Image management method, system and device for removing illumination influence and storage medium

Abstract

The invention relates to the technical field of image processing, and particularly provides an image management method, system and device for removing illumination influence, and a storage medium, and the method comprises the steps: carrying out the color correction of a to-be-processed image through a color constancy convolution algorithm, and obtaining a standard image; extracting a feature value of the standard image by using a deep learning neural network model; calculating the similarity between the feature value and a feature value of an image stored in a database; if the maximum similarity reaches a set threshold value, it is judged that an image matched with the to-be-processed image exists in the database; and if the maximum similarity does not reach a set threshold value, storing the characteristic value of the standard image, and storing the to-be-processed image to the database. The repeated images can be accurately recognized, and the repeated images are effectively prevented from occupying a large number of storage resources.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and in particular relates to an image management method, system, device and storage medium for removing illumination influence. Background Art

[0002] Image storage and search technology is widely used in modern life, whether it is personal photo management or large-scale image database retrieval, it is inseparable from the support of this technology. However, most current image search technologies focus on precise search of the same image, and are unable to search video images of objects shot at the same location at different time points.

[0003] Especially when there are large differences in lighting conditions, the image features of the same object may change significantly, resulting in a significant drop in search accuracy. Such lighting changes not only affect the distribution of light and dark in the image, but may also distort the shape and color of the object, making repeated images difficult to recognize under different lighting conditions.

[0004] This limitation not only reduces the practicality of image search, but also brings another serious problem: the database is filled with a large number of duplicate images. These images are stored repeatedly because they are difficult to be effectively identified as duplicate content, thus greatly wasting precious storage resources. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides an image management method, system, device and storage medium for removing the influence of illumination, so as to solve the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides an image management method for removing the influence of illumination, comprising: The color constant convolution algorithm is used to perform color correction on the image to be processed to obtain a standard image; Extracting feature values ​​of the standard image using a deep learning neural network model; Calculating the similarity between the characteristic value and the characteristic value of the image stored in the database; If the maximum similarity reaches a set threshold, it is determined that there is an image in the database that matches the image to be processed; If the maximum similarity does not reach the set threshold, the feature value of the standard image is saved, and the image to be processed is saved in the database.

[0007] In an optional implementation, color correction is performed on the image to be processed using a color constancy convolution algorithm to obtain a standard image, including: intercepting a request to save an image to a database, and extracting the image to be processed from the request; Preprocessing the image to be processed, wherein the preprocessing includes scaling the image to a specified size and performing a normalization operation; Construct a histogram for the preprocessed image to be processed, count the pixels of the histogram and calculate the position of each pixel in the UV color space; Use a convolutional neural network to estimate the illumination based on the position of each pixel in UV color space; Color correction is performed on the image to be processed according to the illumination.

[0008] In an optional implementation, calculating the similarity between the feature value and the feature value of the image stored in the database includes: Pre-defining a vector set in a vector database, constructing feature values ​​of the image and metadata of the image stored in the database into a vector index, and inserting the vector index into the vector set; A nearest neighbor search algorithm is used to search the vector set for a vector index that is closest to the characteristic value of the standard image, and the corresponding Euclidean distance similarity is output.

[0009] In an optional implementation, saving the characteristic value of the standard image and saving the image to be processed to the database includes: Saving the image to be processed to the database, and acquiring metadata of the image to be processed; Saving the metadata and the feature values ​​of the standard image as new vector indexes; Inserting the new vector index into the vector set; Use a unique key to establish a connection between the vector database and the image database.

[0010] In a second aspect, the present invention provides an image management system for removing the influence of illumination, comprising: A color correction module, used to perform color correction on the image to be processed by using a color constancy convolution algorithm to obtain a standard image; A feature extraction module, used to extract feature values ​​of the standard image using a deep learning neural network model; A similarity search module, used for calculating the similarity between the feature value and the feature value of the image stored in the database; A repeated determination module, used for determining that there is an image in the database that matches the image to be processed if the maximum similarity reaches a set threshold; The image storage module is used to save the feature value of the standard image and save the image to be processed into the database if the maximum similarity does not reach the set threshold.

