WO2023027503A1 - 화상회의시 ai(artificial intelligence)를 이용한 다운스케일 및 업스케일 방법 및 장치 - Google Patents
화상회의시 ai(artificial intelligence)를 이용한 다운스케일 및 업스케일 방법 및 장치 Download PDFInfo
- Publication number
- WO2023027503A1 WO2023027503A1 PCT/KR2022/012646 KR2022012646W WO2023027503A1 WO 2023027503 A1 WO2023027503 A1 WO 2023027503A1 KR 2022012646 W KR2022012646 W KR 2022012646W WO 2023027503 A1 WO2023027503 A1 WO 2023027503A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- electronic device
- dnn
- data
- information
- Prior art date
Links
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 1189
- 238000000034 method Methods 0.000 title description 123
- 238000003672 processing method Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 description 166
- 230000000875 corresponding effect Effects 0.000 description 104
- 230000008569 process Effects 0.000 description 87
- 238000012545 processing Methods 0.000 description 48
- 238000007906 compression Methods 0.000 description 44
- 230000006835 compression Effects 0.000 description 44
- 238000010586 diagram Methods 0.000 description 43
- 230000004913 activation Effects 0.000 description 35
- 230000005540 biological transmission Effects 0.000 description 31
- 230000006870 function Effects 0.000 description 24
- 238000012546 transfer Methods 0.000 description 20
- 238000013139 quantization Methods 0.000 description 17
- 238000006243 chemical reaction Methods 0.000 description 16
- 230000008859 change Effects 0.000 description 12
- 238000013528 artificial neural network Methods 0.000 description 7
- 238000013507 mapping Methods 0.000 description 7
- 230000003287 optical effect Effects 0.000 description 6
- 230000000153 supplemental effect Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 238000013500 data storage Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 3
- 230000003213 activating effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 210000003792 cranial nerve Anatomy 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000010839 reverse transcription Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/14—Systems for two-way working
- H04N7/15—Conference systems
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/59—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/30—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using hierarchical techniques, e.g. scalability
-
- 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
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/136—Incoming video signal characteristics or properties
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/164—Feedback from the receiver or from the transmission channel
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
- H04N19/436—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation using parallelised computational arrangements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/46—Embedding additional information in the video signal during the compression process
-
- 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
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/234—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs
- H04N21/2343—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs involving reformatting operations of video signals for distribution or compliance with end-user requests or end-user device requirements
-
- 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
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
- H04N21/4402—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving reformatting operations of video signals for household redistribution, storage or real-time display
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/01—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level
- H04N7/0117—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level involving conversion of the spatial resolution of the incoming video signal
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/01—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level
- H04N7/0127—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level by changing the field or frame frequency of the incoming video signal, e.g. frame rate converter
Definitions
- the present disclosure relates to a method and apparatus for processing an image during a video conference, and more specifically, the present disclosure relates to a method and apparatus for downscaling or upscaling an image using artificial intelligence (AI) during a video call. .
- AI artificial intelligence
- the video is encoded by a codec that complies with a predetermined data compression standard, for example, the Moving Picture Expert Group (MPEG) standard, and then stored in a bit stream in a recording medium or transmitted through a communication channel.
- MPEG Moving Picture Expert Group
- an image during a video conference is downscaled and transmitted using artificial intelligence (AI), and the downscaled image is adaptively transmitted using AI according to the importance or priority of the transmitted video conference video. It is a technical challenge to upscale to .
- AI artificial intelligence
- An electronic device participating in a video conference using AI includes a display; and a processor that executes one or more instructions stored in the electronic device, wherein the processor includes image data generated as a result of a first encoding of a first image related to another electronic device participating in the video conference, and an original image.
- AI artificial intelligence
- Acquiring AI data related to AI downscaling to the first image from a server first decoding the image data to obtain a second image corresponding to the first image, and obtaining a second image corresponding to the first image, based on the degree of importance to the other electronic device Therefore, it is determined whether to perform AI upscaling on the second image, and if it is determined to perform AI upscaling, AI upscaling the second image through a DNN for upscaling to obtain a third image , When it is determined that the third image is provided to the display and the AI upscaling is not performed, the second image may be provided to the display.
- a server for managing a video conference using artificial intelligence (AI), proposed in the present disclosure to solve the above technical problem, includes a processor that executes one or more instructions stored in the server, wherein the processor comprises: First image data generated as a result of first encoding a first image and AI data related to AI downscaling from an original image to the first image are obtained from a first electronic device participating in a video conference, and the first image Data is first decoded to obtain a second image corresponding to the first image, and if the importance of the first electronic device indicates that the user is a listener, the second image obtained by first encoding the second image Data is transmitted to the second electronic device, and if the importance indicates that the user of the first electronic device is a presenter, AI upscales the second image through a DNN for upscaling to obtain a third image, and the third image is obtained. Third image data obtained by first encoding the 3 images may be transmitted to the second electronic device.
- a first related to another electronic device participating in the video conference acquiring image data generated as a result of first encoding an image and AI data related to AI downscaling from an original image to the first image from a server; obtaining a second image corresponding to the first image by first decoding the image data; determining whether to perform AI upscaling on the second image based on the importance of the other electronic device; if it is determined that the AI upscaling is to be performed, AI upscaling the second image through a DNN for upscaling to obtain a third image, and providing the third image to the display; and providing the second image to the display when it is determined that the AI upscaling is not performed.
- AI artificial intelligence
- AI up-scale or AI down-scale is applied to the video conference video based on whether the electronic device of the video conference participant and the video conference server can perform AI up-scale or AI down-scale, so that a number of existing users are fixed. It is possible to reduce the user's data usage while maintaining high-quality video conferencing video, unlike exchanging video at the specified bit rate and resolution.
- FIG. 1 is a diagram for explaining an AI encoding process and an AI decoding process according to an embodiment.
- FIG. 2 is a block diagram showing the configuration of an AI decoding apparatus according to an embodiment.
- FIG. 3 is an exemplary diagram illustrating a second DNN for AI up-scaling of a second image.
- FIG. 4 is a diagram for explaining a convolution operation by a convolution layer.
- FIG. 5 is an exemplary diagram illustrating a mapping relationship between various pieces of image-related information and various pieces of DNN setting information.
- FIG. 6 is a diagram illustrating a second image composed of a plurality of frames.
- FIG. 7 is a block diagram showing the configuration of an AI encoding apparatus according to an embodiment.
- FIG. 8 is an exemplary diagram illustrating a first DNN for AI downscaling of an original video.
- FIG. 9 is a diagram showing the structure of AI encoded data according to an embodiment.
- FIG. 10 is a diagram showing the structure of AI encoded data according to another embodiment.
- 11 is a diagram for explaining a method of training a first DNN and a second DNN.
- FIG. 12 is a diagram for explaining a training process of a first DNN and a second DNN by a training device.
- FIG. 13 is a block diagram illustrating a configuration of an electronic device that AI upscales a video conference video according to importance of other electronic devices participating in the video conference according to an embodiment.
- FIG. 14 is a block diagram illustrating a configuration of an electronic device that AI downscales a video conference video of an electronic device participating in a video conference during a video conference according to an embodiment.
- 15 is a block diagram illustrating a configuration of a server that AI upscales a video conference video according to whether an electronic device of a participant supports AI upscaling and importance information during a video conference according to an embodiment.
- 16 is a block diagram illustrating a configuration of a server for managing a video conference that AI downscales a video conference video according to an exemplary embodiment
- 17 is a diagram for explaining a data transmission relationship between an electronic device and a server in a conventional video conference.
- 18 is a diagram for explaining a data transmission relationship between an electronic device supporting AI downscaling and AI upscaling in a videoconference and a server according to an embodiment.
- 19 is an electronic device that supports AI downscale and AI upscale in a video conference, an electronic device that does not support AI downscale and AI upscale, and a server that supports AI downscale and AI upscale according to an embodiment. It is a diagram for explaining the data transmission relationship between
- 20 is an electronic device supporting AI downscaling and AI upscaling in a video conference, an electronic device not supporting AI downscaling and AI upscaling, and a server supporting AI downscaling and AI upscaling according to another embodiment. It is a diagram for explaining the data transmission relationship between
- 21 is an electronic device supporting AI downscaling and AI upscaling in a video conference, an electronic device not supporting AI downscaling and AI upscaling, and a server supporting AI downscaling and AI upscaling according to another embodiment; It is a diagram for explaining the data transmission relationship between
- 22 is a diagram for explaining arrangement and importance of video conference images displayed on electronic devices participating in a video conference.
- FIG. 23 is a flowchart illustrating a method of AI upscaling a video conference image according to the importance of other electronic devices by an electronic device participating in a video conference during a video conference according to an embodiment.
- FIG. 24 illustrates a video conference video by a server managing a video conference, acquiring an AI downscaled image from an electronic device participating in the video conference, and determining whether to support AI upscaling according to importance information of the electronic device. It is a flowchart for explaining a method of AI up-scaling and transmitting to another electronic device.
- the expression “at least one of a, b, or c” means “a”, “b”, “c”, “a and b”, “a and c”, “b and c”, “a, b” and c”, or variations thereof.
- one component when one component is referred to as “connected” or “connected” to another component, the one component may be directly connected or directly connected to the other component, but in particular Unless otherwise described, it should be understood that they may be connected or connected via another component in the middle.
- components expressed as ' ⁇ unit (unit)', 'module', etc. are two or more components combined into one component, or one component is divided into two or more components for each more subdivided function. may be differentiated into.
- each of the components to be described below may additionally perform some or all of the functions of other components in addition to its own main function, and some of the main functions of each component may be different from other components. Of course, it may be performed exclusively by a component.
- 'image' or 'picture' may indicate a still image, a moving image composed of a plurality of continuous still images (or frames), or a video.
- 'DNN deep neural network
- 'DNN deep neural network
- a 'parameter' is a value used in the calculation process of each layer constituting the neural network, and may include, for example, a weight used when an input value is applied to a predetermined calculation expression. Also, parameters may be expressed in a matrix form. A parameter is a value set as a result of training and can be updated through separate training data as needed.
- 'first DNN' refers to a DNN used for AI downscaling of an image
- 'second DNN' refers to a DNN used for AI upscaling of an image
- 'DNN setting information' is information related to elements constituting a DNN and includes the aforementioned parameters.
- a first DNN or a second DNN may be configured using the DNN configuration information.
- 'original video' means an image to be subjected to AI encoding
- 'first video' means an image obtained as a result of AI downscaling of the original video in the AI encoding process
- 'second image' means an image obtained through the first decoding in the AI decoding process
- 'third image' means an image obtained by AI up-scaling the second image in the AI decoding process.
- 'AI downscale' means a process of reducing the resolution of an image based on AI
- 'first encoding' means an encoding process by a frequency conversion-based video compression method
- 'first decoding' means a decoding process by a frequency conversion-based image restoration method
- 'AI upscaling' means a process of increasing the resolution of an image based on AI.
- AI 1 is a diagram for explaining an artificial intelligence (AI) encoding process and an AI decoding process according to an embodiment.
- AI artificial intelligence
- an original image 105 having a high resolution is AI downscaled 110 to obtain a first image 115 .
- the first encoding 120 and the first decoding 130 are performed on the first image 115 having a relatively small resolution, the first encoding 120 and the first decoding 120 on the original image 105 Compared to the case where the first decoding 130 is performed, the bit rate can be greatly reduced.
- an original image 105 is AI downscaled 110 to obtain a first image 115, and the first image 115 is Encode (120).
