WO2019209007A1 - Ai 기반의 영상 압축 및 복원 시스템 - Google Patents
Ai 기반의 영상 압축 및 복원 시스템 Download PDFInfo
- Publication number
- WO2019209007A1 WO2019209007A1 PCT/KR2019/004894 KR2019004894W WO2019209007A1 WO 2019209007 A1 WO2019209007 A1 WO 2019209007A1 KR 2019004894 W KR2019004894 W KR 2019004894W WO 2019209007 A1 WO2019209007 A1 WO 2019209007A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- data
- resolution
- image data
- file
- Prior art date
Links
Images
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
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/21—Server components or server architectures
- H04N21/222—Secondary servers, e.g. proxy server, cable television Head-end
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/47—End-user applications
- H04N21/472—End-user interface for requesting content, additional data or services; End-user interface for interacting with content, e.g. for content reservation or setting reminders, for requesting event notification, for manipulating displayed content
Definitions
- the present invention relates to an AI-based image compression and decompression system.
- content is produced in only one resolution.
- the video content may be serviced at a lower resolution.
- the network environment may be unstable, and thus the resolution of the serviced video content may be lowered even if there is a risk of interruption of data transmission and reception.
- the bilinear interpolation method and the bicubic interpolation method are mainly used.
- the conventional image interpolation method adds arbitrary pixels according to the pattern of the surrounding pixels, not just the existing pixels. Since the image is crushed, the upscaling technique is not suitable.
- the present invention is designed to solve such a problem, and based on the data of the artificial neural network learning based on the original video data, an object of the present invention is to effectively restore a low quality compressed file of the video.
- the image compression and decompression system compresses and stores the acquired image data through artificial neural network-based image learning, decompresses the compressed image at the user's request, and restores the resolution to the user device.
- a user device that controls to transmit the image data to the image processing server and the image processing server, and receives a result of image learning and restoration performed by the image processing server.
- the embodiment of the present invention performs a compression operation to lower the resolution and capacity of the video data, and can easily restore the resolution of the compressed file to the original level resolution through an artificial neural network-based operation to use the storage space more efficiently. Can be.
- the present invention can effectively extract information about an image object requested by a user based on the reconstructed video data and provide the same to the user.
- the present invention can store and store a large amount of data for later retrieval of information, even for a large amount of video data having a long playback time and high quality.
- FIG. 1 is a diagram illustrating a configuration of an AI-based image compression and decompression system according to an embodiment of the present invention.
- FIG. 2 is a diagram illustrating a configuration of an image processing server according to an exemplary embodiment of the present invention.
- FIG. 3 is a diagram illustrating a configuration of a data determination unit according to an exemplary embodiment of the present invention.
- FIG. 4 is a diagram illustrating a configuration of an AI image reconstruction unit according to an embodiment of the present invention.
- FIG. 5 is a flowchart illustrating a sequence of image restoration and data operation according to an embodiment of the present invention.
- FIG. 6 is a diagram illustrating an autoscaling operation according to an exemplary embodiment of the present invention.
- the image compression and decompression system compresses and stores the acquired image data through artificial neural network-based image learning, decompresses the compressed image at the user's request, and restores the resolution to the user device.
- a user device that controls to transmit the image data to the image processing server and the image processing server, and receives a result of image learning and restoration performed by the image processing server.
- FIG. 1 is a diagram illustrating a configuration of an AI-based image compression and decompression system according to an embodiment of the present invention.
- An image compression and decompression system includes an image processing server 100 that performs an image compression and decompression function, an image capturing apparatus 200 that provides a captured image to the image processing server 100, and an image processing server 100. It may include a user device 300 and a companion server 400 that receive the performed image processing result.
- the image capturing apparatus 200 may include a drone 201, a CCTV 202, and the like.
- the image capturing apparatus 200 and the companion server 400 may be, for example, an institution or a group server that requires an image processing result performed by the image processing server 100.
- the image processing server 100 may obtain moving image data from the image capturing apparatus 200, and may perform a function of compressing and restoring the obtained moving image data.
- the image processing server 100 may provide a cloud service to a user side of the user device 300 or an institution or a group of the companion server 400.