[0011] In an optional implementation, the color correction module includes: An image acquisition unit, used for intercepting a request to save an image to a database, and extracting the image to be processed from the request; A preprocessing unit, used for preprocessing the image to be processed, wherein the preprocessing includes scaling the image to a specified size and performing a normalization operation; A pixel processing unit, used for constructing a histogram for the preprocessed image to be processed, counting the pixels of the histogram and calculating the position of each pixel in the UV color space; An illumination estimation unit, for estimating illumination based on the position of each pixel in UV color space using a convolutional neural network; A color correction unit is used to perform color correction on the image to be processed according to the illumination.

[0012] In an optional implementation, the similarity search module includes: A vector pre-storage unit, used to pre-define a vector set in a vector database, construct a feature value of an image and metadata of the image stored in the database into a vector index, and insert the vector index into the vector set; The vector search unit is used to search the vector index closest to the characteristic value of the standard image from the vector set by using the nearest neighbor search algorithm, and output the corresponding Euclidean distance similarity.

[0013] In an optional implementation, the image storage module includes: An image storage unit, used for storing the image to be processed in the database and obtaining metadata of the image to be processed; An index building unit, used for saving the metadata and the feature value of the standard image as a new vector index; An index inserting unit, used to insert the new vector index into the vector set; The relationship establishment unit is used to establish a connection relationship between the vector database and the image database using a unique key.

[0014] In a third aspect, a device is provided, comprising: A memory for storing an image management program for removing the influence of illumination; The processor is used to implement the steps of the image management method for removing the influence of lighting as provided in the first aspect when executing the image management program for removing the influence of lighting.

[0015] In a fourth aspect, a computer-readable storage medium is provided, on which is stored an image management program for removing the influence of lighting. When the image management program for removing the influence of lighting is executed by a processor, the steps of the image management method for removing the influence of lighting provided in the first aspect are implemented.

[0016] The beneficial effect of the present invention is that the image management method, system, device and storage medium for removing the influence of lighting provided by the present invention, combined with the color constancy convolution algorithm and the deep learning neural network, can accurately extract the real features of the image, eliminate lighting interference, and then accurately identify duplicate images, effectively avoiding duplicate images occupying a large amount of storage resources.

[0017] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention.

[0020] Figure 2 is another schematic flow chart of a method according to an embodiment of the present invention.

[0021] Figure 3 The figure is a color correction effect diagram of a method according to an embodiment of the present invention.

[0022] Figure 4 It is a schematic architecture diagram of a vector database of a method according to an embodiment of the present invention.

[0023] Figure 5 is a schematic block diagram of a system according to an embodiment of the present invention.

[0024] Figure 6 A schematic diagram of the structure of a device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0027] The key terms appearing in the present invention are explained below.

[0028] 1. Color constancy technology is an important research direction in the field of computer vision and image processing. It aims to simulate the human visual system's ability to perceive color, that is, to maintain stable recognition of object colors under different lighting conditions, solve the problem that image colors are affected by lighting changes, and enable accurate recognition and processing of colors in images under different lighting conditions. Among them, color constancy methods based on deep neural networks (DNNs) have made significant progress. These methods jointly learn image representations and estimation functions, and can automatically learn complex patterns of color constancy from large amounts of data.

[0029] 2. Neural network feature extraction technology is a method used in the field of deep learning to automatically extract useful features from raw data. Among them, CNN is a method specifically used to extract image features. It extracts local to global features of an image through structures such as convolutional layers, pooling layers, and fully connected layers. The convolutional layer uses convolution kernels to perform convolution operations on the image to extract features such as edges and textures in the image. The pooling layer downsamples the feature map to reduce the data dimension and improve computational efficiency.

[0030] 3. Feature vector database storage technology is a technology used to efficiently store, manage and retrieve high-dimensional vector data, which involves vector storage, similarity calculation, nearest neighbor search, index construction and distributed management.