- AI decoding process AI encoding data including AI data and image data obtained as a result of AI encoding is received, a second image 135 is obtained through the first decoding 130, and the second image 135 is obtained.
- a third image 145 is obtained by AI up-scaling 140 .
- the original image 105 is AI downscaled 110 to obtain the first image 115 with a predetermined resolution and/or a predetermined quality.
- the AI downscale 110 is performed based on AI, and the AI for the AI downscale 110 must be trained in conjunction with the AI for the AI upscale 140 of the second image 135 (joint trained) do. This is because, when AI for AI downscale 110 and AI for AI upscale 140 are trained separately, between the original image 105, which is an AI encoding target, and the third image 145 restored through AI decoding. because the difference between
- AI data may be used to maintain this linkage in the AI encoding process and the AI decoding process. Therefore, the AI data obtained through the AI encoding process must include information indicating the upscale target, and in the AI decoding process, the AI upscales the second image 135 according to the upscale target identified based on the AI data ( 140) should be done.
- the AI for AI downscale 110 and the AI for AI upscale 140 may be implemented as a deep neural network (DNN).
- DNN deep neural network
- the AI encoding apparatus provides target information used when the first DNN and the second DNN are jointly trained. is provided to the AI decoding device, and the AI decoding device may AI upscale 140 the second image 135 to a target quality and/or resolution based on the provided target information.
- the first encoding 120 includes a process of generating predicted data by predicting the first image 115, a process of generating residual data corresponding to a difference between the first image 115 and the predicted data, and a spatial domain component, It may include a process of transforming the residual data into frequency domain components, a process of quantizing the residual data transformed into frequency domain components, and a process of entropy encoding the quantized residual data.
- Such a first encoding process 120 is MPEG-2, H.264 AVC (Advanced Video Coding), MPEG-4, HEVC (High Efficiency Video Coding), VC-1, VP8, VP9 and AV1 (AOMedia Video 1) It may be implemented through one of image compression methods using equal frequency transform.
- the second image 135 corresponding to the first image 115 may be restored through the first decoding 130 of image data.
- the first decoding 130 includes a process of generating quantized residual data by entropy decoding image data, a process of inverse quantizing the quantized residual data, a process of converting residual data of frequency domain components into spatial domain components, and prediction data. It may include a process of generating and a process of restoring the second image 135 using the prediction data and residual data.
- the first decoding (130) process compresses images using frequency conversion such as MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, and AV1 used in the first encoding (120) process. It may be implemented through an image restoration method corresponding to one of the methods.
- the AI encoded data obtained through the AI encoding process may include image data obtained as a result of the first encoding 120 of the first image 115 and AI data related to the AI downscale 110 of the original image 105.
- Image data may be used in the first decoding 130 process, and AI data may be used in the AI upscale 140 process.
- Image data may be transmitted in the form of a bit stream.
- the image data may include data obtained based on pixel values in the first image 115, eg, residual data that is a difference between the first image 115 and prediction data of the first image 115.
- the image data includes information used in the first encoding 120 process of the first image 115 .
- the image data includes prediction mode information used in the first encoding 120 of the first image 115, motion information, and quantization parameter related information used in the first encoding 120. can do.
- the image data is a video compression method used in the first encoding 120 process among video compression methods using frequency conversion such as MPEG-2, H.264 AVC, MPEG-4, HEVC, VC-1, VP8, VP9 and AV1. It may be generated according to rules, for example, syntax.
- AI data is used for AI upscale 140 based on the second DNN.
- the AI data includes information enabling accurate AI upscaling 140 of the second image 135 through the second DNN to be performed.
- the AI upscale 140 may be performed to the target resolution and/or quality of the second image 135 based on the AI data.
- AI data may be transmitted together with image data in the form of a bit stream. Depending on implementation, AI data may be transmitted separately from image data in the form of frames or packets.
- AI data may be included in image data and transmitted.
- Video data and AI data may be transmitted through the same network or different networks.
- FIG. 2 is a block diagram showing the configuration of an AI decoding apparatus 200 according to an embodiment.
- the AI decoding apparatus 200 includes a receiving unit 210 and an AI decoding unit 230.
- the AI decoder 230 may include a parser 232, a first decoder 234, an AI upscaler 236, and an AI setter 238.
- the receiving unit 210 and the AI decoding unit 230 may be implemented by one processor.
- the receiver 210 and the AI decoder 230 may be implemented as a dedicated processor, and may be implemented with a general-purpose processor such as an application processor (AP), central processing unit (CPU), or graphic processing unit (GPU) and S/W. It may be implemented through a combination of
- a dedicated processor may include a memory for implementing an embodiment of the present disclosure or a memory processing unit for using an external memory.
- the receiving unit 210 and the AI decoding unit 230 may be composed of a plurality of processors. In this case, it may be implemented with a combination of dedicated processors, or it may be implemented with a combination of S/W and a plurality of general-purpose processors such as APs, CPUs, or GPUs.
- the receiving unit 210 is implemented with a first processor
- the first decoding unit 234 is implemented with a second processor different from the first processor
- the AI setting unit 238 may be implemented as a third processor different from the first processor and the second processor.
- the receiving unit 210 receives AI encoded data obtained as a result of AI encoding.
- AI-encoded data may be a video file having a file format such as mp4 or mov.
- the receiving unit 210 may receive AI-encoded data transmitted over a network.
- the receiving unit 210 outputs the AI-encoded data to the AI decoding unit 230.
- the AI-encoded data is magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. It may be obtained from a data storage medium including an optical medium) or the like.
- the parsing unit 232 parses the AI-encoded data, transfers image data generated as a result of the first encoding of the first image 115 to the first decoding unit 234, and transfers the AI data to the AI setting unit 238. convey
- the parsing unit 232 may parse image data and AI data separately included in AI-encoded data.
- the parsing unit 232 may read a header in the AI-encoded data and distinguish between AI data and image data included in the AI-encoded data.
- AI data may be included in a Vendor Specific InfoFrame (VSIF) within the HDMI stream.
- VSIF Vendor Specific InfoFrame
- AI encoded data including separated AI data and image data will be described later with reference to FIG. 9 .
- the parsing unit 232 parses video data from AI-encoded data, extracts AI data from the video data, transfers the AI data to the AI setting unit 238, and performs first decoding on the remaining video data. It can be passed to unit 234. That is, AI data may be included in video data. For example, AI data may be included in supplemental enhancement information (SEI), which is an additional information area of a bitstream corresponding to video data.
- SEI Supplemental Enhancement information
- the parsing unit 232 divides a bitstream corresponding to image data into a bitstream to be processed by the first decoding unit 234 and a bitstream corresponding to AI data, and each of the divided bitstreams It can be output to the first decoding unit 234 and the AI setting unit 238.
- the parsing unit 232 acquires video data included in the AI-encoded data through a predetermined codec (eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1). It can also be confirmed that it is image data. In this case, corresponding information may be transmitted to the first decoder 234 so that the video data can be processed with the identified codec.
- a predetermined codec eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1
- the first decoding unit 234 restores the second image 135 corresponding to the first image 115 based on the image data received from the parsing unit 232 .
- the second image 135 obtained by the first decoder 234 is provided to the AI upscaler 236.
- first decoding-related information such as prediction mode information, motion information, and quantization parameter information may be provided from the first decoding unit 234 to the AI setting unit 238.
- the first decoding related information may be used to obtain DNN configuration information.
- the AI data provided to the AI setting unit 238 includes information enabling AI up-scaling of the second image 135 .
- the upscale target of the second image 135 should correspond to the downscale target of the first DNN. Therefore, the AI data must include information capable of identifying the downscale target of the first DNN.
- the information included in the AI data is specifically exemplified, there are difference information between the resolution of the original image 105 and the resolution of the first image 115 and information related to the first image 115 .
- the difference information may be expressed as information about a degree of resolution conversion of the first image 115 compared to the original image 105 (eg, resolution conversion rate information).
- the difference information can be expressed only with the resolution information of the original image 105.
- the resolution information may be expressed as a horizontal/vertical screen size, or as a ratio (16:9, 4:3, etc.) and a size on one axis.
- it may be expressed in the form of an index or flag.
- information related to the first image 115 includes the resolution of the first image 115, the bit rate of image data obtained as a result of the first encoding of the first image 115, and the first encoding of the first image 115. It may include information on at least one of codec types used at the time of writing.
- the AI setting unit 238 may determine an upscale target of the second image 135 based on at least one of difference information included in the AI data and information related to the first image 115 .
- the upscaling target may indicate, for example, to what resolution the second image 135 should be upscaled.
- the AI upscaler 236 AI upscales the second image 135 through a second DNN to obtain a third image 145 corresponding to the upscale target.
- FIG. 3 is an exemplary diagram showing the second DNN 300 for AI up-scaling of the second image 135, and FIG. 4 shows the convolution operation in the first convolution layer 310 shown in FIG. are showing
- the second image 135 is input to the first convolution layer 310 .
- 3X3X4 displayed on the first convolution layer 310 shown in FIG. 3 exemplifies convolution processing on one input image using four filter kernels each having a size of 3x3.
- 4 feature maps are generated by 4 filter kernels.
- Each feature map represents unique characteristics of the second image 135 .
- each feature map may indicate a vertical direction characteristic, a horizontal direction characteristic, or an edge characteristic of the second image 135 .
- one A feature map 450 may be created. Since four filter kernels are used in the first convolution layer 310, four feature maps can be generated through a convolution operation process using the four filter kernels.
- I1 to I49 displayed on the second image 135 represent pixels of the second image 135
- F1 to F9 displayed on the filter kernel 430 represent parameters of the filter kernel 430
- M1 to M9 displayed on the feature map 450 represent samples of the feature map 450 .
- the second image 135 includes 49 pixels, but this is only one example, and when the second image 135 has a resolution of 4K, for example, 3840 X 2160 pixels. may contain pixels.
- pixel values of I1, I2, I3, I8, I9, I10, I15, I16, and I17 of the second image 135 and F1, F2, F3, F4, and F5 of the filter kernel 430 , F6, F7, F8, and F9 are respectively multiplied, and a value obtained by combining (eg, an addition operation) result values of the multiplication operation may be assigned as a value of M1 of the feature map 450.
- pixel values of I3, I4, I5, I10, I11, I12, I17, I18, and I19 of the second image 135 and F1 and F2 of the filter kernel 430 , F3, F4, F5, F6, F7, F8, and F9 are each multiplied, and a value obtained by combining the result values of the multiplication operation may be assigned as the value of M2 of the feature map 450.
- a convolution operation between pixel values in the second image 135 and parameters of the filter kernel 430 is performed.
- a feature map 450 having a predetermined size may be obtained.
- parameters of a second DNN through joint training of the first DNN and the second DNN for example, parameters of a filter kernel used in convolutional layers of the second DNN (eg, filter The values of F1, F2, F3, F4, F5, F6, F7, F8 and F9) of the kernel 430 may be optimized.
- the AI setting unit 238 determines an upscale target corresponding to the downscale target of the first DNN based on the AI data, and parameters corresponding to the determined upscale target are used in convolution layers of the second DNN. It can be determined by the parameters of the filter kernel.
- the convolution layers included in the first DNN and the second DNN can perform processing according to the convolution operation process described in relation to FIG. 4, but the convolution operation process described in FIG. 4 is only an example, and is limited thereto. It is not.
- the feature maps output from the first convolution layer 310 are input to the first activation layer 320 .
- the first activation layer 320 may assign non-linear characteristics to each feature map.
