  • FIG. 7 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of the present application. As shown in FIG. The above image processing method can be realized during execution.
  • An embodiment of the present application provides a computer program product.
  • the terminal device can implement the above image processing method when executed.
  • Module completion means that the internal structure of the terminal device is divided into different functional units or modules, so as to complete all or part of the functions described above.
  • Each functional unit and module in the embodiment may be integrated into one processing unit, or each unit may exist separately physically, or two or more units may be integrated into one unit, and the above-mentioned integrated units may adopt hardware It can also be implemented in the form of software functional units.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Multimedia (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Image Analysis (AREA)

Abstract

一种图像处理方法、终端设备及计算机可读存储介质，包括：将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像，将不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像，通过连续更新网络提取出不同尺度的特征图像，提取出全局信息和局部信息，能够保证恢复得到的正常光照图像不会出现细节缺失的现象，再通过色彩增强网络来增强恢复图像的图像颜色和纹理，保证图像质量，解决了目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。

Description

一种图像处理方法、终端设备及计算机可读存储介质 技术领域
本申请属于图像处理技术领域，尤其涉及一种图像处理方法、终端设备及计算机可读存储介质。
 背景技术
在工业生产、视频监控、智能交通、遥感控制等多个领域，都涉及到对采集到的图像进行图像识别、图像信息提取等多种处理，然而在弱光条件下拍摄到的图像往往会由于光线较暗导致图像存在亮度低、对比度低、噪声点多等问题，导致后续的图像处理效果差，无法识别出有效的图像信息的问题。
为了提高图像识别的可靠性和鲁棒性，需要对弱光条件下拍摄到的图像进行增亮。目前，对弱光条件下拍摄到的图像进行增亮的方法通常是视网膜理论或基于生成对抗式网络来实现增亮，其中，基于视网膜理论的方式是利用两个网络，其中一个网络将弱光图像分解为光照和反射，另一个网络作为一个增强器来改进弱光图像的照明图，然而由于照度和反射的分解存在较大难度，导致处理效果不佳；基于生成对抗式网络的方式是通过寻找弱光和正常光域之间的映射，将弱光增强视为域转移学习任务，利用发生器从弱光图像中估计正常光图像，利用鉴别器估计的视觉质量，然而这发生器和鉴别器需要同时训练且期望是相反的，因此很难得到期望的输出，也会导致处理效果不佳的问题。
综上可知，目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。
 技术问题
有鉴于此，本申请实施例提供了一种图像处理方法、终端设备及计算机可读存储介质，以解决目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。
 技术解决方案