- the image processing server 100 may not only provide a general cloud service provided by an existing cloud server, but also automatically downscale and store and manage the resolution of image data. Through this downscaling, the image processing server 100 may reduce the capacity of the acquired video, thereby more effectively utilizing the storage space of the cloud server.
- the image processing server 100 restores a compressed file when a user downloads or streams a downscaling file that has been compressed and stored later. In this case, unlike the general decompression method, an upscaling process using an artificial neural network is performed. Will be restored.
- the image capturing apparatus 200 may include a user device such as a camera, a smartphone, or the like, which captures image data at the request of an individual user, an institution, or a group using the image processing server 100. Automatically send to the image processing server 100.
- the user device 300 or the companion server 400 may control the image capturing apparatus 200 to transmit image data to the image processing apparatus 100.
- the image processing and cloud services provided by the image processing server 100 may be utilized not only by the individual user (the user device 300 side) but also by organizations and organizations (for example, the interlocking server 400). For convenience, hereinafter, all objects that use the service of the image processing server 100 will be referred to as 'users'.
- FIG. 2 is a diagram illustrating a configuration of an image processing server according to an exemplary embodiment of the present invention.
- the image processing server 100 may include a communication unit 110, a storage unit 120, and a controller 130.
- the communication unit 110 may use a network for data transmission and reception between a user device and a server, and the type of the network is not particularly limited.
- the network may be, for example, an IP (Internet Protocol) network providing a transmission / reception service of a large amount of data through an Internet protocol (IP), or an All IP network integrating different IP networks.
- IP Internet Protocol
- the network may include a wired network, a wireless broadband network, a mobile communication network including WCDMA, a high speed downlink packet access (HSDPA) network, and a long term evolution (LTE) network, LTE advanced (LTE-A). ), Or one of a mobile communication network including 5G (Five Generation), a satellite communication network, and a Wi-Fi network, or a combination of at least one of them.
- 5G Wireless Generation
- satellite communication network and a Wi-Fi network
- the communication unit 110 may perform a data communication function with the image capturing apparatus 200, the user device 300, and the companion server 400.
- the communication unit 300 may receive image data (original image data) received from the image capturing apparatus 200 or the user device 300 and the companion server 400.
- the communication unit 110 may receive a request signal for an AI learning operation or an AI image reconstruction operation for an image of the image processing server 100 from the user device 300 or the companion server 400.
- the communication unit 110 may receive a request signal for information on a specific image object from an image existing in the image processing server 100 from the user device 300 or the companion server 400.
- the communication unit 110 may perform data communication for providing image data after the image restoration operation to the user device 300 or the companion server 400.
- the communication unit 110 provides a data communication operation for providing a signal for requesting an object extraction and data conversion operation based on the restored image data, a response signal, and a result data (object extraction result data) to the user device 300 or the companion server 400. Can be done.
- the storage unit 120 may include, for example, an internal memory or an external memory.
- the internal memory may be, for example, volatile memory (for example, dynamic RAM (DRAM), static RAM (SRAM), or synchronous dynamic RAM (SDRAM), etc.), non-volatile memory (for example, OTPROM (one). time programmable ROM (PROM), programmable ROM (PROM), erasable and programmable ROM (EPROM), electrically erasable and programmable ROM (EEPROM), mask ROM, flash ROM, flash memory (such as NAND flash or NOR flash), hard drives, Or it may include at least one of a solid state drive (SSD).
- volatile memory for example, dynamic RAM (DRAM), static RAM (SRAM), or synchronous dynamic RAM (SDRAM), etc.
- non-volatile memory for example, OTPROM (one).
- the external memory may be a flash drive such as compact flash (CF), secure digital (SD), micro secure digital (Micro-SD), mini secure digital (mini-SD), extreme digital (XD), It may further include a multi-media card (MMC) or a memory stick.
- the external memory may be functionally and / or physically connected to the electronic device through various interfaces.
- the storage unit 120 may include a raw image DB 121 and a training image DB 122.
- the raw image DB 121 may temporarily store raw image data obtained from the image capturing apparatus 200 or the user device 300 (or the companion server 400) before performing image learning.
- raw image data that consumes a relatively large amount of capacity in order to efficiently use a storage function may be checked after a preset time or by generating a downscaling file and a neural network file that are a result of a predetermined learning (eg, a result of image learning). If is satisfied), it can be deleted automatically.