- the first activation layer 320 may include, but is not limited to, a sigmoid function, a Tanh function, a Rectified Linear Unit (ReLU) function, and the like.
- Giving the nonlinear characteristics in the first activation layer 320 means changing and outputting some sample values of the feature map, which is an output of the first convolution layer 310 . At this time, the change is performed by applying nonlinear characteristics.
- the first activation layer 320 determines whether to transfer sample values of feature maps output from the first convolution layer 310 to the second convolution layer 330 . For example, certain sample values among sample values of feature maps are activated by the first activation layer 320 and transferred to the second convolution layer 330, and certain sample values are activated by the first activation layer 320. It is inactivated and not transmitted to the second convolution layer 330 . The unique characteristics of the second image 135 indicated by the feature maps are emphasized by the first activation layer 320 .
- the feature maps 325 output from the first activation layer 320 are input to the second convolution layer 330 .
- One of the feature maps 325 shown in FIG. 3 is a result of processing the feature map 450 described with reference to FIG. 4 in the first activation layer 320 .
- 3X3X4 displayed on the second convolution layer 330 exemplifies convolution processing on the input feature maps 325 using four filter kernels having a size of 3 ⁇ 3.
- the output of the second convolution layer 330 is input to the second activation layer 340 .
- the second activation layer 340 may impart nonlinear characteristics to input data.
- the feature maps 345 output from the second activation layer 340 are input to the third convolution layer 350 .
- 3X3X1 displayed on the third convolution layer 350 shown in FIG. 3 illustrates convolution processing to create one output image using one filter kernel having a size of 3x3.
- the third convolution layer 350 is a layer for outputting a final image and generates one output using one filter kernel. According to the example of the present disclosure, the third convolution layer 350 may output the third image 145 through a convolution operation.
- DNN setting information indicating the number of filter kernels of the first convolution layer 310, the second convolution layer 330, and the third convolution layer 350 of the second DNN 300, parameters of the filter kernels, etc.
- there may be a plurality of DNN configuration information and the plurality of DNN configuration information must be associated with the plurality of DNN configuration information of the first DNN. Association between the plurality of DNN configuration information of the second DNN and the plurality of DNN configuration information of the first DNN may be implemented through association learning of the first DNN and the second DNN.
- FIG. 3 shows that the second DNN 300 includes three convolution layers 310, 330, and 350 and two activation layers 320 and 340, this is only an example and an implementation example. Accordingly, the number of convolution layers and activation layers may be variously changed. Also, depending on implementation, the second DNN 300 may be implemented through a recurrent neural network (RNN). This case means changing the CNN structure of the second DNN 300 according to the example of the present disclosure into an RNN structure.
- RNN recurrent neural network
- the AI upscaler 236 may include at least one arithmetic logic unit (ALU) for the above-described convolution operation and activation layer operation.
- An ALU may be implemented as a processor.
- the ALU may include a multiplier performing a multiplication operation between the sample values of the feature map output from the second image 135 or the previous layer and the sample values of the filter kernel, and an adder adding the resultant values of the multiplication.
- the ALU is a multiplier that multiplies input sample values with weights used in a predetermined sigmoid function, Tanh function, or ReLU function, and compares the multiplication result with a predetermined value to obtain an input sample value.
- a comparator for determining whether to transfer to the next layer may be included.
- the AI setting unit 238 determines an up-scale target and the AI up-scale unit 236 AI up-scales the second image 135 according to the up-scale target will be described.
- the AI setting unit 238 may store a plurality of DNN setting information that can be set in the second DNN.
- the DNN configuration information may include information on at least one of the number of convolution layers included in the second DNN, the number of filter kernels for each convolution layer, and parameters of each filter kernel.
- a plurality of DNN configuration information may correspond to various upscale targets, and the second DNN may operate based on the DNN configuration information corresponding to a specific upscale target.
- the second DNN may have a different structure according to DNN configuration information.
- the second DNN may include three convolutional layers according to certain DNN configuration information, and the second DNN may include four convolutional layers according to other DNN configuration information.
- the DNN configuration information may include only parameters of a filter kernel used in the second DNN.
- the structure of the second DNN is not changed, but only parameters of an internal filter kernel may be changed according to the DNN configuration information.
- the AI setting unit 238 may obtain DNN setting information for AI upscaling of the second image 135 from among a plurality of DNN setting information.
- Each of the plurality of DNN setting information used herein is information for acquiring the third image 145 of a predetermined resolution and/or a predetermined quality, and is trained in association with the first DNN.
- any one DNN setting information among a plurality of DNN setting information is the third image 145 having a resolution twice as large as the resolution of the second image 135, for example, the second image 145 of 2K (2048*1080). It may include information for acquiring the third image 145 of 4K (4096*2160), which is twice as large as the second image 135, and the other DNN setting information is 4 times higher than the resolution of the second image 135.
- Information for acquiring the third image 145 of twice the resolution, for example, the third image 145 of 8K (8192 * 4320) four times larger than the second image 135 of 2K (2048 * 1080) may include
- Each of the plurality of DNN setting information is created in association with the DNN setting information of the first DNN of the AI encoding apparatus 700, and the AI setting unit 238 determines the enlargement ratio corresponding to the reduction ratio of the DNN setting information of the first DNN. Accordingly, one DNN setting information among a plurality of DNN setting information is obtained. To this end, the AI setting unit 238 needs to check the information of the first DNN. In order for the AI setting unit 238 to check the information of the first DNN, the AI decoding apparatus 200 according to an embodiment receives AI data including information of the first DNN from the AI encoding apparatus 700.
- the AI setting unit 238 uses the information received from the AI encoding device 700 to check information targeted by the DNN setting information of the first DNN used to obtain the first image 115, and , it is possible to obtain DNN setting information of the second DNN trained in association with it.
- DNN setting information for AI upscaling of the second image 135 is obtained from among the plurality of DNN setting information
- the DNN setting information is transmitted to the AI upscaling unit 236, and the second DNN operating according to the DNN setting information Based on the input data may be processed.
- the AI upscaler 236 obtains any one DNN setting information
- the first convolution layer 310 and the second convolution layer 330 of the second DNN 300 shown in FIG. ) and the third convolution layer 350 the number of filter kernels included in each layer and parameters of the filter kernels are set to values included in the obtained DNN configuration information.
- the AI upscaler 236 sets the parameters of the 3 X 3 filter kernel used in any one convolution layer of the second DNN shown in FIG. 4 to ⁇ 1, 1, 1, 1, 1, 1 , 1, 1, 1 ⁇ , and when there is a change in the DNN configuration information, the parameters of the filter kernel included in the changed DNN configuration information are ⁇ can be replaced.
- the AI setting unit 238 may obtain DNN setting information for upscaling the second image 135 from among a plurality of DNN setting information based on information included in the AI data, which is used to obtain the DNN setting information.
- AI data is explained in detail.
- the AI setting unit 238 may obtain DNN setting information for upscaling the second image 135 from among a plurality of DNN setting information based on the difference information included in the AI data. For example, based on the difference information, the resolution (eg, 4K (4096*2160)) of the original image 105 is higher than the resolution (eg, 2K (2048*1080)) of the first image 115. If it is confirmed that the resolution is twice as large, the AI setting unit 238 may obtain DNN setting information capable of doubling the resolution of the second image 135 .
- the resolution eg, 4K (4096*2160)
- the AI setting unit 238 may obtain DNN setting information capable of doubling the resolution of the second image 135 .
- the AI setting unit 238 provides DNN setting information for AI up-scaling the second image 135 of the plurality of DNN setting information based on the information related to the first image 115 included in the AI data. can be obtained.
- the AI setting unit 238 may pre-determine a mapping relationship between image related information and DNN setting information, and obtain DNN setting information mapped to the first image 115 related information.
- FIG. 5 is an exemplary diagram illustrating a mapping relationship between various pieces of image-related information and various pieces of DNN setting information.
- the AI encoding/AI decoding process of an embodiment of the present disclosure does not consider only a change in resolution.
- DNN setting information individually or collectively considers resolutions such as SD, HD, and Full HD, bit rates such as 10Mbps, 15Mbps, and 20Mbps, and codec information such as AV1, H.264, and HEVC. of can be made.
- resolutions such as SD, HD, and Full HD
- bit rates such as 10Mbps, 15Mbps, and 20Mbps
- codec information such as AV1, H.264, and HEVC.
- the first image received in the AI decoding process ( 115) DNN setting information for AI up-scaling of the second image 135 may be obtained based on the related information.
- the AI setting unit 238 matches the image related information shown on the left side of the table shown in FIG. 5 with the DNN setting information shown on the right side of the table, so that the DNN setting information according to the image related information can be used.
- the AI setting unit 238 may obtain A DNN setting information from among a plurality of DNN setting information.
- the AI setting unit 238 may obtain B DNN setting information from among a plurality of DNN setting information.
- the AI setting unit 238 obtains C DNN setting information among a plurality of DNN setting information, the resolution of the first image 115 is Full HD, and the first When it is confirmed that the bit rate of the image data obtained as a result of the first encoding of the image 115 is 15 Mbps and that the first image 115 is first encoded with the HEVC codec, the AI setting unit 238 provides a plurality of DNN setting information. Of DNN configuration information can be obtained.
- the bit rate of image data obtained as a result of the first encoding of the first image 115 is 20 Mbps or 15 Mbps
- one of C DNN setting information and D DNN setting information is selected.
- the fact that the bit rates of the image data are different means that the image quality of the restored image is different from each other.
- the first DNN and the second DNN may be jointly trained based on a predetermined image quality, and accordingly, the AI setting unit 238 sets the DNN according to the bit rate of the image data representing the image quality of the second image 135. information can be obtained.
- the AI setting unit 238 relates information provided from the first decoding unit 234 (prediction mode information, motion information, quantization parameter information, etc.) to the first image 115 included in the AI data.
- DNN setting information for AI up-scaling the second image 135 from among a plurality of DNN setting information may be obtained by considering all information.
- the AI setting unit 238 receives quantization parameter information used in the first encoding process of the first image 115 from the first decoding unit 234, and transmits information on the quantization parameter used in the first encoding process of the first image 115 from the AI data.
- a bit rate of image data obtained as a result of encoding may be checked, and DNN setting information corresponding to the quantization parameter and bit rate may be obtained.
- bit rate Even if the bit rate is the same, there may be a difference in the quality of the reconstructed image according to the complexity of the image, and the bit rate is a value representing the entire first image 115 to be first encoded, and the Picture quality may vary. Therefore, considering prediction mode information, motion information, and/or quantization parameters that can be obtained for each frame from the first decoder 234 together, a DNN setting more suitable for the second image 135 than using only AI data information can be obtained.
- the AI data may include an identifier of mutually agreed DNN configuration information.
- the identifier of the DNN configuration information is an upscale target corresponding to the downscale target of the first DNN, and DNN configuration information that has been jointly trained between the first DNN and the second DNN so that the second image 135 can be AI upscaled. This is information for distinguishing a pair of .
- the AI setting unit 238 obtains the identifier of the DNN setting information included in the AI data, and then obtains the DNN setting information corresponding to the identifier of the DNN setting information, and the AI upscaler 236 converts the corresponding DNN setting information.
- the second image 135 can be upscaled using AI.
- an identifier indicating each of a plurality of DNN configuration information configurable in the first DNN and an identifier indicating each of a plurality of DNN configuration information configurable in the second DNN may be designated in advance.
- the same identifier may be designated for a pair of DNN configuration information that can be set for each of the first DNN and the second DNN.
- the AI data may include an identifier of DNN setting information set in the first DNN for AI downscaling of the original video 105 .