第一方面，本申请实施例提供一种图像处理方法，包括：
将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像；
将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
可选的，所述图像处理方法还包括：
构建连续更新网络和色彩增强网络；
获取样本数据集；
基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
可选的，连续更新网络包括N个U型网络结构和特征融合结构，其中，N为大于等于2的正整数。
可选的，所述U型网络结构包括下采样结构和上采样结构，所述下采样结构包括四个不同尺度的特征块，所述上采样结构包括四个不同尺度的特征块，不同特征块之间使用跳过连接进行连接。
可选的，所述特征融合结构包括卷积层、平均池层和完全连通层。
可选的，所述色彩增强网络包括残差获取模块、卷积模块和权重确定模块。
可选的，所述不同尺度的特征图像包括第一尺度特征图像、第二尺度特征图像以及第三特征尺度图像。
第二方面，本申请实施例提供一种终端设备，包括：
所述不同尺度的特征图像包括第一尺度特征图像、第二尺度特征图像以及第三特征尺度图像。
第三方面，本申请实施例提供一种终端设备，所述终端设备包括处理器、存储器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如第一方面或第一方面的任意可选方式所述的方法。
第四方面，本申请实施例提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现如第一方面或第一方面的任意可选方式所述的方法。
第五方面，本申请实施例提供一种计算机程序产品，当计算机程序产品在终端设备上运行时，使得终端设备执行上述第一方面或第一方面的任意可选方式所述的方法。
 有益效果
实施本申请实施例提供的一种图像处理方法、终端设备、计算机可读存储介质及计算机程序产品具有以下有益效果：
通过连续更新网络提取出不同尺度的特征图像，提取出全局信息和局部信息，能够保证恢复得到的正常光照图像不会出现细节缺失的现象，再通过色彩增彩网络来增强恢复图像的图像颜色和纹理，保证图像质量，解决了目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。
 附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本申请实施例提供的一种图像处理方法的示意性流程图；
图2是本申请实施例中连续更新网络的U型网络结构的结构示意图；
图3是本申请实施例提供的特征融合过程的流程示意图；
图4是本申请实施例提供的色彩增强网络对不同尺度的特征图像进行处理的过程示意图；
图5是本申请实施例提供的一种终端设备的结构示意图；
图6是本申请另一实施例提供的一种终端设备的结构示意图；
图7是本申请实施例提供的一种计算机可读存储介质的结构示意图。
 本发明的实施方式
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的系统、装置、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。
应当理解，在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。另外，在本申请说明书和所附权利要求书的描述中，术语“第一”、“第二”、“第三”等仅用于区分描述，而不能理解为指示或暗示相对重要性。
还应当理解，在本申请说明书中描述的参考“一个实施例”或“一些实施例”等意味着在本申请的一个或多个实施例中包括结合该实施例描述的特定特征、结构或特点。由此，在本说明书中的不同之处出现的语句“在一个实施例中”、“在一些实施例中”、“在其他一些实施例中”、“在另外一些实施例中”等不是必然都参考相同的实施例，而是意味着“一个或多个但不是所有的实施例”，除非是以其他方式另外特别强调。术语“包括”、“包含”、“具有”及它们的变形都意味着“包括但不限于”，除非是以其他方式另外特别强调。
弱光环境是指自然照明条件不佳，影响人们视觉的特定环境，通常把光照度在50流明以下的环境称为弱光环境。弱光环境下拍摄到的图像通常会存在亮度低、对比度低、噪声点多等问题。本申请实施例提供的一种图像处理方法，通过连续更新网络提取出不同尺度的特征图像，提取出全局信息和局部信息，能够保证恢复得到的正常光照图像不会出现细节缺失的现象，再通过色彩增彩网络来增强恢复图像的图像颜色和纹理，保证图像质量，解决了目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。