[0031] 4. Feature vector query technology is the basic function of vector database. Feature vector query technology is the key technology for efficiently retrieving similar vectors in vector database. Commonly used nearest neighbor search algorithm, approximate nearest neighbor search algorithm, approximate nearest neighbor search algorithm based on small world graph theory, hash approximate nearest neighbor search algorithm, etc., can be used to efficiently process and retrieve large-scale high-dimensional vector data through vector database.

[0032] The image management method for removing the influence of illumination provided by the embodiment of the present invention is executed by a computer device. Accordingly, the image management system for removing the influence of illumination runs in the computer device.

[0033] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention. Figure 1The execution subject may be an image management system that removes the influence of illumination. According to different requirements, the order of the steps in the flow chart may be changed, and some may be omitted.

[0034] like Figure 1 As shown, the method includes: S1. Use the color constancy convolution algorithm to perform color correction on the image to be processed to obtain a standard image.

[0035] The color constancy convolution algorithm is used to perform color correction on the processed image. The purpose of this step is to eliminate the impact of changes in the lighting environment on the image color, so that the images of the same object taken under different lighting conditions can show consistent color characteristics. Through the processing of this algorithm, we can obtain a color-standardized image, that is, a standard image, which lays a good foundation for subsequent feature extraction and matching.

[0036] S2. Using a deep learning neural network model to extract feature values ​​of the standard image.

[0037] The deep learning neural network model is used to extract feature values ​​from the color-corrected standard image. After extensive training, this neural network model can accurately capture key information in the image, such as the shape, texture, and edges of the object, and encode this information into a series of high-dimensional feature vectors. These feature vectors can effectively represent the content of the image and provide a reliable basis for subsequent image matching.

[0038] S3. Calculate the similarity between the feature value and the feature value of the image stored in the database.

[0039] After extracting the eigenvalues ​​of the standard image, we compare it with the eigenvalues ​​of all images stored in the database one by one and calculate the similarity between them. The similarity calculation method can use common metrics such as cosine similarity and Euclidean distance. The specific choice depends on the actual application scenario and algorithm performance requirements.

[0040] S4. If the maximum similarity reaches a set threshold, it is determined that there is an image in the database that matches the image to be processed.

[0041] By comparison, if the maximum similarity reaches the set threshold, then we can determine that there is an image in the database that matches the image to be processed. This threshold is usually set according to the needs of actual applications to ensure the accuracy and efficiency of matching.

[0042] S5. If the maximum similarity does not reach the set threshold, the feature value of the standard image is saved, and the image to be processed is saved in the database.

[0043] If the maximum similarity does not reach the set threshold, it means that there is no image in the database that completely matches the image to be processed. At this point, we need to save the feature values ​​of the standard image to the database, and also save the image to be processed itself to the database for more extensive searches and matches in the future. This step not only enriches the content of the database, but also improves the flexibility and adaptability of the image search system.

[0044] Please refer to Figure 2 To facilitate understanding of the present invention, the following further describes the image management method for removing the influence of illumination provided by the present invention based on the principle of the image management method for removing the influence of illumination and the process of managing the image in the embodiment.

[0045] In an embodiment of the present invention, based on step S1, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0046] S101. Before processing the image data and preparing to save it to the database, we first intercept the request to save the image to the database and accurately extract the image to be processed from the request. This step is the basis for ensuring that the subsequent processing flow can proceed smoothly.

[0047] S102. After extracting the image to be processed, we immediately perform a series of preprocessing operations on it. These preprocessing steps are designed to ensure that the image can maintain consistency in subsequent processing and improve the efficiency and accuracy of the algorithm. Specifically, the preprocessing includes scaling the image to a specified size, which helps to reduce the calculation error caused by different image sizes. At the same time, we also normalize the image so that the pixel values ​​of the image are distributed in a reasonable range for subsequent processing.

[0048] S103. After completing the preprocessing, we start to build a histogram for the image to be processed. Convert the image RGB to YUV, calculate the feature histogram M, count the number of pixels close to the color (u, v) in the histogram M, set N bins in the histogram, N can be selected as 64, 128 or 256, the width of the bins is 0.025 by default, traverse each pixel in the image, and calculate the position of each pixel in the UV color space.