- a predetermined learning eg, a result of image learning
- the training image DB 122 may store a downscaling file and a neural network file generated as a result of image learning.
- the training image DB 122 may store and set the same names of the downscaling file and the neural network file in order to be able to restore the resolution immediately when the user later selects and plays a specific video. have.
- the training image DB 122 may store not only a neural network file corresponding to each video file, but also a general purpose neural network file that is a result of learning based on an arbitrary video data set.
- the controller 130 may also be referred to as a processor, a controller, a microcontroller, a microprocessor, a microcomputer, or the like.
- the control unit may be implemented by hardware (hardware) or firmware (firmware), software, or a combination thereof.
- an embodiment of the present invention may be implemented in the form of a module, procedure, function, etc. that performs the functions or operations described above.
- the software code may be stored in a memory and driven by the controller.
- the memory may be located inside or outside the user terminal and the server, and may exchange data with the controller by various known means.
- the controller 130 may include a data determination unit 131, an AI learning support unit 132, an AI image restoration unit 133, and an object data conversion unit 134.
- the data determination unit 131 may determine information on the raw image received from the outside.
- the data determination unit 131 may include meta information (eg, resolution) of an image file received from an external image capturing apparatus 200 (images may be received from a user device 300 or a companion server 400 according to various embodiments). , Capacity, name, key, etc.) and files corresponding to the same classification item can be classified.
- the data determination unit 131 may manage data related to a member registered in the image processing server 100.
- the data determination unit 131 may determine management information related to an image data provider, an image photographing device identification value, member registration information, a usage history, and the like.
- the data determiner 131 may identify and classify the image data based on the side providing the image data and the meta information of the image data.
- the data determination unit 131 may include at least one of a predetermined criterion (eg, an image data providing device property, an attribute of a member (eg, an individual or an institution), a service use level, and a use purpose). Based on the video obtained based on the image processing (eg, image compression through image learning, image restoration, etc.) can be classified before.
- the data determination unit 131 classifies video data uploaded by the member to the image processing server 100 based on two types of resolutions (for example, 540p or more), and class A in the case of a B-class member. Members can control to classify based on four types of resolutions (eg 270p, 540p, 720p, 1080p). Such a classification operation for each resolution of a video may be required for preprocessing for image learning.
- the data determination unit 131 may mainly perform resolution-based classification or image type-based classification through the resolution determination unit 131a and the image type determination unit 131b.
- the AI learning support unit 132 of the configuration of the controller 130 may support a compression function through AI-based image learning for the obtained video data.
- the AI learning support unit 132 may generate a downscaling file from the obtained video file and a neural network file that is a file required to restore the quality of the downscaling file to an original level later.
- the downscaling file refers to a compressed file
- the AI learning support unit 132 may perform video file compression by performing such image learning.
- the video file downscaled and compressed by such AI image learning may be reconstructed (decompressed) by a reconstruction operation performed based on a neural network file in the AI image reconstruction unit 133.
- the image learning performed by the AI learning support unit 132 may be performed using a multilayer perceptron implemented to include at least one hidden layer, according to an exemplary embodiment.
- the perceptron can input a plurality of signals and output one signal.
- the weight and bias required in the calculation process using the artificial neural network model can be calculated as appropriate values through the backpropagation method.
- appropriate weight and bias data are extracted through this backpropagation technique.
- a neural network file calculated through an artificial neural network for resolution improvement and restoration of a compressed file may include information about such weight data and bias data.
- the image learning process according to an embodiment of the present invention may be performed after generating a basic neural network file through a basic data set (which may be composed of any of various image or video data according to an embodiment).
- the AI learning support unit 132 inputs the low quality version (downscaling) data of the image to be newly trained as an input value to the artificial neural network algorithm, and the result value when performing the artificial neural network calculation process using the basic neural network file. This input value confirms that the matching ratio of the original image and the reference target value is achieved.
- the AI learning support unit 132 performs the backpropagation technique for calculating the parameters of the artificial neural network for achieving the match rate higher than the reference value. Etc.) can be obtained.
- the file storing the values of the appropriate parameters obtained through this process becomes a neural network file, and the input data restored to the original data when the artificial neural network operation is performed through the neural network file becomes a downscaling file. That is, the AI image learning support unit 132 performs image learning on arbitrary video data (original), and as a result, a 'downscaling file' having a lower resolution and capacity than the original data, and the downscaling file as an original image. You will be able to create a neural network file, the file required to restore it.