以下将对本申请实施例提供的图像处理方法进行详细的说明：
请参阅图1，图1是本申请实施例提供的一种图像处理方法的示意性流程图。本申请实施例提供的图像处理方法的执行主体为终端设备，终端设备可以是智能手机、平板电脑或可穿戴设备等移动终端，也可以是各种应用场景下的电脑、云服务器、辅助计算机等。
如图1所示的图像处理方法可以包括S11~S12，详述如下：
S11：将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像。
在本申请实施例中，通过连续更新网络提取低光照图像中的局部特征和全局特征，然后将局部特征和全局特征进行特征融合，得到不同尺度的特征图像。
在本申请实施例中，上述连续更新网络可以包括N个U型网络结构，通过U型网络提取出不同尺度的图像特征，上述连续更新网络中还包括了特征融合结构，通过特征融合结构将U型网络提取出的不同尺度的图像特征进行特征融合，最终得到不同尺度的特征图像。
需要说明的是，N为大于等于2的正整数。
在本申请一实施例中，上述连续更新网络可以包括4个U型网络结构，每个U型网络结构的结构可以参见图2，图2示出了本申请实施例中连续更新网络的U型网络结构的结构示意图（以两个U型网络进行示意）。
如图2所示，上述U型网络结构可以包括下采样结构21和上采样结构22，上述下采样结构包括四个不同尺度的特征块，四个不同尺度的特征块（如图2中的A1、A2、A3、A4）的尺度比例由高到低可以是8:4:2:1，即两个相邻特征块相同规模的卷积的步长为1，两个相邻的特性与减少尺度卷积的步长为2。上采样结构同样包括四个不同尺度的特征块，四个不同尺度的特征块（如图2中的B1、B2、B3、B4）的尺度比例由低到高可以是1:2:4:8，即两个相邻特征块相同规模的卷积的步长为1，两个相邻特征块尺度得到的增加反褶积的步长为2。
不同特征块之间使用跳过连接（skip connection）进行连接，具体地，通过将下采样结构获取到的不同尺度特征与上采样结算获得的相同尺度特征直接合并，将上采样阶段获得的不同尺度特征通过上采样的方式与更大尺度的特征进行融合，通过跳过连接，上采样阶段可以同时获取全局特征和局部特征。
对于两个U型结构的组合，连接方式可以是，先将两个U型网络得到的四个不同尺度的特征放大一倍或两倍，然后通过特征融合结构（FF结构）将获得的四个相同尺度的特征进行融合，得到单个特征，得到的单个特征会被重新传输会U型网络中继续进行操作，这种连接方式使得第二个U型网络的特征能够获取到更多的全局特征和局部特征。
在本申请一实施例中，上述特征融合结构可以包括卷积层、平均池层和完全连通层。
具体的，请参阅图3，图3示出了本申请实施例提供的特征融合过程的流程示意图。如图像所示，在具体应用中，先将U型结构提取出的特征输入到特征融合模块中，特征融合模块中的卷积层进行处理，得到初始卷积，然后将初始卷积经过平均池层就会将特征进行压缩，继而得到单个特征值，然后通过两个完全连通层来确定不同信道之间的相关性，以得到权重值，然后将得到的权重值与初始卷积后得到的特征值相乘，得到第一加权特征，将第一加权特征加上初始卷积后得到的特征值，重复上述操作，就可以得到第二特加权特征值，最后，对第二加权特征进行压缩，就可以得到这四个特征融合后对应的特征图像。将不同尺度的特征都进行这样的操作，最后就可以得到不同尺度的特征图像。
在本申请一实施例中，上述不同尺度的特征图像包括第一尺度特征图像、第二尺度特征图像和第三尺度特征图像。
S12：将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
在本申请实施例中，色彩增强网络能够基于不同尺度的特征图像的残差来确定出每个特征图像的权重，之后使得每个特征图像与对应的权重相乘，合并出一张图像，即目标图像。
在本申请一实施例中，请参阅图4，图4是本申请实施例提供的色彩增强网络对不同尺度的特征图像进行处理的过程示意图。如图4所示，上述色彩增强网络400可以包括残差获取模块401、卷积模块402和权重确定模块403。
将第一尺度特征图像Y1、第二尺度特征图像Y2、第三特征尺度图像Y3输入到残差获取模块401中，就可以得到第一残差Y2-Y1、第二残差Y3-Y2以及第三残差Y3-Y1。然后将第一残差Y2-Y1、第二残差Y3-Y2以及第三残差Y3-Y1输入到卷积单元中进行处理，就可以得到第一特征D1和第二特征D2、将第一特征D1和第二特征D2进行卷积就得到第三特征D3、第四特征D4，然后通过权重确定模块将第二特征D2和第四特征D4进行融合，卷积得到第五特征D5，再将第五特征D5经过权重确定模块403进行权重分解，就得到得到第一权重G1、第二权重G2以及第三权重G3，将第一权重G1与第一尺度特征图像Y1相乘、将第二权重G2与第二尺度特征图像Y2相乘、将第三权重G3与第三尺度特征图像Y3相乘，然后将相乘得到的图像进行融合，就得到了目标图像。