[0049] The whole process can be understood as inputting the original image I and calculating the white balance image W by estimating the value of L. Therefore, the key to solving the problem lies in how to estimate the value of illuminance L. The relationship between the white balance image and illuminance is: , Where I represents the RGB pixel value of the input image, W represents the pixel value after white balance, and L represents the brightness information of the image.

[0050] The function relationship between RGB and UV value is: , Convert RGB pixel values ​​to UV values ​​through the function corresponding to RGB to UV values. UV values ​​are usually used to represent a point in the color space. Compared with RGB, UV values ​​can better reflect the brightness information of the color. Here, Iu and Iv represent the pixel values ​​of the U and V channels respectively. It can be known that Ig, Ir, and Ib correspond to the three channel pixel values ​​of the original image, and Iu and Iv can be obtained: , Obtain Iy, convert RGB to UV pixel values, and further express Lu and Lv in the following way: ; By performing log transformation on the relationship between the white balance image and illumination, we can get the white balance solution formula: ; Convert the final problem to Lu and Lv. Since the absolute ratio is unknown, the mapping of RGB to UV is uncertain. Assuming L is a unit norm, Lr, Lg, and Lb can be represented by Lu and Lv.

[0051] ; On the other hand, by constructing a histogram M of the number of pixels close to (u, v), as the scoring standard, the number of pixels close to (u, v) is counted, where ε is the bin width, the default is 0.025, and the histogram has 256 bins: This histogram can be used as a scoring standard to evaluate the white balance effect.

[0052] S104. Next, we use a convolutional neural network (CNN) to estimate the illumination of an image based on the position of each pixel in the UV color space. A convolutional neural network is a deep learning model with powerful feature extraction and pattern recognition capabilities. Through training, it can accurately estimate the illumination value of an image based on the UV color information of the pixel.

[0053] The constructed histogram M is convolved with the convolution kernel F, and the Lu and Lv with the highest scores are used to generate a white balanced image. The purpose of this step is to smooth the histogram and reduce noise, making it easier to find the optimal Lu and Lv values.

[0054] S105. Finally, we perform color correction on the image to be processed based on the estimated illumination value. Color correction is a commonly used image processing technique that aims to eliminate image color distortion caused by illumination changes. By adjusting parameters such as image brightness, contrast, and color balance, we can make the image present a consistent color effect under different illumination conditions.

[0055] Use the highest-scoring Lu and Lv values ​​to generate the white-balanced image W. This image should be more natural in color, with white objects appearing white and other colors retaining natural tones. Figure 3 shown.

[0056] In an embodiment of the present invention, based on step S2, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0057] When dealing with image feature extraction tasks, we used Efficientnet-B5 as the feature network to extract the eigenvalues ​​of the image. The Efficientnet series of models are known for their efficient computing performance and excellent classification accuracy. The B5 version has achieved outstanding performance in multiple image classification tasks with its deep network structure and optimized parameter configuration.

[0058] First, we applied a color constancy algorithm to the processed image. The color constancy algorithm is an image processing technique that aims to eliminate color distortion caused by lighting changes in the image, so that the image can maintain consistent color effects under different lighting conditions. This step is crucial for subsequent feature extraction and image encoding because it ensures that the image input to the Efficientnet-B5 model has stable color features.

[0059] In order to ensure the efficiency of image encoding, we need to resize the image before inputting it into the Efficientnet-B5 model. Specifically, we resize the image to 224*224 pixels. This size selection is based on the input requirements of the Efficientnet-B5 model, and is also to minimize the amount of calculation and increase the processing speed while ensuring the integrity of the image information.

[0060] We then use the Efficientnet-B5 pre-trained model to extract features from the resized images. Since the Efficientnet-B5 model has been fully trained on a large-scale image dataset, it has high classification accuracy and generalization ability. This allows us to use the model to extract feature vectors with rich information from the image.

[0061] After being processed by the Efficientnet-B5 model, we encode the image into a 1*1024-dimensional vector. This vector contains the main feature information of the image and can be used for subsequent image classification, retrieval and other tasks. Since the vector dimension is moderate, it contains enough information while avoiding excessive computational complexity, so it has good performance and efficiency in practical applications.