- the AI image restoration unit 133 may restore (decompress) the downscaling file generated as a result of image learning by upscaling the neural network file.
- the AI image restoration unit 133 may provide the user device 300 with data whose upscaling is completed and the image quality is restored according to a user request.
- the AI image restoration unit 133 may use a cloud service provided by the image processing server 100, and may request a request for downloading or playing (streaming) a video file previously stored in the cloud.
- the AI image restoration unit 133 may restore (decompress) the quality of the pre-stored downscaling file through the neural network file.
- the object dataizer 134 may extract target object data based on the image data reconstructed by the AI image reconstructor 133 or image data of each frame constituting the corresponding image.
- the object data may refer to a specific object that is requested to be detected by a user such as a person, a car, a number, and the like.
- the object dataizer 134 may automatically extract an object such as a car, a person, or a license plate from the restored image, and record and store the extracted information as data.
- the object data extracted and stored as described above may be delivered to the user at regular or user request later.
- the object dataizer 134 detects an object by dividing it into a case in which a specific identification of an individual object is not required, such as the number and number of objects, and a case in which a specific identification is required, such as a face or a car number of a specific suspect.
- the process can be performed.
- the object data unit 134 may determine the number of objects recognized as a person or the object of a car recognized from a reconstructed image data. Only numbers can be checked and dataized.
- the object data unit 134 extracts data recognized as the number of vehicles from the image data whose image quality is restored. , You can list it. Accordingly, the user can immediately know whether there is a vehicle having a number to be searched among the plurality of vehicles existing in the image.
- FIG. 3 is a diagram illustrating a configuration of a data determination unit according to an exemplary embodiment of the present invention.
- the data determination unit 131 may include a resolution determination unit 131a and an image type determination unit 131b.
- the data determiner 131 may determine and classify types of images for preprocessing of image learning through AI calculation. In addition, the data determination unit 131 may perform classification of the type of image data to effectively check the data and select an appropriate reconstruction mode or a reconstruction option when reconstructing the resolution of the image through an AI operation. In addition, after the resolution reconstruction operation is performed, the data determination unit 131 may perform a classification operation of the image data in order to effectively recognize the object from the reconstructed image data and perform a data operation on the recognized object item.
- the resolution determining unit 131a may classify the obtained data based on the resolution. First, the resolution determining unit 131a may classify the acquired image data according to the resolution in order to determine whether image learning is possible. For example, when the resolution of the determination target image data is set to be learnable only when the resolution is greater than or equal to the reference value (eg, 720p), the resolution determination unit 131a may check whether the resolution of the acquired image data is greater than or equal to the reference value 720p. .
- the reference value eg, 720p
- the resolution of the input data during image learning may be standardized to have a specific value collectively.
- the resolution determining unit 131a checks whether the resolution of the acquired image data is a preset reference value (for example, 720p), and determines that the data less than the reference value is a data group that does not perform image learning separately, and the resolution corresponds to the reference value.
- the corresponding data may be determined as a group performing image learning without a separate preprocessing process.
- the resolution determination unit 131a may classify the image into a group for performing image learning after performing a preprocessing operation of converting the data into a reference value.
- the data determination unit 131 may acquire not only the obtained data but also a downscaling file that is pre-compressed in a format according to an embodiment of the present invention. Accordingly, the resolution determiner 131a may classify the acquired image data according to the resolution in order to extract an appropriate neural network file required for upscaling and decompressing the obtained downscaling file.
- the image type determiner 131b may determine the type of the acquired image data and classify and cluster the obtained image data based on the type.
- the image type determination unit 131b may classify the type of image data according to a criterion such as a photographing device, a photographing purpose, a viewing purpose, or the like.
- the type of the image data may be confirmed from meta information of a file (for example, a photographing device identification value, etc.), or an arbitrary thumbnail image based on a key frame (a key frame may be extracted by a basic function of a codec). It may be confirmed based on.
- the image type determiner 131b may determine and classify whether the acquired image data is an animation based on a thumbnail image of a keyframe.