- the AI setting unit 238 Upon receiving the AI data, the AI setting unit 238 acquires DNN setting information indicated by an identifier included in the AI data among a plurality of DNN setting information, and the AI upscaling unit 236 uses the corresponding DNN setting information to obtain DNN setting information.
- the image 135 may be AI upscaled.
- AI data may include DNN configuration information.
- the AI setting unit 238 obtains DNN setting information included in the AI data, and the AI upscaling unit 236 may AI up-scale the second image 135 using the corresponding DNN setting information.
- DNN setting information eg, the number of convolution layers, the number of filter kernels for each convolution layer, parameters of each filter kernel, etc.
- the AI setting unit 238 obtains DNN setting information by combining some of the lookup table values based on the information included in the AI data, and the AI upscaling unit 236 uses the corresponding DNN setting information to generate a second image. (135) can also be AI upscaled.
- the AI configuration unit 238 may obtain DNN configuration information corresponding to the determined structure of the DNN, for example, filter kernel parameters, when the structure of the DNN corresponding to the upscale target is determined.
- the AI setting unit 238 obtains DNN setting information of the second DNN through AI data including information related to the first DNN, and the AI upscaling unit 236 sets the corresponding DNN setting information.
- the second image 135 is upscaled by AI through the second DNN, which can reduce the amount of memory and computation compared to upscaling by directly analyzing the characteristics of the second image 135 .
- the AI setting unit 238 may independently obtain DNN setting information for each of a predetermined number of frames, or a common DNN for all frames. It is also possible to obtain setting information.
- FIG. 6 is a diagram illustrating a second image 135 composed of a plurality of frames.
- the second image 135 may include frames corresponding to t0 to tn.
- the AI setting unit 238 obtains DNN setting information of the second DNN through AI data, and the AI upscaling unit 236 AI frames corresponding to t0 to tn based on the corresponding DNN setting information. can be upscaled. That is, frames corresponding to t0 to tn may be AI upscaled based on common DNN configuration information.
- the AI setting unit 238 obtains 'A' DNN setting information from AI data for some of the frames corresponding to t0 to tn, for example, frames corresponding to t0 to ta And 'B' DNN setting information can be obtained from AI data for frames corresponding to ta+1 to tb.
- the AI setting unit 238 may obtain 'C' DNN setting information from AI data for frames corresponding to tb+1 to tn.
- the AI setting unit 238 independently acquires DNN setting information for each group including a predetermined number of frames among a plurality of frames, and the AI upscaling unit 236 independently acquires the frames included in each group. AI can be upscaled with the acquired DNN setting information.
- the AI setting unit 238 may independently obtain DNN setting information for each frame constituting the second image 135 .
- the AI setting unit 238 obtains DNN setting information in relation to the first frame and DNN setting information in relation to the second frame.
- DNN configuration information may be obtained in relation to the third frame. That is, DNN configuration information can be obtained independently for each of the first frame, the second frame, and the third frame.
- DNN setting information is obtained based on information (prediction mode information, motion information, quantization parameter information, etc.) provided from the above-described first decoding unit 234 and information related to the first image 115 included in the AI data Depending on the method, DNN setting information may be obtained independently for each frame constituting the second image 135 . This is because mode information, quantization parameter information, and the like can be independently determined for each frame constituting the second image 135 .
- the AI data may include information indicating up to which frame the DNN setting information obtained based on the AI data is valid. For example, if information that the DNN setting information is valid until frame ta is included in the AI data, the AI setting unit 238 obtains the DNN setting information based on the AI data, and the AI upscaler 236 AI upscales t0 to ta frames with corresponding DNN configuration information. And, if information that the DNN setting information is valid until the tn frame is included in the other AI data, the AI setting unit 238 obtains the DNN setting information based on the other AI data, and the AI upscaler 236 can AI upscale ta + 1 to tn frames with the acquired DNN configuration information.
- the AI encoding apparatus 700 for AI encoding of the original video 105 will be described with reference to FIG. 7 .
- FIG. 7 is a block diagram showing the configuration of an AI encoding apparatus 700 according to an embodiment.
- an AI encoding apparatus 700 may include an AI encoding unit 710 and a transmission unit 730.
- the AI encoder 710 may include an AI downscaling unit 712, a first encoding unit 714, a data processing unit 716, and an AI setting unit 718.
- FIG. 7 shows the AI encoder 710 and the transmitter 730 as separate devices
- the AI encoder 710 and the transmitter 730 may be implemented by one processor.
- it may be implemented as a dedicated processor, or may be implemented through a combination of S/W and a general-purpose processor such as an AP, CPU, or GPU.
- a dedicated processor may include a memory for implementing an embodiment of the present disclosure or a memory processing unit for using an external memory.
- the AI encoder 710 and the transmitter 730 may be composed of a plurality of processors. In this case, it may be implemented with a combination of dedicated processors, or it may be implemented with a combination of S/W and a plurality of general-purpose processors such as APs, CPUs, or GPUs.
- the first encoding unit 714 is composed of a first processor, and the AI downscaling unit 712, the data processing unit 716 and the AI setting unit 718 are configured by a second processor different from the first processor. and the transmitter 730 may be implemented as a third processor different from the first processor and the second processor. Performs first encoding of and transfers AI-encoded data to the transmission unit 730. The transmitter 730 transmits the AI-encoded data to the AI decoding device 200.
- the image data includes data obtained as a result of the first encoding of the first image 115 .
- the image data may include data obtained based on pixel values in the first image 115, eg, residual data that is a difference between the first image 115 and prediction data of the first image 115.
- the image data includes information used in the first encoding process of the first image 115 .
- the image data may include prediction mode information used for first encoding the first image 115, motion information, and quantization parameter related information used for first encoding the first image 115. .
- the AI data includes information enabling the AI upscaler 236 to AI upscale the second image 135 to an upscale target corresponding to the downscale target of the first DNN.
- the AI data may include difference information between the original image 105 and the first image 115 .
- AI data may include information related to the first image 115 .
- the information related to the first image 115 includes the resolution of the first image 115, the bit rate of the image data obtained as a result of the first encoding of the first image 115, and the first image 115 used in the first encoding. It may include information on at least one of the codec types.
- the AI data may include an identifier of mutually agreed DNN setting information so that the second image 135 can be AI upscaled to an upscale target corresponding to a downscale target of the first DNN. .
- the AI data may include DNN setting information that can be set in the second DNN.
- the AI downscaling unit 712 may obtain the AI downscaled first image 115 from the original image 105 through the first DNN.
- the AI downscaling unit 712 may AI downscale the original video 105 using the DNN setting information provided from the AI setting unit 718 .
- the AI setting unit 718 may determine a downscale target of the original image 105 based on a predetermined criterion.
- the AI setting unit 718 may store a plurality of DNN setting information that can be set in the first DNN.
- the AI setting unit 718 obtains DNN setting information corresponding to a downscale target from among a plurality of DNN setting information, and provides the obtained DNN setting information to the AI downscaling unit 712.
- Each of the plurality of DNN setting information may be trained to acquire the first image 115 having a predetermined resolution and/or a predetermined quality.
- any one DNN setting information among a plurality of DNN setting information is a first image 115 having a resolution that is 1/2 smaller than the resolution of the original image 105, for example, 4K (4096*2160) It may include information for acquiring the first image 115 of 2K (2048 * 1080), which is 1/2 smaller than the original image 105 of , and another piece of DNN setting information is the resolution of the original image 105.
- a first image 115 having a resolution 1/4 smaller than that of the first image 115 may include information for obtaining.
- the AI setting unit 718 may provide the AI downscaling unit 712 with DNN setting information obtained by combining selected parts of the lookup table values according to the downscale target.
- the AI configuration unit 718 may determine the structure of the DNN corresponding to the downscale target and obtain DNN configuration information corresponding to the determined structure of the DNN, for example, filter kernel parameters.
- the plurality of DNN setting information for AI downscaling of the original video 105 may have an optimized value by performing joint training of the first DNN and the second DNN.
- each DNN configuration information includes at least one of the number of convolution layers included in the first DNN, the number of filter kernels for each convolution layer, and parameters of each filter kernel.
- the AI downscaling unit 712 sets the first DNN with the DNN setting information determined for AI downscaling of the original image 105, and converts the first image 115 of a predetermined resolution and/or quality to the first DNN. can be obtained through When DNN setting information for AI downscaling of the original image 105 is obtained from among a plurality of DNN setting information, each layer in the first DNN may process input data based on information included in the DNN setting information. .
- the downscale target may indicate, for example, how much the first image 115 whose resolution is reduced from the original image 105 should be acquired.
- the AI setting unit 718 obtains one or more pieces of input information.
- the input information includes a target resolution of the first image 115, a target bitrate of image data, a bitrate type of image data (eg, variable bitrate type, constant bitrate type, average bitrate type, etc.), A color format to which AI downscaling is applied (luminance component, chrominance component, red component, green component, blue component, etc.), codec type for first encoding of the first image 115, compression history information, original image 105 It may include at least one of a resolution of and a type of the original image 105 .
- One or more pieces of input information may include information previously stored in the AI encoding device 700 or input from a user.
- the AI setting unit 718 controls the operation of the AI downscaling unit 712 based on the input information.
- the AI setting unit 718 may determine a downscaling target according to input information and provide DNN setting information corresponding to the determined downscaling target to the AI downscaling unit 712 .
- the AI setting unit 718 transfers at least a portion of the input information to the first encoding unit 714 so that the first encoding unit 714 has a bit rate of a specific value, a bit rate of a specific type, and a specific codec.
- the first image 115 may be subjected to first encoding.
- the AI setting unit 718 is a compression rate (eg, resolution difference between the original video 105 and the first video 115, target bit rate), compression quality (eg, bit rate type ), compression history information, and the type of the original video 105, the downscale target may be determined.
- a compression rate eg, resolution difference between the original video 105 and the first video 115, target bit rate
- compression quality eg, bit rate type
- compression history information e.g, compression history information, and the type of the original video 105, the downscale target may be determined.
- the AI setting unit 718 may determine a downscale target based on a compression rate or compression quality that is preset or input from a user.
- the AI setting unit 718 may determine a downscale target using compression history information stored in the AI encoding device 700. For example, according to the compression history information available to the AI encoding apparatus 700, a user's preferred encoding quality or compression rate may be determined, and a downscale target may be determined according to the encoding quality determined based on the compression history information. can For example, the resolution, image quality, etc. of the first image 115 may be determined according to the most frequently used encoding quality according to compression history information.
- the AI setting unit 718 determines the encoding quality that has been used more than a predetermined threshold value (eg, the average quality of the encoding qualities that have been used more than a predetermined threshold value) according to the compression history information. Based on this, a downscale target may be determined.
- a predetermined threshold value eg, the average quality of the encoding qualities that have been used more than a predetermined threshold value
- the AI setting unit 718 may determine a downscale target based on the resolution and type (eg, file format) of the original video 105 .
- the AI setting unit 718 independently obtains DNN setting information for each of a predetermined number of frames, and uses the independently acquired DNN setting information for AI downscaling. It may be provided as part 712 .
- the AI setting unit 718 may divide the frames constituting the original image 105 into a predetermined number of groups, and independently obtain DNN setting information for each group. The same or different DNN setting information may be obtained for each group. The number of frames included in the groups may be the same or different for each group.
- the AI setting unit 718 may independently determine DNN setting information for each frame constituting the original image 105 .
- the same or different DNN configuration information may be obtained for each frame.