[0062] In summary, by using the Efficientnet-B5 feature network, color constancy algorithm to process images, resize images, and encode images into 1*1024 dimensional vectors, we can effectively extract the eigenvalues ​​of images and provide strong support for subsequent image processing tasks.

[0063] In one embodiment of the present invention, based on step S3, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner. Please refer to Figure 4 .

[0064] First, we created a vector database instance using Milvus. In this instance, we pre-defined a vector collection to store image feature vectors and related metadata. This vector collection is the core data structure of the image retrieval system, which allows us to efficiently store, index, and retrieve image feature vectors.

[0065] Next, we extract the feature values ​​of the stored images from the database. These feature values ​​are extracted by the Efficientnet-B5 feature network mentioned above. At the same time, we also extract the metadata of the image, such as the image ID, file name, shooting time, etc. This information is crucial for subsequent image retrieval and display.

[0066] We then combined these feature values ​​and metadata to construct a vector index. The vector index is a key data structure used in Milvus to store and retrieve vector data, which allows us to quickly search based on the similarity between vectors. When building the vector index, we chose an index type that suits our application scenario, such as HNSW (Hierarchical Navigable Small World graphs) or IVF (Inverted File), to optimize search performance and accuracy.

[0067] After completing the construction of the vector index, we insert these indexes into the previously defined vector set. In this way, we have successfully stored the image's feature values ​​and metadata in the Milvus vector database, preparing for subsequent image retrieval.

[0068] In the image retrieval stage, we first extract feature values ​​from the standard image to be retrieved, and then use the nearest neighbor search algorithm (such as Euclidean distance, cosine similarity, etc.) to search for the vector index closest to the feature value of the standard image in the vector set of the Milvus vector database. The nearest neighbor search algorithm is a commonly used similarity search method that can quickly find the vector that is most similar to the target vector in a large amount of vector data.

[0069] Finally, we output the corresponding Euclidean distance similarity based on the searched vector index. Euclidean distance similarity is an indicator that measures the similarity between two vectors. The smaller it is, the more similar the two vectors are. By comparing the Euclidean distance similarity, we can determine which images are closest to the standard image, thereby completing the image retrieval task.

[0070] In an embodiment of the present invention, based on step S4, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0071] If the maximum similarity reaches the set threshold, metadata is extracted from the searched adjacent indexes, and the corresponding image is queried from the database based on the metadata. If the image is successfully obtained, it means that the metadata is valid, and it is considered that there is an image identical to the image to be processed in the database, and the current image to be processed is ignored; if the corresponding image cannot be obtained from the database, it means that the metadata is invalid, the vector index is deleted from the vector database, and the image to be processed is stored in the database.

[0072] In an embodiment of the present invention, based on step S5, a possible embodiment is given below to illustrate its specific implementation in a non-limiting manner.

[0073] In order to establish a connection between the vector database and the image database, we use a unique key to associate the two. This key can be the image ID, file name, or other unique identifier. With this key, we can retrieve the vector index of the image from the vector database when needed, and get the corresponding image data from the image database. This connection method ensures the consistency between the vector index and the image data, allowing the image retrieval system to work accurately and efficiently.

[0074] Save the images to be processed into an image database. This database can be a relational database, such as MySQL, or a non-relational database, such as MongoDB. The specific choice depends on the needs and performance requirements of the system. While saving the images, we also need to obtain and store the metadata of these images. Metadata usually includes the image ID, file name, shooting time, location, device information, etc. This information is crucial for subsequent image retrieval and management.

[0075] Combine the metadata of the image to be processed with the feature values ​​of the standard image to generate a new vector index. This vector index contains not only the feature information of the image, but also the metadata of the image, so that in the subsequent retrieval process, both the content and attributes of the image can be used for search. After completing the construction of the new vector index, we insert it into the previously defined vector set.

[0076] In some embodiments, the image management system for removing the influence of illumination may include a plurality of functional modules composed of computer program segments. The computer programs of the various program segments in the image management system for removing the influence of illumination may be stored in a memory of a computer device and executed by at least one processor to perform (see Figure 1 Description) Image management functions to remove the effects of lighting.