- the image type determiner 131b may determine whether the photographing device is an attached device such as a CCTV or a black box, or a mobile device such as a drone or a smartphone, based on a change rate of the background image identified through a keyframe. Accordingly, the image type determiner 131b may classify the image based on the type of the photographing apparatus (attached or mobile). The image type determiner 131b may more easily perform image data classification for each photographing device through photographing device information included in metadata of the received image data.
- the image type determiner 131b may classify and classify the image data based on a use preset by the user. As such, the image data classified into types based on a predetermined criterion may also be performed for each data group.
- the image type determiner 131b may perform data classification for preprocessing and object data before the image learning step when the resolution of the acquired data is at a level that can be learned.
- the image type determination unit 131b performs an image reconstruction by the AI image reconstruction unit 133 and then performs a classification operation for performing object data reconstruction by the AI data reconstruction unit 133 when the resolution of the acquired data is not attainable. can do.
- the image type determination unit 131b may provide an environment in which the object dataizer 134 may perform an operation (data extraction operation corresponding to a user's request) through the classification operation for each type of image data.
- the image type determination unit 131b may include image data received by a corresponding user (eg, a black box, a CCTV photographed on a road, or a personal broadcasting person when the image processing purpose of the user is a vehicle license plate search). Only black box and CCTV video can be extracted and classified. Accordingly, the object dataizer 134 may automatically select a black box and a group of data grouped by CCTV image data in advance to search for a license plate, and perform object extraction and data operation. Accordingly, the object dataizer 134 may quickly execute a user request.
- FIG. 4 is a diagram illustrating a configuration of an AI image reconstruction unit according to an embodiment of the present invention.
- the AI image reconstruction unit 133 may include an image type-based application unit 133a and a resolution-based application unit 133b.
- the AI image restoration unit 133 restores the downscaling file (compression file) to its original state through AI calculation using a neural network file (meta file for decompression) generated as a result of image learning in the AI learning support unit 132.
- the AI image reconstruction unit 133 may perform resolution reconstruction of the image data by applying parameter data stored in a neural network file corresponding to the image data to be reconstructed to drive an artificial neural network algorithm.
- the type of artificial neural network may correspond to, for example, SRCNN.
- the image type-based application unit 133a determines a set value (eg, active function type and neural network file of the artificial neural network, the final resolution of the restoration result, etc.) of the AI image reconstruction operation according to the image type information, and corresponds to the image.
- the restoration operation may be performed.
- the image type-based application unit 133a may perform an operation of selecting a neural network file item required for reconstructing the resolution of the compressed image and selecting a setting value according to the image type related to the resolution reconstruction.
- the image type-based application unit 133a may be divided into materials for which the type of image has no viewing purpose such as CCTV, and materials for viewing purpose such as movie or animation. Alternatively, the image type-based application unit 133a may distinguish between data for data extraction purpose and data not.
- the image type-based application unit 133a may determine the resolution reconstruction level of the image data based on the classification information of the image data. For example, the image type-based application unit 133a may perform image restoration by applying a general-purpose neural network file, not a neural network file that has been specifically learned to the corresponding image data, for a material for which there is no purpose of viewing. On the other hand, the image type-based application unit 133a may perform image reconstruction by searching for and extracting a neural network file learned for the image data. This may be an operation reflecting the fact that the matching ratio with the original data may be regarded as important for image data such as a movie or a drama that is intended for viewing. However, this is only one embodiment, it is possible to apply a different way of differentially enough. As described above, the image type-based application unit 133a may apply different setting data to be selected in the image restoration process based on the image type, thereby providing efficiency of the restoration operation.
- the resolution-based application unit 133b may differentially set a setting value (eg, an active function type of an artificial neural network and a neural network file, a final resolution of a restoration result, etc.) based on the resolution of the image data to be restored. For example, the resolution-based application unit 133b sets a target resolution differently according to a fee paid by the user or a type of service selected by the user according to a service policy (for example, in the case of A service, 270p is restored to 4K resolution, and B service is performed. In this case, 270p may be set to restore the resolution up to FHD).
- a setting value eg, an active function type of an artificial neural network and a neural network file, a final resolution of a restoration result, etc.