- FIG. 8 is an exemplary diagram illustrating a first DNN 800 for AI downscaling of an original image 105 .
- the original image 105 is input to the first convolution layer 810 .
- the first convolution layer 810 performs convolution on the original image 105 using 32 filter kernels of 5x5 size.
- 32 feature maps generated as a result of the convolution process are input to the first activation layer 820 .
- the first activation layer 820 may assign non-linear characteristics to 32 feature maps.
- the first activation layer 820 determines whether to transfer sample values of feature maps output from the first convolution layer 810 to the second convolution layer 830 . For example, certain sample values among sample values of feature maps are activated by the first activation layer 820 and transferred to the second convolution layer 830, and certain sample values are activated by the first activation layer 820. It is inactivated and not transmitted to the second convolution layer 830. Information indicated by feature maps output from the first convolution layer 810 is emphasized by the first activation layer 820 .
- the output 825 of the first activation layer 820 is input to the second convolution layer 830 .
- the second convolution layer 830 performs convolution processing on input data using 32 filter kernels of 5x5 size.
- the 32 feature maps output as a result of the convolution process are input to the second activation layer 840, and the second activation layer 840 may assign nonlinear characteristics to the 32 feature maps.
- the output 845 of the second activation layer 840 is input to the third convolution layer 850 .
- the third convolution layer 850 performs convolution processing on the input data using one filter kernel having a size of 5 x 5. As a result of the convolution process, one image may be output from the third convolution layer 850 .
- the third convolution layer 850 is a layer for outputting a final image and obtains one output by using one filter kernel. According to the example of the present disclosure, the third convolution layer 850 may output the first image 115 through a convolution operation result.
- DNN setting information indicating the number of filter kernels of the first convolution layer 810, the second convolution layer 830, and the third convolution layer 850 of the first DNN 800, parameters of the filter kernels, etc.
- the first DNN 800 includes three convolution layers 810, 830, and 750 and two activation layers 820 and 840, but this is only an example and an implementation example. Accordingly, the number of convolution layers and activation layers may be variously changed. Also, depending on implementation, the first DNN 800 may be implemented through a recurrent neural network (RNN). This case means changing the CNN structure of the first DNN 800 according to the example of the present disclosure into an RNN structure.
- RNN recurrent neural network
- the AI downscaling unit 712 may include at least one ALU for convolution operation and activation layer operation.
- An ALU may be implemented as a processor.
- the ALU may include a multiplier that performs a multiplication operation between the sample values of the feature map output from the original image 105 or the previous layer and the sample values of the filter kernel, and an adder that adds the resultant values of the multiplication. there is.
- the ALU is a multiplier that multiplies input sample values with weights used in a predetermined sigmoid function, Tanh function, or ReLU function, and compares the multiplication result with a predetermined value to obtain an input sample value.
- a comparator for determining whether to transfer to the next layer may be included.
- the AI setting unit 718 transfers AI data to the data processing unit 716 .
- the AI data includes information enabling the AI upscaler 236 to AI upscale the second image 135 to an upscale target corresponding to the downscale target of the first DNN.
- the first encoder 714 receiving the first image 115 from the AI downscaler 712 first encodes the first image 115 according to a frequency conversion-based image compression method to obtain the first image 115 ) can reduce the amount of information.
- a predetermined codec eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1
- Image data is obtained according to the rules of a predetermined codec, that is, syntax.
- the image data includes residual data that is a difference between the first image 115 and predicted data of the first image 115, prediction mode information used to first encode the first image 115, motion information, and Information related to a quantization parameter used for first encoding the first image 115 may be included.
- Image data obtained as a result of the first encoding of the first encoder 714 is provided to the data processor 716 .
- the data processing unit 716 generates AI-encoded data including the video data received from the first encoding unit 714 and the AI data received from the AI setting unit 718 .
- the data processor 716 may generate AI-encoded data including image data and AI data in a separated state.
- AI data may be included in a Vendor Specific InfoFrame (VSIF) within an HDMI stream.
- VSIF Vendor Specific InfoFrame
- the data processing unit 716 may include AI data in image data obtained as a result of the first encoding by the first encoding unit 714 and generate AI-encoded data including the corresponding image data. .
- the data processor 716 may combine a bitstream corresponding to image data and a bitstream corresponding to AI data to generate image data in the form of a bitstream.
- the data processing unit 716 may express AI data as bits having a value of 0 or 1, that is, as a bitstream.
- the data processor 716 may include a bitstream corresponding to AI data in supplemental enhancement information (SEI), which is an additional information area of the bitstream obtained as a result of the first encoding.
- SEI Supplemental Enhancement Information
- AI-encoded data is transmitted to the transmitter 730.
- the transmission unit 730 transmits the AI-encoded data obtained as a result of the AI-encoding through a network.
- the AI-encoded data is a hard disk, a magnetic medium such as a floppy disk and a magnetic tape, such as a CD-ROM and a DVD. It may be stored in a data storage medium including an optical recording medium, a magneto-optical medium such as a floptical disk, and the like.
- FIG. 9 is a diagram showing the structure of AI encoded data 900 according to an embodiment.
- the AI data 912 and the image data 932 may be separately included in the AI encoded data 900.
- the AI-encoded data 900 may be in a container format such as MP4, AVI, MKV, or FLV.
- the AI encoded data 900 may be composed of a metadata box 910 and a media data box 930.
- the metadata box 910 includes information about the video data 932 included in the media data box 930 .
- the metadata box 910 may include information about the type of the first image 115, the type of codec used for encoding the first image 115, and the playback time of the first image 115. there is.
- AI data 912 may be included in the metadata box 910 .
- the AI data 912 may be encoded according to an encoding method provided by a predetermined container format and stored in the metadata box 910 .
- the media data box 930 may include image data 932 generated according to the syntax of a predetermined image compression method.
- FIG. 10 is a diagram showing the structure of AI encoded data 1000 according to another embodiment.
- AI data 1034 may be included in image data 1032 .
- the AI-encoded data 1000 may include a metadata box 1010 and a media data box 1030.
- the AI data box 1010 contains AI data.
- Data 1034 may not be included.
- the media data box 1030 includes image data 1032 including AI data 1034.
- the AI data 1034 may be included in the additional information area of the image data 1032 .
- 11 is a diagram for explaining a method of training the first DNN 800 and the second DNN 300.
- the AI-encoded original image 105 through the AI encoding process is restored to the third image 145 through the AI decoding process.
- the third image 145 obtained as a result of AI decoding and the original image 105
- correlation between the AI encoding process and the AI decoding process is required. That is, information lost in the AI encoding process should be able to be restored in the AI decoding process, and for this purpose, joint training of the first DNN 800 and the second DNN 300 is required.
- the quality loss information 1130 is used for both training of the first DNN 800 and the second DNN 300 .
- an original training image 1101 is an image to be AI downscaled
- a first training image 1102 is an AI downscaled image from the original training image 1101. It's a video.
- a third training image 1104 is an AI upscaled image from the first training image 1102 .
- the original training image 1101 includes a still image or a video consisting of a plurality of frames.
- the original training image 1101 may include a luminance image extracted from a still image or a video consisting of a plurality of frames.
- the original training image 1101 may include a patch image extracted from a still image or a video consisting of a plurality of frames.
- the first training image 1102, the second training image 1104, and the third training image 1104 also consist of a plurality of frames.
- the first training image 1102 and the second training image are generated through the first DNN 800 and the second DNN 300.
- a plurality of frames of the third training image 1104 may be sequentially obtained.
- the original training image 1101 is input to the first DNN 800.
- the original training image 1101 input to the first DNN 800 is downscaled by AI and output as the first training image 1102, and the first training image 1102 is input to the second DNN 300.
- a third training image 1104 is output.
- a first training image 1102 is input to the second DNN 300, and according to an embodiment, a first training image 1102 obtained through a first encoding and a first decoding process
- a second training image may be input to the second DNN 300 .
- any one of MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, and AV1 codec may be used.
- MPEG-2, H.264, MPEG-4, HEVC, VC-1 , VP8, VP9, and AV1 may be used.
- a legacy downscaled downscaled training image 1103 is obtained from the original training image 1101.
- the legacy downscale may include at least one of a bilinear scale, a bicubic scale, a lanczos scale, and a stair step scale.
- a reduced training image 1103 preserving the structural features of the original training image 1101 is obtained.
- the first DNN 800 and the second DNN 300 may be set with predetermined DNN setting information.
- structural loss information 1110, complexity loss information 1120, and quality loss information 1130 may be determined.
- the structural loss information 1110 may be determined based on a comparison result between the reduced training image 1103 and the first training image 1102 .
- the structural loss information 1110 may correspond to a difference between structural information of the reduced training image 1103 and structural information of the first training image 1102 .
- Structural information may include various features extractable from an image, such as luminance, contrast, and histogram of the image.
- the structural loss information 1110 represents how much structural information of the original training image 1101 is maintained in the first training image 1102 . As the structural loss information 1110 is smaller, the structural information of the first training image 1102 becomes similar to the structural information of the original training image 1101.
- the complexity loss information 1120 may be determined based on the spatial complexity of the first training image 1102 . In one example, as the spatial complexity, a total variance value of the first training image 1102 may be used.
- the complexity loss information 1120 is related to the bitrate of image data obtained by first encoding the first training image 1102 . The smaller the complexity loss information 1120 is, the smaller the bit rate of image data is defined.
- the quality loss information 1130 may be determined based on a comparison result between the original training image 1101 and the third training image 1104 .
- the quality loss information 1130 includes an L1-norm value, an L2-norm value, a Structural Similarity (SSIM) value, and a Peak Signal-To-Value (PSNR-HVS) for the difference between the original training image 1101 and the third training image 1104.
- -It may include at least one of a Noise Ratio-Human Vision System (MS-SSIM) value, a Multiscale SSIM (MS-SSIM) value, a Variance Inflation Factor (VIF) value, and a Video Multimethod Assessment Fusion (VMAF) value.
- the quality loss information 1130 indicates how similar the third training image 1104 is to the original training image 1101 . The smaller the quality loss information 1130 is, the more similar the third training image 1104 is to the original training image 1101.
- structural loss information 1110, complexity loss information 1120, and quality loss information 1130 are used to train the first DNN 800, and the quality loss information 1130 is used for training the second DNN ( 300) is used for training. That is, the quality loss information 1130 is used for training both the first DNN 800 and the second DNN 300 .
- the first DNN 800 may update parameters such that final loss information determined based on the structural loss information 1110 , complexity loss information 1120 , and quality loss information 1130 is reduced or minimized. Also, the second DNN 300 may update a parameter such that the quality loss information 1130 is reduced or minimized.
- Final loss information for training of the first DNN 800 and the second DNN 300 may be determined as shown in Equation 1 below.
- LossDS represents final loss information to be reduced or minimized for training of the first DNN 800
- LossUS represents final loss information to be reduced or minimized for training of the second DNN 300.
- a, b, c, and d may correspond to predetermined weights.
- the first DNN 800 updates parameters in a direction in which LossDS of Equation 1 decreases
- the second DNN 300 updates parameters in a direction in which LossUS decreases.
- the parameters of the first DNN 800 are updated according to the LossDS derived in the training process
- the first training image 1102 obtained based on the updated parameters is different from the first training image 1102 in the previous training process.
- the third training image 1104 also becomes different from the third training image 1104 in the previous training process. If the third training image 1104 is different from the third training image 1104 in the previous training process, the quality loss information 1130 is also newly determined, and the second DNN 300 updates parameters accordingly.