[0077] In this embodiment, the image management system for removing the influence of illumination can be divided into multiple functional modules according to the functions it performs, such as Figure 5 As shown. The functional modules of the system 500 may include: a color correction module 510, a feature extraction module 520, a similarity search module 530, a duplication determination module 540 and an image storage module 550. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0078] A color correction module, used to perform color correction on the image to be processed by using a color constancy convolution algorithm to obtain a standard image; A feature extraction module, used to extract feature values ​​of the standard image using a deep learning neural network model; A similarity search module, used for calculating the similarity between the feature value and the feature value of the image stored in the database; A repeated determination module, used for determining that there is an image in the database that matches the image to be processed if the maximum similarity reaches a set threshold; The image storage module is used to save the feature value of the standard image and save the image to be processed into the database if the maximum similarity does not reach the set threshold.

[0079] Optionally, as an embodiment of the present invention, the color correction module includes: An image acquisition unit, used for intercepting a request to save an image to a database, and extracting the image to be processed from the request; A preprocessing unit, used for preprocessing the image to be processed, wherein the preprocessing includes scaling the image to a specified size and performing a normalization operation; A pixel processing unit, used for constructing a histogram for the preprocessed image to be processed, counting the pixels of the histogram and calculating the position of each pixel in the UV color space; An illumination estimation unit, for estimating illumination based on the position of each pixel in UV color space using a convolutional neural network; A color correction unit is used to perform color correction on the image to be processed according to the illumination.

[0080] Optionally, as an embodiment of the present invention, the similarity search module includes: A vector pre-storage unit, used to pre-define a vector set in a vector database, construct a feature value of an image and metadata of the image stored in the database into a vector index, and insert the vector index into the vector set; The vector search unit is used to search the vector index closest to the characteristic value of the standard image from the vector set by using the nearest neighbor search algorithm, and output the corresponding Euclidean distance similarity.

[0081] Optionally, as an embodiment of the present invention, the image storage module includes: An image storage unit, used for storing the image to be processed in the database and obtaining metadata of the image to be processed; An index building unit, used for saving the metadata and the feature value of the standard image as a new vector index; An index inserting unit, used to insert the new vector index into the vector set; The relationship establishment unit is used to establish a connection relationship between the vector database and the image database using a unique key.

[0082] Figure 6 The image management method for removing the influence of illumination provided in the embodiment of the present application can be applied to the device. It can be understood by those skilled in the art that the device structure involved in the embodiment of the present invention does not constitute a limitation on the device, and the device may include more or less components than shown in the figure, or combine certain components, or arrange different components. In the embodiment of the present invention, the device includes but is not limited to a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the embodiments of the present application described and / or required herein.

[0083] The device 600 may include: a processor 610, a memory 620 and a communication unit 630. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention, and it may be a bus structure or a star structure, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0084] The memory 620 may be used to store the execution instructions of the processor 610, and the memory 620 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. When the execution instructions in the memory 620 are executed by the processor 610, the device 600 is enabled to perform some or all of the steps in the following method embodiments.

[0085] The processor 610 is the control center of the storage device, and uses various interfaces and lines to connect various parts of the entire electronic device. It runs or executes software programs and / or modules stored in the memory 620, and calls data stored in the memory to perform various functions of the electronic device and / or process data. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of a plurality of packaged ICs with the same or different functions. For example, the processor 610 can include only a central processing unit (CPU). In an embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0086] The communication unit 630 is used to establish a communication channel so that the storage device can communicate with other devices, receive user data sent by other devices or send user data to other devices.

[0087] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0088] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes, including several instructions for enabling a computer device (which can be a personal computer, a server, or a second device, a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0089] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0090] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or modules, which can be electrical, mechanical or other forms.

[0091] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0092] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0093] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. An image management method for removing the influence of illumination, characterized in that: include: The color constant convolution algorithm is used to perform color correction on the image to be processed to obtain a standard image; Extracting feature values ​​of the standard image using a deep learning neural network model; Calculating the similarity between the characteristic value and the characteristic value of the image stored in the database; If the maximum similarity reaches a set threshold, it is determined that there is an image in the database that matches the image to be processed; If the maximum similarity does not reach the set threshold, the feature value of the standard image is saved, and the image to be processed is saved in the database.