- the resolution-based application unit 133b may designate a target resolution corresponding to the type of resolution before performing the restoration, and search and apply an appropriate neural network file to perform the restoration operation to the target resolution. For example, the resolution-based application unit 133b automatically performs upscaling on the obtained reconstruction image data (eg, downscaling file), but 270p is FHD (1080p), 540p is 4K (2060p), and 720p. Is 5K (2880p), and 1080p may perform upscaling (or decompression) to 4320p.
- FIG. 6 A schematic of such an automatic upscaling operation is shown through FIG. 6.
- FIG. 6 is a diagram illustrating an autoscaling operation according to an exemplary embodiment of the present invention.
- the resolution-based application unit 133b may perform an automatic upscaling operation according to the resolution of the obtained image data.
- the classification operation according to the resolution may be performed by the resolution determination unit 131a of the data determination unit 131.
- the operation of upscaling the classified image data to each target resolution may be performed by the resolution-based application unit 133b of the AI image restoration unit 133. It can be done through.
- FIG. 5 is a flowchart illustrating a sequence of image restoration and data operation according to an embodiment of the present invention.
- the controller 130 of the image processing server 100 may first perform operation 505 of receiving image data from an external image capturing apparatus 200.
- the controller 130 checks the received image data, and performs operation 510 of restoring the resolution by upscaling the corresponding image data.
- resolution upscaling or upscaling and object recognition
- separate image learning downscaling file and neural network
- the file generation operation may be omitted, and the upscaling operation may be immediately performed.
- the controller 130 performs object recognition from the reconstructed image data or frame images of the image data and classifies the recognized object. Can be.
- the controller 130 may search for an object using a specific technique (for example, YOLO; You Only Look Once, deep learning based real-time object search technique, etc.) from the reconstructed image.
- a specific technique for example, YOLO; You Only Look Once, deep learning based real-time object search technique, etc.
- the controller 130 may not only extract the image object, but also classify the extracted image object by items.
- the controller 130 may classify an object extracted from the restored image data into a person, a car, and the like according to a predetermined criterion.
- the controller 130 may classify the extracted human object into items such as a man, a woman, an adult, and a child according to a more detailed classification criterion.
- the controller 130 may perform operation 520 of data (for example, listing) the recognized object information and outputting the same. Since it is difficult for a user to catch an object image detected and identified in real time without directly missing it, the controller 130 may list the recognized object information and provide it to the user. Such an operation may allow a user to search through the documented data instead of directly checking the image data in order to confirm the presence or absence of a specific object (eg, a car) in the image data. Accordingly, the user can perform an object search operation from the image data in a more effective manner.
- data for example, listing
- the controller 130 may list the recognized object information and provide it to the user.
- Such an operation may allow a user to search through the documented data instead of directly checking the image data in order to confirm the presence or absence of a specific object (eg, a car) in the image data. Accordingly, the user can perform an object search operation from the image data in a more effective manner.
- the controller 130 of the image processing server 100 may provide an effective data storage and management function to the user in more various ways.
- the controller 130 may include a remaining capacity controller (not shown).
- the controller 130 automatically checks the remaining capacity (eg, cloud capacity) allocated to each user (account), and newly uploads the The image data may be compressed and stored, and a compression ratio (degree of reduction in capacity compared to the original of the downscaling file) of the image data may be automatically set according to the confirmed remaining capacity. If it is determined that there is little remaining capacity for a particular user, it may be controlled to improve the downscaling level of the image data received by the image processing server 100.