- LossDS is also newly determined, so the first DNN 800 updates parameters according to the newly determined LossDS. That is, updating parameters of the first DNN 800 causes updating of parameters of the second DNN 300 , and updating parameters of the second DNN 300 causes updating of parameters of the first DNN 800 .
- the parameters of the first DNN 800 and the parameters of the second DNN 300 are They can be correlated with each other and optimized.
- LossUS is determined according to quality loss information 1130, but this is an example, and LossUS includes at least one of structural loss information 1110 and complexity loss information 1120, It may be determined based on the quality loss information 1130 .
- the AI setting unit 238 of the AI decoding apparatus 200 and the AI setting unit 718 of the AI encoding apparatus 700 have been described as storing a plurality of DNN setting information. A method of training each of the plurality of DNN setting information stored in the setting unit 718 will be described.
- the similarity between the structural information of the first training image 1102 and the structural information of the original training image 1101 (structural loss information 1110) ), the bit rate (complexity loss information 1120) of image data obtained as a result of the first encoding of the first training image 1102 and the difference between the third training image 1104 and the original training image 1101 (quality loss Parameters are updated in consideration of the information 1130).
- the first training image 1102 which is similar to the structural information of the original training image 1101 and has a small bit rate of the image data obtained when the first encoding is performed, can be obtained, and the first training image Parameters of the first DNN 800 may be updated so that the second DNN 300 that AI upscales 1102 can obtain a third training image 1104 similar to the original training image 1101.
- the directions in which the parameters of the first DNN 800 are optimized become different.
- parameters of the first DNN 800 may be updated with more importance placed on the lowering of the bitrate than the quality of the third training image 1104 .
- increasing the quality of the third training image 1104 is more important than increasing the bit rate or maintaining structural information of the original training image 1101. 1 Parameters of the DNN 800 may be updated.
- the direction in which the parameters of the first DNN 800 are optimized may be different according to the type of codec used to first encode the first training image 1102 . This is because the second training image to be input to the second DNN 300 may vary according to the type of codec.
- the parameters of the first DNN 800 and the parameters of the second DNN 300 are linked based on the weight a, the weight b, the weight c, and the type of codec for the first encoding of the first training image 1102. so that it can be updated. Therefore, after determining the weight a, the weight b, and the weight c as predetermined values, and determining the type of codec as a predetermined type, and then training the first DNN 800 and the second DNN 300, they are linked to each other. Optimized parameters of the first DNN 800 and parameters of the second DNN 300 may be determined.
- the parameters of the first DNN 800 are optimized in association with each other. and parameters of the second DNN 300 may be determined.
- a plurality of DNN setting information trained in conjunction with each other is It can be determined by the DNN (800) and the second DNN (300).
- a plurality of DNN setting information of the first DNN 800 and the second DNN 300 may be mapped to first image related information.
- the first training image 1102 output from the first DNN 800 is first encoded with a specific codec according to a specific bit rate, and the bitstream obtained as a result of the first encoding is first encoded.
- the acquired second training image may be input to the second DNN 300.
- the first DNN settings mapped to the resolution of the training image 1102, the type of codec used for the first encoding of the first training image 1102, and the bitrate of the bitstream obtained as a result of the first encoding of the first training image 1102. information pairs can be determined.
- the bit rate of the bit stream obtained according to the resolution of the first training image 1102, the type of codec used for the first encoding of the first training image 1102, and the first encoding of the first training image 1102 varies.
- a mapping relationship between a plurality of pieces of DNN setting information of the first DNN 800 and the second DNN 300 and the first image related information may be determined.
- FIG. 12 is a diagram for explaining a training process of the first DNN 800 and the second DNN 300 by the training apparatus 1200.
- Training of the first DNN 800 and the second DNN 300 described with reference to FIG. 11 may be performed by the training device 1200 .
- the training device 1200 includes a first DNN 800 and a second DNN 300 .
- the training device 1200 may be, for example, the AI encoding device 700 or a separate server.
- DNN setting information of the second DNN 300 obtained as a result of training is stored in the AI decoding apparatus 200.
- the training apparatus 1200 initially sets DNN configuration information of the first DNN 800 and the second DNN 300 (S1240 and S1245). Accordingly, the first DNN 800 and the second DNN 300 may operate according to predetermined DNN setting information.
- the DNN setting information includes at least one of the number of convolution layers included in the first DNN 800 and the second DNN 300, the number of filter kernels for each convolution layer, the size of each filter kernel for each convolution layer, and the parameters of each filter kernel. It may contain information about one.
- the training device 1200 inputs the original training image 1101 to the first DNN 800 (S1250).
- the original training image 1101 may include at least one frame constituting a still image or a moving image.
- the first DNN 800 processes the original training image 1101 according to the initially set DNN setting information, and outputs the first training image 1102, which is downscaled by AI, from the original training image 1101 (S1255).
- 12 shows that the first training image 1102 output from the first DNN 800 is directly input to the second DNN 300, but the first training image 1102 output from the first DNN 800 ) may be input to the second DNN 300 by the training apparatus 1200.
- the training device 1200 may first encode and first decode the first training image 1102 using a predetermined codec, and then input the second training image to the second DNN 300 .
- the second DNN 300 processes the first training image 1102 or the second training image according to the initially set DNN setting information, and obtains a third AI upscaled image from the first training image 1102 or the second training image.
- the training image 1104 is output (S1260).
- the training device 1200 calculates complexity loss information 1120 based on the first training image 1102 (S1265).
- the training apparatus 1200 compares the reduced training image 1103 and the first training image 1102 to calculate structural loss information 1110 (S1270).
- the training device 1200 compares the original training image 1101 and the third training image 1104 to calculate quality loss information 1130 (S1275).
- the first DNN 800 updates the initially set DNN configuration information through a back propagation process based on the final loss information (S1280).
- the training apparatus 1200 may calculate final loss information for training of the first DNN 800 based on the complexity loss information 1120 , the structural loss information 1110 , and the quality loss information 1130 .
- the second DNN 300 updates the initially set DNN configuration information through a reverse transcription process based on the quality loss information or the final loss information (S1285).
- the training apparatus 1200 may calculate final loss information for training of the second DNN 300 based on the quality loss information 1130 .
- the training device 1200, the first DNN 800, and the second DNN 300 update DNN setting information while repeating processes S1250 to S1285 until the final loss information is minimized.
- the first DNN 800 and the second DNN operate according to the DNN configuration information updated in the previous process.
- Table 1 shows effects when AI encoding and AI decoding the original video 105 according to an embodiment of the present disclosure and when the original video 105 is encoded and decoded using HEVC.
- An electronic device and/or server described below may reduce the amount of data transmitted and received through a network by performing AI upscaling and/or AI downscaling of video conference images.
- the electronic devices and/or servers described below do not AI downscale/AI upscale all video conference videos collectively, but rather perform video conferencing according to the importance of electronic devices of video conferencing participants, that is, the importance of video conferencing videos.
- By adaptively performing AI upscaling and/or AI downscaling of video a smooth video conference can be conducted.
- the electronic device or the server managing the video conference depends on whether AI downscale or AI upscale is supported and the importance of the electronic device participating in the video conference. It relates to a method of performing AI downscaling or AI upscaling of a video conference video.
- FIG. 13 is a block diagram illustrating a configuration of an electronic device that AI upscales a video conference video according to importance of other electronic devices participating in the video conference according to an embodiment.
- an electronic device 1300 may include a receiving unit 1310, an AI decoding unit 1330, and a display unit 1350.
- the AI decoder 1330 may include a parser 1331, a first decoder 1332, an AI upscaler 1333, and an AI setter 1334.
- the receiver 1310, the AI decoder 1330, and the display unit 1350 are one unit. It can be implemented through a processor. In this case, it may be implemented as a dedicated processor, or may be implemented through a combination of S/W and a general-purpose processor such as an AP, CPU, or GPU. In addition, a dedicated processor may include a memory for implementing an embodiment of the present disclosure or a memory processing unit for using an external memory.
- the receiving unit 1310, the AI decoding unit 1330, and the display unit 1350 may be configured with a plurality of processors. In this case, it may be implemented with a combination of dedicated processors, or it may be implemented with a combination of S/W and a plurality of general-purpose processors such as APs, CPUs, or GPUs.
- the receiver 1310 is implemented as a first processor
- the first decoder 1332 is implemented as a second processor different from the first processor
- the AI setting unit 1334 is implemented by a third processor different from the first processor and the second processor
- the display unit 1350 is implemented by a fourth processor different from the first processor, the second processor, and the third processor.
- the receiving unit 1310 receives AI-encoded data obtained as a result of AI encoding and importance information of other electronic devices participating in the video conference from a server managing the video conference.
- AI-encoded data may be a video file having a file format such as mp4 or mov.
- the AI-encoded data obtained by the receiving unit 1310 from the server is data related to a first image obtained by AI downscaling an original image of another electronic device by the server, or data related to a first image obtained by AI downscaling an original image of another electronic device by another electronic device. It is data.
- the receiving unit 1310 may receive AI-encoded data transmitted from a server managing a video conference through a communication network and importance information of other electronic devices participating in the video conference.
- the receiving unit 1310 outputs AI encoded data and importance information to the AI decoding unit 1330.
- the importance information may be included in AI data of AI-encoded data and transmitted. In one embodiment, the importance information may be received from a server or other electronic device as meta data separately from AI-encoded data.
- the parsing unit 1331 parses the AI-encoded data and transfers the video data generated as a result of the first encoding of the video conference video to the first decoding unit 1332, and the AI data and importance information to the AI setting unit 1334. convey
- the parsing unit 1331 may parse image data and AI data separately included in AI-encoded data.
- the parsing unit 1331 may read a header in the AI-encoded data and distinguish between AI data and image data included in the AI-encoded data.
- AI-encoded data including AI data and image data separated from each other is omitted as described above with reference to FIG. 9 .
- the parsing unit 1331 parses video data from AI-encoded data, extracts AI data from the video data, transfers the AI data to the AI setting unit 1334, and performs first decoding on the remaining video data. It can be delivered to the unit 1332. That is, AI data may be included in video data. For example, AI data may be included in supplemental enhancement information (SEI), which is an additional information area of a bitstream corresponding to video data.
- SEI Supplemental Enhancement information
- the parsing unit 1831 divides a bitstream corresponding to image data into a bitstream to be processed in the first decoding unit 1332 and a bitstream corresponding to AI data, and each of the divided bitstreams It can be output to the first decoder 1332 and the AI setting unit 1334.
- the parsing unit 1331 acquires video data included in AI-encoded data through a predetermined codec (eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1). It can also be confirmed that it is image data. In this case, corresponding information may be transmitted to the first decoder 1332 so that the video data can be processed with the identified codec.
- a predetermined codec eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1
- the first decoding unit 1332 restores a second video corresponding to the first video, which is a video conference video, based on the video data received from the parsing unit 1331 .
- the second image obtained by the first decoder 1332 is provided to the AI upscaler 1333.
- the importance information from the AI setting unit 1334 is also provided to the AI upscaling unit 1333.
- the AI upscaling unit 1333 provides the second image as it is to the display unit 1350 or provides the third image obtained by AI upscaling the second image to the display unit 1350 according to the degree of importance.
- the importance information is information indicating whether a user of an electronic device participating in a video conference is a presenter or a listener, and is information used to determine whether or not to perform AI upscaling on video in a video conference.