2. The method according to claim 1, characterized in that The color constant convolution algorithm is used to perform color correction on the image to be processed to obtain a standard image, including: intercepting a request to save an image to a database, and extracting the image to be processed from the request; Preprocessing the image to be processed, wherein the preprocessing includes scaling the image to a specified size and performing a normalization operation; Construct a histogram for the preprocessed image to be processed, count the pixels of the histogram and calculate the position of each pixel in the UV color space; Use a convolutional neural network to estimate the illumination based on the position of each pixel in UV color space; Color correction is performed on the image to be processed according to the illumination.

3. The method according to claim 1, characterized in that Calculating the similarity between the feature value and the feature value of the image stored in the database, comprising: Pre-defining a vector set in a vector database, constructing feature values ​​of the image and metadata of the image stored in the database into a vector index, and inserting the vector index into the vector set; A nearest neighbor search algorithm is used to search the vector set for a vector index that is closest to the characteristic value of the standard image, and the corresponding Euclidean distance similarity is output.

4. The method according to claim 3, characterized in that Saving the characteristic value of the standard image and saving the image to be processed to the database includes: Saving the image to be processed to the database, and acquiring metadata of the image to be processed; Saving the metadata and the feature values ​​of the standard image as new vector indexes; Inserting the new vector index into the vector set; Use a unique key to establish a connection between the vector database and the image database.

5. An image management system for removing the influence of illumination, characterized in that: include: A color correction module, used to perform color correction on the image to be processed by using a color constancy convolution algorithm to obtain a standard image; A feature extraction module, used to extract feature values ​​of the standard image using a deep learning neural network model; A similarity search module, used for calculating the similarity between the feature value and the feature value of the image stored in the database; A repeated determination module, used for determining that there is an image in the database that matches the image to be processed if the maximum similarity reaches a set threshold; The image storage module is used to save the feature value of the standard image and save the image to be processed into the database if the maximum similarity does not reach the set threshold.

6. The system according to claim 5, characterized in that The color correction module comprises: An image acquisition unit, used for intercepting a request to save an image to a database, and extracting the image to be processed from the request; A preprocessing unit, used for preprocessing the image to be processed, wherein the preprocessing includes scaling the image to a specified size and performing a normalization operation; A pixel processing unit, used for constructing a histogram for the preprocessed image to be processed, counting the pixels of the histogram and calculating the position of each pixel in the UV color space; An illumination estimation unit, for estimating illumination based on the position of each pixel in UV color space using a convolutional neural network; A color correction unit is used to perform color correction on the image to be processed according to the illumination.

7. The system according to claim 5, characterized in that The similarity search module comprises: A vector pre-storage unit, used to pre-define a vector set in a vector database, construct a feature value of an image and metadata of the image stored in the database into a vector index, and insert the vector index into the vector set; The vector search unit is used to search the vector index closest to the characteristic value of the standard image from the vector set by using the nearest neighbor search algorithm, and output the corresponding Euclidean distance similarity.

8. The system according to claim 3, characterized in that The image storage module includes: An image storage unit, used for storing the image to be processed in the database and obtaining metadata of the image to be processed; An index building unit, used for saving the metadata and the feature value of the standard image as a new vector index; An index inserting unit, used to insert the new vector index into the vector set; The relationship establishment unit is used to establish a connection relationship between the vector database and the image database using a unique key.

9. A device, characterized in that: include: A memory for storing an image management program for removing the influence of illumination; A processor is used to implement the steps of the image management method for removing the influence of lighting as described in any one of claims 1 to 4 when executing the image management program for removing the influence of lighting.

10. A computer-readable storage medium storing a computer program, characterized in that: The readable storage medium stores an image management program for removing the influence of lighting, and when the image management program for removing the influence of lighting is executed by a processor, the steps of the image management method for removing the influence of lighting as claimed in any one of claims 1 to 4 are implemented.

Images