在本申请另一实施例中，上述图像处理方法还可以包括以下步骤：
构建连续更新网络和色彩增强网络；
获取样本数据集；
基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
本申请实施例中，可以构建网络结构如图2所示的连续更新网络和如图4所示的色彩增强网络，然后利用样本数据对其进行训练，以得到训练完成的连续更新网络和色彩增强网络。
在具体应用中，上述样本数据可以包括历史低光照图像和对应的正常光照图像。
在实际应用中，可以选用不小于1000组的样本数据，得到样本数据集。将样本数据集分为训练集、验证集和测试集。为了达到训练要求，可以将50%的样本数据作为训练集，剩余部分作为验证集和测试集。
在获取到样本数据后，通过训练集数据对该连续更新网络和进行训练，并使用验证集进行快速调参，再使用测试集对连续更新网络进行测试，得到训练完成的连续更新网络。
在训练连续更新网络时，可以将样本图像中的历史低光照图像输入预先构建的连续更新网络进行处理，得到不同尺度的特征图像，然后基于正常光照图像和得到的不同尺度的特征图像的损失函数对连续更新网络中的网络参数进行调整，具体可以将第三尺度特征图像与对应的正常光照图像的结构相似性损失和全变分（Total Variation，TV）损失作为损失函数进行调整，当该损失函数收敛时，再基于验证集和测试集中的样本数据对调整完网络参数的连续更新网络进行验证和测试，验证和测试通过即说明该连续更新网络训练完成，训练完成而得连续更新网络就可以在S11中使用。
需要说明的是，由于在低光照条件下拍摄的图像通常会有明显的结构失真问题，为了定量和定性地提高图像的质量，故使用结构相似性（SSIM）损失，而低光照图像复原出来的正常光照图像可能具有不稳定的照明和噪声，因此使用全变分损失作为平滑度先验，以最小化整个图像的梯度。
在训练色彩增强网络时，可以将训练完成的连续更新网络输出的不同尺度的特征图像输入预先构建的色彩增强网络进行处理，得到目标图像，然后基于正常光照图像和得到的目标图像的损失函数对色彩增强网络中的网络参数进行调整，具体可以将目标图像与对应的正常光照图像的感知损失（perception loss）和全变分（Total Variation，TV）损失作为损失函数对网络参数进行调整，当该损失函数收敛时，再基于验证集和测试集中的样本数据对调整完网络参数的色彩增强网络进行验证和测试，验证和测试通过即说明色彩增强网络训练完成，训练完成而得色彩增强网络就可以在S12中使用。
需要说明的是上述感知损失包括特征重构损失（feature reconstruction loss）和风格重建损失（style reconstruction  loss），上述特征重构损失可以表示特征之间欧几里得距离，以此衡量特征的相似性，风格重建损失主要是为了得到更好地图像颜色和纹理，风格重建损失是输出图像（输出的目标图像）与正常光照图像的格拉姆矩阵只差的平方佛罗贝尼斯范数，最小化风格重建损失的图像保留了正常光照图像的风格特征，但是不保留其空间结构。
感知损失就是将特征重构损失和风格重建损失结合在一起的损失函数。而同样的，为了最小化整个图像的梯度，还基于全变分损失来构建色彩增强网络的损失函数。
以上可以看出，本申请实施例提供的图像处理方法，通过连续更新网络提取出不同尺度的特征图像，提取出全局信息和局部信息，能够保证恢复得到的正常光照图像不会出现细节缺失的现象，再通过色彩增彩网络来增强恢复图像的图像颜色和纹理，保证图像质量，解决了目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
基于上述实施例所提供的图像处理方法，本发明实施例进一步给出实现上述方法实施例的终端设备的实施例。
请参阅图5，图5是本申请实施例提供的一种终端设备的结构示意图。本申请实施例中，终端设备包括的各单元用于执行图1至图4对应的实施例中的各步骤。具体请参阅图1至图4以及图1至图4对应的实施例中的相关描述。为了便于说明，仅示出了与本实施例相关的部分。如图5所示，终端设备50包括：特征提取模块51和色彩增强模块52。其中：
特征提取模块51用于将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像。
色彩增强模块52用于将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
可选的，所述终端设备还包括：
构建模块，用于构建连续更新网络和色彩增强网络；
获取模块，用于获取样本数据集；