- the presenter refers to a participant who explains important conference contents in a video conference and has the right to control the screen
- the listener means a participant who simply listens to the contents explained by the presenter in the video conference and does not have the right to control the screen. Therefore, since the presenter's video conference video contains a relatively large amount of important information, it is important to restore the downscaled video to the original video quality in order to reduce data usage due to its high importance, and the listener's video conference video contains important information. Since is included relatively little, there is no need to restore the downscaled image to the original image due to its low importance.
- the AI upscaler 1333 provides a third image obtained by upscaling the second image by AI to the display unit 1350 when the user of the electronic device having different importance information indicates that the presenter is a presenter. If the user of the electronic device indicates that he or she is a listener, the second image is provided to the display unit 1350. That is, since the presenter's video conference video contains a lot of important information, the third video obtained by AI upscaling is displayed to correspond to the original video of the presenter's electronic device, and the listener's video conference video is less important than the presenter's video conference video. Therefore, the second image is displayed as it is.
- the importance information may be flag information set to 1 if a user of another electronic device is a presenter and set to 0 if the user is a listener.
- the server when starting a video conference, other electronic devices participating in the video conference generate and transmit importance information of the other electronic devices to the server, and the server determines that the electronic device 1300 can use the importance information.
- Priority information may be transmitted to the electronic device 1300 .
- the other electronic device when an input for changing the importance of another electronic device is received during a video conference, the other electronic device generates the changed importance information or updates the importance information and transmits the updated importance information to the server, and the server sends the electronic device 1300 the importance information.
- importance information may be transmitted to the electronic device 1300 . That is, the importance information is not always transmitted, but changed or updated importance information may be transmitted when a video conference starts or when there is a change in importance.
- the electronic device 1300 acquires and stores the importance information of other electronic devices transmitted through the server when the video conference starts, performs AI upscaling based on the importance information, and then changes the importance of other electronic devices. , the changed importance information of the other electronic device is acquired from the server and updated, and AI upscaling is performed based on the updated importance information.
- first decoding-related information such as prediction mode information, motion information, and quantization parameter information may be provided from the first decoding unit 1332 to the AI setting unit 1334.
- the first decoding related information may be used to obtain DNN setting information.
- the AI data provided to the AI setting unit 1334 includes information enabling AI up-scaling of the second image.
- the upscale target of the second image must correspond to the downscale target of the first DNN. Therefore, the AI data must include information capable of identifying the downscale target of the first DNN.
- Specific examples of the information included in the AI data include information about a difference between the resolution of an original video during a video conference and the resolution of a first video obtained by AI downscaling the original video, and information related to the first video.
- the difference information may be expressed as information about a degree of resolution conversion of the first image compared to the original image (eg, resolution conversion rate information).
- the difference information may be expressed only with the resolution information of the original image.
- the resolution information may be expressed as a horizontal/vertical screen size, or as a ratio (16:9, 4:3, etc.) and a size on one axis.
- it may be expressed in the form of an index or flag.
- the first image-related information may include information on at least one of a resolution of the first image, a bit rate of image data obtained as a result of the first encoding of the first image, and a codec type used in the first encoding of the first image.
- the AI setting unit 1334 may determine an upscale target of the second image based on at least one of difference information included in the AI data and information related to the first image.
- the upscaling target may indicate, for example, to what resolution the second image should be upscaled.
- the AI upscaler 1333 AI upscales the second image through a second DNN to obtain a third image corresponding to the upscale target.
- the AI setting unit 1334 determines an up-scale target and the AI up-scale unit 1333 AI up-scales the second video according to the up-scale target will be described.
- the AI setting unit 1334 may store a plurality of DNN setting information that can be set in the second DNN.
- the DNN configuration information may include information on at least one of the number of convolution layers included in the second DNN, the number of filter kernels for each convolution layer, and parameters of each filter kernel.
- a plurality of DNN configuration information may correspond to various upscale targets, and the second DNN may operate based on the DNN configuration information corresponding to a specific upscale target.
- the second DNN may have a different structure according to DNN configuration information.
- the second DNN may include three convolutional layers according to certain DNN configuration information, and the second DNN may include four convolutional layers according to other DNN configuration information.
- the DNN configuration information may include only parameters of a filter kernel used in the second DNN.
- the structure of the second DNN is not changed, but only parameters of an internal filter kernel may be changed according to the DNN configuration information.
- the AI setting unit 1335 may obtain DNN setting information for AI upscaling of the second image from among a plurality of DNN setting information.
- Each of the plurality of DNN setting information used herein is information for obtaining a third image having a predetermined resolution and/or a predetermined quality, and is trained in association with the first DNN.
- any one DNN setting information among a plurality of DNN setting information is a third image having a resolution twice as large as the resolution of the second image, for example, a 1080p third image twice as large as the 540p second image. It may include information for acquiring or information for acquiring a third image with a resolution four times greater than the resolution of the second image, for example, a third image of 1080p four times larger than the second image of 270p. .
- each DNN can be trained in consideration of the resolution of an image used in a video conference. For example, when the size of the original video used in video conferencing is 1080p, the DNN for AI downscaling converts a first video of 540p having a size of 1/2 of the original video or a size of 1/4 of the original video from the original video. outputs a first 270p image having a 270p image, and the DNN for AI upscaling may be trained to output a 540p second image or a 1080p third image from a 270p second image.
- Each of the plurality of DNN setting information is created in association with the DNN setting information of the first DNN of another electronic device or server, and the AI setting unit 1334 configures the DNN setting information according to the magnification ratio corresponding to the reduction ratio of the DNN setting information of the first DNN. Acquire one DNN setting information among a plurality of DNN setting information. To this end, the AI setting unit 1334 needs to check the information of the first DNN. In order for the AI setting unit 1334 to check information of the first DNN, the electronic device 1300 according to an embodiment receives AI data including information of the first DNN from a server managing a video conference.
- the AI setting unit 1334 uses the information received from the server to identify information targeted by the DNN setting information of the first DNN used to acquire the first image, and to obtain a second image trained in conjunction therewith. DNN configuration information of the DNN can be acquired.
- DNN setting information for AI upscaling of the second image is obtained from among the plurality of DNN setting information
- the DNN setting information is transmitted to the AI upscaler 1333, and based on the second DNN operating according to the DNN setting information Input data can be processed.
- the AI upscaler 1333 obtains any one DNN setting information
- the first convolution layer 310 and the second convolution layer 330 of the second DNN 300 shown in FIG. ) and the third convolution layer 350 the number of filter kernels included in each layer and parameters of the filter kernels are set to values included in the obtained DNN configuration information.
- the AI upscaler 1333 sets the parameters of the 3 X 3 filter kernel used in any one convolution layer of the second DNN shown in FIG. 4 to ⁇ 1, 1, 1, 1, 1, 1 , 1, 1, 1 ⁇ , and when there is a change in the DNN configuration information, the parameters of the filter kernel included in the changed DNN configuration information are ⁇ can be replaced.
- the AI setting unit 1334 may obtain DNN setting information for upscaling the second image from among a plurality of DNN setting information based on information included in the AI data. explain in detail.
- the AI setting unit 1334 may obtain DNN setting information for upscaling the second image from among a plurality of DNN setting information based on difference information included in AI data. For example, when it is determined that the resolution of the original image (eg, 1080p) is twice as large as the resolution (eg, 540p) of the first image based on the difference information, the AI setting unit 1835 It is possible to obtain DNN setting information capable of doubling the resolution of 2 images.
- the AI setting unit 1334 may obtain DNN setting information for AI upscaling a second image among a plurality of DNN setting information based on information related to the first image included in the AI data.
- the AI setting unit 1334 may pre-determine a mapping relationship between the image-related information and the DNN setting information, and obtain DNN setting information mapped to the first image-related information.
- a plurality of DNN setting information for AI downscaling and a plurality of DNN setting information for AI upscaling may be obtained through the training process described with reference to FIGS. 11 and 12 described above.
- one DNN setting information for AI downscaling and one DNN setting information for AI upscaling may be obtained through the training process described with reference to FIGS. 11 and 12 described above. That is, a pair of DNN setting information may be acquired for AI downscaling and AI upscaling, and in this case, one DNN setting information for AI upscaling may be stored in the AI setting unit.
- the AI setting unit 1334 provides one DNN setting information for AI upscaling a second image corresponding to a downscaled first image of 540p to a third image of 1080p based on information included in AI data. can be obtained.
- the information included in the AI data is information indicating that the image data transmitted together is for an image obtained through AI downscaling
- the AI setting unit 1334 converts one DNN setting information stored in the AI setting unit to AI. It is provided to the upscale unit 1333.
- the AI upscaler 1333 obtains a third image by performing AI upscale based on the DNN setting information and importance provided from the AI setter 1334, and provides the third image to the display unit 1350.
- the AI upscaler 1333 converts the DNN setting information provided from the AI setting unit 1334 into a 1080p third image corresponding to a 540p first image downscaled from a 1080p original image.
- the reason for performing AI upscaling twice is to match the resolution of the third image to the original image.
- the AI upscaler 1333 AI upscales the second image corresponding to the downscaled first image of 540p to the third image of 1080p based on the DNN setting information provided from the AI setting unit 1334. If the first image is 270p and the original image is 1080p and the importance is low, AI upscaling is not performed and the second image of 270p is provided to the display unit 1350.
- the AI upscaler 1333 AI upscales the second image corresponding to the downscaled first image of 540p to the third image of 1080p based on the DNN setting information provided from the AI setting unit 1334. If the first image is 270p and the original image is 1080p and the importance is low, AI upscaling is performed once using the predetermined DNN setting information to obtain a third image of 540p. acquired, and provides the third image to the display unit 1350.
- the display unit 1350 displays a second image corresponding to the downscaled image of another electronic device participating in the video conference or a third image corresponding to the original image.
- FIG. 14 is a block diagram illustrating a configuration of an electronic device that AI downscales a video conference video of an electronic device participating in a video conference during a video conference according to an embodiment.
- the electronic device 1300 of FIG. 13 receives a video conference video from another electronic device, determines whether to perform AI upscaling according to the importance of the other electronic device, and performs AI upscaling, whereas the electronic device 1400 of FIG. 14 performs AI upscaling.
- the video conference video of the electronic device 1400 is downscaled by AI and transmitted to another electronic device.
- the electronic device 1300 of FIG. 13 and the electronic device 1400 of FIG. 14 represent components that perform different roles in the electronic device participating in the video conference, respectively, and the electronic device 1300 of FIG. 13 and the electronic device 1400 of FIG. 14
- the electronic device 1400 of may be the same electronic device.
- the electronic device 1300 of FIG. 13 and the electronic device 1400 of FIG. 14 may be different electronic devices participating in the video conference.
- the electronic device 1400 of FIG. 14 has the same overall configuration as the above-described AI encoding device 700, and is different in that it additionally transmits importance information to the server.
- the electronic device 1400 participating in the video conference performs AI downscaling regardless of importance information and transmits the AI downscaled video to the server, but the electronic device 1400 transmits the video conference video
- the receiving electronic device 1300 or the server determines whether to upscale the video conference by AI based on the importance.
- an electronic device 1400 may include an AI encoder 1410 and a transmitter 1430.
- the AI encoder 1410 may include an AI downscaling unit 1411, a first encoding unit 1412, a data processing unit 1413, and an AI setting unit 1414.
- FIG. 14 illustrates the AI encoder 1410 and the transmitter 1430 as separate devices
- the AI encoder 1410 and the transmitter 1430 may be implemented by one processor.
- it may be implemented as a dedicated processor, or may be implemented through a combination of S/W and a general-purpose processor such as an AP, CPU, or GPU.