- the image processing server 100 may check the remaining capacity level of the storage in real time and change a setting value such as a downscaling target value to automatically set the image processing level in response to the storage situation according to various embodiments.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Description
Claims (9)
- 획득된 영상 데이터를 인공신경망 기반의 영상 학습을 통해 압축하여 저장하고, 사용자 요청에 따라 압축된 영상의 압축을 해제하여 해상도를 복원하고 사용자 기기에 제공하는 영상 처리 서버; 및상기 영상 처리 서버로 영상 데이터를 전송하도록 제어하고, 상기 영상 처리 서버에서 수행된 영상 학습 및 복원이 수행된 결과물을 수신하는 사용자 기기;를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 1항에 있어서,상기 영상 처리 서버는영상 데이터를 외부로부터 수신하는 데이터 통신부;상기 영상 데이터를 기반으로 영상 학습을 수행하여, 상기 영상 데이터의 해상도와 용량을 저감한 압축파일인 다운스케일링 파일과, 상기 다운스케일링 파일을 인공 신경망 연산을 통해 복원하는 데 요구되는 파일인 신경망 파일을 생성하는 제어부; 및영상 학습의 결과로 생성된 상기 다운스케일링 파일과 상기 신경망 파일을 저장하는 저장부;를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 2항에 있어서,상기 제어부는영상 데이터를 기 설정된 기준에 기반하여 분류하는 데이터 판단부;를 포함하고,상기 기 설정된 기준은 영상 데이터 제공 장치의 속성, 회원의 속성, 서비스 이용 등급, 이용 목적 중 적어도 하나를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 3항에 있어서,상기 데이터 판단부는영상 학습의 가능 여부를 판단하기 위해 해상도가 기준값 이상인지 여부를 판단하며, 기준값 미만인 영상 데이터에 대하여는 영상 학습을 수행하지 않는 그룹으로 분류하고, 기준값 이상인 데이터는 영상 학습을 수행하는 그룹으로 분류하되, 기준값보다 높은 해상도를 갖는 데이터는 해상도를 기준값으로 변환하는 전처리를 수행하는 해상도 판단부;영상 데이터의 메타 데이터에 포함된 촬영 장치 정보를 기반으로 촬영 장치별 분류를 수행하는 영상 종류 판단부;를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 2항에 있어서,상기 제어부는원본 영상 데이터를 기 설정된 해상도로 낮춘 다운스케일링 파일과, 다운스케일링 파일을 복원하는 데 요구되는 파일인 신경망 파일을 생성하는 AI 학습 지원부;를 포함하되,상기 신경망 파일은상기 다운스케일링 파일을 대상으로 인공신경망 연산을 수행하였을 때 원본 영상 데이터와의 일치율이 기준치 이상이 되도록 하는 인공신경망 파라미터로 구성되는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 2항에 있어서,상기 제어부는영상 학습 결과로 생성된 다운스케일링 파일의 해상도를 인공신경망 연산을 통해 복원하는 AI 영상 복원부;를 포함하되,영상 데이터의 종류에 따라 AI 영상 복원 동작에 관한 설정값을 결정하고, 결정된 설정값을 적용하여 영상 복원 동작을 수행하는 영상 종류 기반 적용부;영상 데이터의 해상도에 대응하여 복원 동작에 관한 설정값을 결정하고, 결정된 설정값을 적용하여 영상 복원 동작을 수행하는 해상도 기반 적용부;를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 2항에 있어서,상기 제어부는영상 데이터의 해상도 복원이 수행된 이후, 복원된 영상 데이터의 각 프레임 이미지들로부터 특정 이미지 객체를 추출하고, 추출된 이미지 객체의 항목을 리스트화하는 객체 데이터화부;를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 2항에 있어서,상기 제어부는각 사용자에게 할당된 잔여 용량을 확인하고, 신규 업로드된 영상 데이터를 압축하여 저장하되, 확인된 잔여 용량에 대응하여 상기 영상 데이터의 압축률을 자동 설정하는 잔여 용량 제어부;를 포함하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
- 제 1항에 있어서,상기 영상 처리 서버는영상 데이터를 수신하여 저장하는 클라우드 서비스를 제공하는 것을 특징으로 하는 영상 압축 및 복원 시스템.