训练模块，用于基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
可选的，连续更新网络包括N个U型网络结构和特征融合结构，其中，N为大于等于2的正整数。
可选的，所述U型网络结构包括下采样结构和上采样结构，所述下采样结构包括四个不同尺度的特征块，所述上采样结构包括四个不同尺度的特征块，不同特征块之间使用跳过连接进行连接。
可选地，所述特征融合结构包括卷积层、平均池层和完全连通层。
可选地，所述色彩增强网络包括残差获取模块、卷积模块和权重确定模块。
需要说明的是，上述模块之间的信息交互、执行过程等内容，由于与本申请方法实施例基于同一构思，其具体功能及带来的技术效果，具体可参照方法实施例部分，此处不再赘述。
基于此，本申请实施例提供的终端设备，同样能够通过连续更新网络提取出不同尺度的特征图像，提取出全局信息和局部信息，能够保证恢复得到的正常光照图像不会出现细节缺失的现象，再通过色彩增彩网络来增强恢复图像的图像颜色和纹理，保证图像质量，解决了目前对弱光条件下拍摄到的图像进行增亮的方法存在图像处理效果不佳的问题。
图6是本申请另一实施例提供的一种终端设备的结构示意图。如图6所示，该实施例提供的终端设备6包括：处理器60、存储器61以及存储在所述存储器61中并可在所述处理器60上运行的计算机程序62，例如图像处理程序。处理器60执行所述计算机程序62时实现上述各个图像处理方法实施例中的步骤，例如图1所示的S11~S12。或者，所述处理器60执行所述计算机程序62时实现上述各终端设备实施例中各模块/单元的功能，例如图5所示单元51~52的功能。
示例性的，所述计算机程序62可以被分割成一个或多个模块/单元，所述一个或者多个模块/单元被存储在所述存储器61中，并由处理器60执行，以完成本申请。所述一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述所述计算机程序62在所述终端设备6中的执行过程。例如，所述计算机程序62可以被分割成第一获取单元和第一处理单元，各单元具体功能请参阅图5对应地实施例中的相关描述，此处不赘述。
所述终端设备可包括但不仅限于，处理器60、存储器61。本领域技术人员可以理解，图6仅仅是终端设备6的示例，并不构成对终端设备6的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述终端设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器60可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器 (Digital Signal Processor，DSP)、专用集成电路 (Application Specific Integrated Circuit，ASIC)、现成可编程门阵列 (Field-Programmable Gate Array，FPGA) 或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器61可以是所述终端设备6的内部存储单元，例如终端设备6的硬盘或内存。所述存储器61也可以是所述终端设备6的外部存储设备，例如所述终端设备6上配备的插接式硬盘，智能存储卡（Smart Media Card, SMC），安全数字（Secure Digital, SD）卡，闪存卡（Flash Card）等。进一步地，所述存储器61还可以既包括所述终端设备6的内部存储单元也包括外部存储设备。所述存储器61用于存储所述计算机程序以及所述终端设备所需的其他程序和数据。所述存储器61还可以用于暂时地存储已经输出或者将要输出的数据。
本申请实施例还提供了一种计算机可读存储介质。请参阅图7，图7是本申请实施例提供的一种计算机可读存储介质的结构示意图，如图7所示，计算机可读存储介质70中存储有计算机程序71，计算机程序71被处理器执行时可实现上述图像处理方法。
本申请实施例提供了一种计算机程序产品，当计算机程序产品在终端设备上运行时，使得终端设备执行时实现可实现上述图像处理方法。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述终端设备的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中，上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参照其它实施例的相关描述。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围，均应包含在本申请的保护范围之内。

Claims (17)

  1. 一种图像处理方法，其特征在于，包括：
    将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像；
    将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
  2. 根据权利要求1所述的图像处理方法，其特征在于，还包括：
    构建连续更新网络和色彩增强网络；
    获取样本数据集；
    基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
  3. 根据权利要求1所述的图像处理方法，其特征在于，连续更新网络包括N个U型网络结构和特征融合结构，其中，N为大于等于2的正整数。
  4. 根据权利要求3所述的图像处理方法，其特征在于，所述U型网络结构包括下采样结构和上采样结构，所述下采样结构包括四个不同尺度的特征块，所述上采样结构包括四个不同尺度的特征块，不同特征块之间使用跳过连接进行连接。
  5. 根据权利要求3所述的图像处理方法，其特征在于，所述特征融合结构包括卷积层、平均池层和完全连通层。
  6. 根据权利要求1所述的图像处理方法，其特征在于，所述色彩增强网络包括残差获取模块、卷积模块和权重确定模块。
  7. 根据权利要求1至6任意一项所述的图像处理方法，其特征在于，所述不同尺度的特征图像包括第一尺度特征图像、第二尺度特征图像以及第三特征尺度图像。
  8. 一种终端设备，其特征在于，包括：
    特征提取模块，用于将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像；
    色彩增强模块，用于将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
  9. 如权利要求8所述的终端设备，其特征在于，所述终端设备还包括：
    构建模块，用于构建连续更新网络和色彩增强网络；
    获取模块，用于获取样本数据集；
    训练模块，用于基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
  10. 如权利要求8所述的终端设备，其特征在于，连续更新网络包括N个U型网络结构和特征融合结构，其中，N为大于等于2的正整数。
  11. 如权利要求10 所述的终端设备，其特征在于，所述U型网络结构包括下采样结构和上采样结构，所述下采样结构包括四个不同尺度的特征块，所述上采样结构包括四个不同尺度的特征块，不同特征块之间使用跳过连接进行连接。
  12. 如权利要求10所述的终端设备，其特征在于，所述特征融合结构包括卷积层、平均池层和完全连通层。
  13. 如权利要求8所述的终端设备，其特征在于，所述色彩增强网络包括残差获取模块、卷积模块和权重确定模块。
  14. 一种终端设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机可读指令时实现如下步骤：
    将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像；
    将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
  15. 如权利要求14所述的终端设备，其特征在于，所述处理器执行所述计算机可读指令时还实现如下步骤：
    构建连续更新网络和色彩增强网络；
    获取样本数据集；
    基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
  16. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其特征在于，所述计算机可读指令被处理器执行时实现如下步骤：
    将低光照图像输入至已完成训练的连续更新网络中进行局部特征和全局特征提取，得到不同尺度的特征图像；
    将所述不同尺度的特征图像输入至已完成训练的色彩增强网络中进行处理，得到目标图像。
  17. 如权利要求16所述的计算机可读存储介质，其特征在于，还包括：
    构建连续更新网络和色彩增强网络；
    获取样本数据集；
    基于样本数据集对所述连续更新网络和色彩增强网络进行训练，得到所述完成训练的连续更新网络和完成训练的色彩增强网络。
PCT/CN2021/133130 2021-11-25 2021-11-25 一种图像处理方法、终端设备及计算机可读存储介质 WO2023092386A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/133130 WO2023092386A1 (zh) 2021-11-25 2021-11-25 一种图像处理方法、终端设备及计算机可读存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/133130 WO2023092386A1 (zh) 2021-11-25 2021-11-25 一种图像处理方法、终端设备及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2023092386A1 true WO2023092386A1 (zh) 2023-06-01