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR20180047305 | 2018-04-24 | ||
KR10-2018-0047305 | 2018-04-24 | ||
KR10-2019-0046955 | 2019-04-22 | ||
KR1020190046955A KR102190483B1 (ko) | 2018-04-24 | 2019-04-22 | Ai 기반의 영상 압축 및 복원 시스템 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2019209007A1 true WO2019209007A1 (ko) | 2019-10-31 |
Family
ID=68294157
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2019/004894 WO2019209007A1 (ko) | 2018-04-24 | 2019-04-23 | Ai 기반의 영상 압축 및 복원 시스템 |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2019209007A1 (ko) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20220215198A1 (en) * | 2020-12-26 | 2022-07-07 | International Business Machines Corporation | Dynamic multi-resolution processing for video classification |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040045030A1 (en) * | 2001-09-26 | 2004-03-04 | Reynolds Jodie Lynn | System and method for communicating media signals |
US20130034313A1 (en) * | 2011-08-05 | 2013-02-07 | Zhe Lin | Regression-Based Learning Model for Image Upscaling |
US20160180502A1 (en) * | 2014-12-22 | 2016-06-23 | Thomson Licensing | Method for upscaling an image and apparatus for upscaling an image |
KR20170019159A (ko) * | 2015-08-11 | 2017-02-21 | 삼성전자주식회사 | 전자 장치 및 이미지 처리 방법 |
US20170347061A1 (en) * | 2015-02-19 | 2017-11-30 | Magic Pony Technology Limited | Machine Learning for Visual Processing |
-
2019
- 2019-04-23 WO PCT/KR2019/004894 patent/WO2019209007A1/ko active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040045030A1 (en) * | 2001-09-26 | 2004-03-04 | Reynolds Jodie Lynn | System and method for communicating media signals |
US20130034313A1 (en) * | 2011-08-05 | 2013-02-07 | Zhe Lin | Regression-Based Learning Model for Image Upscaling |
US20160180502A1 (en) * | 2014-12-22 | 2016-06-23 | Thomson Licensing | Method for upscaling an image and apparatus for upscaling an image |
US20170347061A1 (en) * | 2015-02-19 | 2017-11-30 | Magic Pony Technology Limited | Machine Learning for Visual Processing |
KR20170019159A (ko) * | 2015-08-11 | 2017-02-21 | 삼성전자주식회사 | 전자 장치 및 이미지 처리 방법 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20220215198A1 (en) * | 2020-12-26 | 2022-07-07 | International Business Machines Corporation | Dynamic multi-resolution processing for video classification |
US11954910B2 (en) * | 2020-12-26 | 2024-04-09 | International Business Machines Corporation | Dynamic multi-resolution processing for video classification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102190483B1 (ko) | Ai 기반의 영상 압축 및 복원 시스템 | |
CN110633669B (zh) | 家居环境中基于深度学习的移动端人脸属性识别方法 | |
WO2018150083A1 (en) | A method and technical equipment for video processing | |
WO2021164216A1 (zh) | 一种视频编码方法、装置、设备及介质 | |
Yu et al. | Pinto: enabling video privacy for commodity iot cameras | |
EP3405889A1 (en) | Cloud platform with multi camera synchronization | |
CN114679607A (zh) | 一种视频帧率控制方法、装置、电子设备及存储介质 | |
US20240007720A1 (en) | Peer-to-Peer Data Transfer for Video on Demand from an Edge Data Storage Device to a Browser | |
US11095901B2 (en) | Object manipulation video conference compression | |
WO2019209007A1 (ko) | Ai 기반의 영상 압축 및 복원 시스템 | |
US20220006960A1 (en) | Camera analyzing images on basis of artificial intelligence, and operating method therefor | |
US20190370553A1 (en) | Filtering of false positives using an object size model | |
KR20220070866A (ko) | 딥러닝 기술이 적용된 영상 개선 방법, 장치 및 프로그램 | |
KR102130076B1 (ko) | 특징 영역의 학습 중요도를 바탕으로 스트리밍 파일의 해상도를 개선하는 방법 | |
CN110765869A (zh) | 分渠道采集数据的唇语活体检测方法、系统和计算机设备 | |
CN107979568A (zh) | 一种视频监控应用系统 | |
WO2019135270A1 (ja) | 動画像解析装置、動画像解析システム、動画像解析方法、及びプログラム | |
WO2019209008A1 (ko) | 변화 매크로블록 추출 기법을 이용한 동영상 화질 개선 시스템 | |
WO2019209005A1 (ko) | 인공지능 기반 해상도 개선 시스템 | |
KR102130077B1 (ko) | 격자 생성 패턴 정보를 바탕으로 해상도를 개선하는 시스템 | |
CN114513681A (zh) | 一种视频处理系统、方法、装置、电子设备和存储介质 | |
CN113792021A (zh) | 一种降低图片存储空间的方法 | |
WO2019209006A1 (ko) | 스트리밍 파일의 해상도 개선 방법 | |
KR20220021494A (ko) | Ai 기반 해상도 개선 시스템 | |
US9451275B2 (en) | System and method for storing and moving graphical image data sets with reduced data size requirements |
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: 19794057 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19794057 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 20.04.2021) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19794057 Country of ref document: EP Kind code of ref document: A1 |