Family

ID=86538510

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/133130 WO2023092386A1 (zh) 2021-11-25 2021-11-25 一种图像处理方法、终端设备及计算机可读存储介质

Country Status (1)

Country Link
WO (1) WO2023092386A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116758617A (zh) * 2023-08-16 2023-09-15 四川信息职业技术学院 一种低光照度场景下的校园学生签到方法和校园签到系统
CN117391995A (zh) * 2023-10-18 2024-01-12 山东财经大学 渐进式人脸图像复原方法、系统、设备及存储介质
CN117522742A (zh) * 2024-01-04 2024-02-06 深圳市欧冶半导体有限公司 图像处理方法、架构、装置和计算机设备

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111986120A (zh) * 2020-09-15 2020-11-24 天津师范大学 一种基于帧累加和多尺度Retinex的低光照图像增强优化方法
CN112001863A (zh) * 2020-08-28 2020-11-27 太原科技大学 一种基于深度学习的欠曝光图像恢复方法
CN112019827A (zh) * 2020-09-02 2020-12-01 上海网达软件股份有限公司 视频图像色彩增强的方法、装置、设备及存储介质
CN112508815A (zh) * 2020-12-09 2021-03-16 中国科学院深圳先进技术研究院 模型的训练方法和装置、电子设备、机器可读存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112001863A (zh) * 2020-08-28 2020-11-27 太原科技大学 一种基于深度学习的欠曝光图像恢复方法
CN112019827A (zh) * 2020-09-02 2020-12-01 上海网达软件股份有限公司 视频图像色彩增强的方法、装置、设备及存储介质
CN111986120A (zh) * 2020-09-15 2020-11-24 天津师范大学 一种基于帧累加和多尺度Retinex的低光照图像增强优化方法
CN112508815A (zh) * 2020-12-09 2021-03-16 中国科学院深圳先进技术研究院 模型的训练方法和装置、电子设备、机器可读存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CHEN MINGSONG; OUYANG ZUWEI; ZHANG QIESHI; REN ZILIANG; CHENG JUN; YUAN SHUAI: "BEN: Brightness Enhancement Network for Low-Light Image Enhancement in Complex Environment", 2021 INTERNATIONAL CONFERENCE ON SENSING, MEASUREMENT & DATA ANALYTICS IN THE ERA OF ARTIFICIAL INTELLIGENCE (ICSMD), 21 October 2021 (2021-10-21), pages 1 - 5, XP034064504, DOI: 10.1109/ICSMD53520.2021.9670842 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116758617A (zh) * 2023-08-16 2023-09-15 四川信息职业技术学院 一种低光照度场景下的校园学生签到方法和校园签到系统
CN116758617B (zh) * 2023-08-16 2023-11-10 四川信息职业技术学院 一种低光照度场景下的校园学生签到方法和校园签到系统
CN117391995A (zh) * 2023-10-18 2024-01-12 山东财经大学 渐进式人脸图像复原方法、系统、设备及存储介质
CN117522742A (zh) * 2024-01-04 2024-02-06 深圳市欧冶半导体有限公司 图像处理方法、架构、装置和计算机设备
CN117522742B (zh) * 2024-01-04 2024-03-22 深圳市欧冶半导体有限公司 图像处理方法、架构、装置和计算机设备

Similar Documents

Publication Publication Date Title
WO2023092386A1 (zh) 一种图像处理方法、终端设备及计算机可读存储介质
US11373275B2 (en) Method for generating high-resolution picture, computer device, and storage medium
CN110197229B (zh) 图像处理模型的训练方法、装置及存储介质
US11037278B2 (en) Systems and methods for transforming raw sensor data captured in low-light conditions to well-exposed images using neural network architectures
CN108876792B (zh) 语义分割方法、装置和系统及存储介质
WO2022105125A1 (zh) 图像分割方法、装置、计算机设备及存储介质
CN110163801B (zh) 一种图像超分辨和着色方法、系统及电子设备
US20200252555A1 (en) Method for processing image data by using information generated from external electronic device, and electronic device
WO2022022288A1 (zh) 一种图像处理方法以及装置
JP6978542B2 (ja) 電子装置及びその制御方法
CN112348747A (zh) 图像增强方法、装置及存储介质
US11972543B2 (en) Method and terminal for improving color quality of images
CN109977832B (zh) 一种图像处理方法、装置及存储介质
CN112906721B (zh) 图像处理方法、装置、设备及计算机可读存储介质
WO2023284401A1 (zh) 图像美颜处理方法、装置、存储介质与电子设备
CN113939845A (zh) 提高图像色彩质量的方法、系统和计算机可读介质
CN112508812A (zh) 图像色偏校正方法、模型训练方法、装置及设备
Liu et al. Learning noise-decoupled affine models for extreme low-light image enhancement
CN115601820A (zh) 一种人脸伪造图像检测方法、装置、终端及存储介质
Song et al. Multi-scale joint network based on Retinex theory for low-light enhancement
CN114830168A (zh) 图像重建方法、电子设备和计算机可读存储介质
CN116645598A (zh) 一种基于通道注意力特征融合的遥感图像语义分割方法
WO2024041108A1 (zh) 图像矫正模型训练及图像矫正方法、装置和计算机设备
CN114219725A (zh) 一种图像处理方法、终端设备及计算机可读存储介质
WO2023010981A1 (zh) 编解码方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21965122

Country of ref document: EP

Kind code of ref document: A1