- a dedicated processor may include a memory for implementing an embodiment of the present disclosure or a memory processing unit for using an external memory.
- the AI encoder 1410 and the transmitter 1430 may be composed of a plurality of processors. In this case, it may be implemented with a combination of dedicated processors, or it may be implemented with a combination of S/W and a plurality of general-purpose processors such as APs, CPUs, or GPUs.
- the first encoding unit 1412 is composed of a first processor
- the AI downscaling unit 1411, the data processing unit 1413 and the AI setting unit 1414 are configured by a second processor different from the first processor.
- the transmitter 1430 may be implemented as a third processor different from the first processor and the second processor.
- the AI encoder 1410 performs AI downscaling of the original video of the electronic device 1400 participating in the video conference and first encoding of the AI downscaled first video, and transmits AI-encoded data and importance information to the transmission unit ( 1430).
- the transmission unit 1430 transmits AI-encoded data and importance information to a server managing a video conference.
- the image data includes data obtained as a result of the first encoding of the first image.
- the image data may include data obtained based on pixel values in the first image, eg, residual data that is a difference between the first image and prediction data of the first image.
- the image data includes information used in the first encoding process of the first image.
- the image data may include prediction mode information, motion information, and quantization parameter-related information used for first encoding of the first image.
- the AI upscale unit 1333 of the electronic device 1300 or the AI upscale unit 1533 of the server 1500 converts the second image to an AI upscale target corresponding to the downscale target of the first DNN.
- the AI data may include difference information between the original image and the first image.
- the AI data may include information related to the first image.
- the first image related information may include information on at least one of a resolution of the first image, a bit rate of image data obtained as a result of the first encoding of the first image, and a codec type used in the first encoding of the first image.
- the AI data may include an identifier of mutually promised DNN setting information so that the second image can be AI upscaled to an upscale target corresponding to a downscale target of the first DNN.
- the AI data may include DNN setting information that can be set in the second DNN.
- the AI downscaling unit 1411 may obtain a first AI downscaled image from an original image through the first DNN.
- the AI downscaling unit 1411 may AI downscale the original video using DNN setting information provided from the AI setting unit 1414.
- the AI setting unit 1414 may determine a downscale target of the original video based on a predetermined criterion.
- the AI setting unit 1414 may store a plurality of DNN setting information that can be set in the first DNN.
- the AI setting unit 1414 obtains DNN setting information corresponding to a downscale target among a plurality of DNN setting information, and provides the obtained DNN setting information to the AI downscaling unit 1411.
- Each of the plurality of DNN setting information may be trained to acquire a first image having a predetermined resolution and/or a predetermined quality.
- any one DNN setting information among a plurality of DNN setting information is a first image having a resolution that is 1/2 times smaller than the resolution of the original image, for example, 540p, which is 1/2 times smaller than the original 1080p image. It may include information for acquiring the first image, and the other piece of DNN setting information is the first image having a resolution that is 1/4 times smaller than the resolution of the original image, for example, 1/4 of the original 1080p image. It may include information for acquiring the first image of 270p smaller than the size.
- the AI setting unit 1414 may provide the AI downscaling unit 1411 with DNN setting information obtained by combining selected parts of the lookup table values according to the downscale target.
- the AI configuration unit 1414 may determine the structure of the DNN corresponding to the downscale target and obtain DNN configuration information corresponding to the determined structure of the DNN, for example, filter kernel parameters.
- the plurality of DNN setting information for AI downscaling of the original video may have an optimized value by performing joint training of the first DNN and the second DNN.
- each DNN configuration information includes at least one of the number of convolution layers included in the first DNN, the number of filter kernels for each convolution layer, and parameters of each filter kernel.
- the AI downscaling unit 1411 may obtain a first image having a predetermined resolution and/or a predetermined quality through the first DNN by setting the first DNN with the DNN setting information determined for AI downscaling of the original video.
- DNN setting information for AI downscaling of an original video is obtained from among a plurality of DNN setting information
- each layer in the first DNN may process input data based on information included in the DNN setting information.
- the AI setting unit 1414 may store one piece of DNN setting information that can be set in the first DNN in order to obtain the first image. For example, the AI setting unit 1414 obtains one DNN setting information for AI downscaling a 1080p original video to a first 540p video, and the AI downscaling unit 1411 uses the acquired DNN setting information. provided by
- the downscale target may indicate, for example, how much the first image whose resolution is reduced from the original image should be acquired.
- the AI setting unit 1414 obtains one or more pieces of input information.
- the input information may include a target resolution of the first video, a target bitrate of the video data, a bitrate type of the video data (eg, variable bitrate type, constant bitrate type, average bitrate type, etc.), AI downscale At least one of the applied color format (luminance component, chrominance component, red component, green component, blue component, etc.), codec type for first encoding of the first image, compression history information, original image resolution, and original image type. may contain one.
- One or more pieces of input information may include information previously stored in the electronic device 1400 or input from a user.
- the AI setting unit 1414 controls the operation of the AI downscaling unit 1411 based on input information.
- the AI setting unit 1414 may determine a downscale target according to input information and provide DNN setting information corresponding to the determined downscale target to the AI downscale unit 1411 .
- the AI setting unit 1414 transfers at least a portion of the input information to the first encoding unit 1412 so that the first encoding unit 1412 has a bit rate of a specific value, a bit rate of a specific type, and a specific codec.
- the first image may be subjected to first encoding.
- the AI setting unit 1414 may include compression rate (eg, resolution difference between the original video and the first video, target bitrate), compression quality (eg, bitrate type), compression history information, and A downscale target may be determined based on at least one type of original video.
- compression rate eg, resolution difference between the original video and the first video
- target bitrate e.g., resolution difference between the original video and the first video
- compression quality e.g., bitrate type
- compression history information e.g., compression history information
- a downscale target may be determined based on at least one type of original video.
- the AI setting unit 1414 may determine a downscale target based on a compression rate or compression quality that is previously set or input from a user.
- the AI setting unit 1414 may determine a downscale target using compression history information stored in the electronic device 1400 . For example, according to compression history information available to the electronic device 1400, a user's preferred encoding quality or compression rate may be determined, and a downscale target may be determined according to the encoding quality determined based on the compression history information. there is. For example, the resolution and quality of the first image may be determined according to the most frequently used encoding quality according to the compression history information.
- the AI setting unit 1414 determines the encoding quality that has been used more than a predetermined threshold value (eg, the average quality of encoding qualities that have been used more than a predetermined threshold value) according to the compression history information. Based on this, a downscale target may be determined.
- a predetermined threshold value eg, the average quality of encoding qualities that have been used more than a predetermined threshold value
- the AI setting unit 1414 may determine a downscale target based on the resolution and type (eg, file format) of the original video.
- the AI setting unit 718 independently obtains DNN setting information for each of a predetermined number of frames, and the independently acquired DNN setting information is used by the AI downscaling unit 1411 ) can also be provided.
- the AI setting unit 1414 may divide the frames constituting the original video into a predetermined number of groups, and independently obtain DNN setting information for each group.
- the same or different DNN setting information may be obtained for each group.
- the number of frames included in the groups may be the same or different for each group.
- the AI setting unit 1414 may independently determine DNN setting information for each frame constituting the original video. The same or different DNN configuration information may be obtained for each frame.
- the exemplary structure of the first DNN which is the basis of AI downscaling, is omitted as described above in FIG. 8 .
- the AI downscaling unit 1411 may include at least one ALU for convolution operation and activation layer operation.
- An ALU may be implemented as a processor.
- the ALU may include a multiplier for performing a multiplication operation between sample values of a feature map output from an original image or a previous layer and sample values of a filter kernel, and an adder for adding multiplication result values.
- the ALU is a multiplier that multiplies input sample values with weights used in a predetermined sigmoid function, Tanh function, or ReLU function, and compares the multiplication result with a predetermined value to obtain an input sample value.
- a comparator for determining whether to transfer to the next layer may be included.
- the AI downscaling unit 1411 AI downscales the original video based on one DNN setting information for AI downscaling the 1080p original video provided by the AI setting unit 1414 to the first 540p video. do.
- the AI downscaling unit 1411 additionally receives importance information from the AI setting unit 1414 and converts the 1080p original video into a 540p first video in order to further reduce data usage in the case of an original video having low importance.
- a first image of 270p may be obtained by performing AI downscaling twice using one DNN setting information for AI downscaling with .
- the AI downscaling unit 1411 AI downscales the 1080p original video provided by the AI setting unit 1414 to the first 540p video regardless of importance, based on one DNN setting information. AI downscales 2 times. In this case, data usage may be further reduced by AI downscaling the 1080p original video to the 270p first video.
- the AI setting unit 1414 transfers AI data to the data processing unit 1413 .
- the AI data includes information enabling the AI upscaling units 1333 and 1533 to AI upscale the second video to an upscale target corresponding to the downscale target of the first DNN.
- the first encoder 1412 receiving the first image from the AI downscaler 1411 first encodes the first image according to a frequency conversion-based image compression method to reduce the amount of information of the first image. .
- a predetermined codec eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1
- Image data is obtained according to the rules of a predetermined codec, that is, syntax.
- the image data may include residual data that is a difference between a first image and predicted data of the first image, prediction mode information used for first encoding the first image, motion information, and used for first encoding the first image. quantization parameter related information, etc. may be included.
- Image data obtained as a result of the first encoding of the first encoder 1412 is provided to the data processor 1413 .
- the AI setting unit 1414 generates importance information indicating whether the user of the electronic device 1400 is a presenter or a listener and transmits it to the data processing unit 1413.
- the AI setting unit 1414 may transmit importance information by including it in AI data.
- the AI setting unit 1414 generates the importance information as metadata separate from the AI-encoded data and transmits it to the data processing unit 1413.
- the data processing unit 1413 generates AI-encoded data including the image data received from the first encoding unit 1412 and the AI data received from the AI setting unit 1414 .
- the data processing unit 1413 may generate AI-encoded data including image data and AI data in a separated state.
- AI data may be included in a Vendor Specific InfoFrame (VSIF) within an HDMI stream.
- VSIF Vendor Specific InfoFrame
- the data processing unit 1413 may include AI data in image data obtained as a result of the first encoding by the first encoding unit 1412 and generate AI-encoded data including the corresponding image data. .
- the data processing unit 1413 may combine a bitstream corresponding to image data and a bitstream corresponding to AI data to generate image data in the form of a single bitstream.
- the data processor 1413 may express AI data as bits having a value of 0 or 1, that is, as a bitstream.
- the data processor 1413 may include a bitstream corresponding to AI data in supplemental enhancement information (SEI), which is an additional information area of the bitstream obtained as a result of the first encoding.
- SEI Supplemental Enhancement Information
- AI-encoded data is transmitted to the transmitter 1430.
- the transmitter 1430 transmits AI-encoded data obtained as a result of AI-encoding through a network.
- the AI encoded data is magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. medium) and the like may be stored in a data storage medium.
- the data processing unit 1413 transmits AI-encoded data based on AI data including importance information to the transmission unit 1430. According to another embodiment, the data processing unit 1413 transmits the importance information to the transmission unit 1430 together with the AI-encoded data. The transmission unit 1430 transmits AI-encoded data obtained as a result of AI-encoding and importance information through a network.
- the importance information is generated when the electronic device 1400 first participates in a video conference, transmitted to a server managing the video conference, and transmitted to other electronic devices capable of using the importance information through the server.
- the server may generate and store importance information of devices participating in the video conference, and then distribute the importance information to the devices when videos are transmitted and received between the devices. For example, when a video conference video for a first device is transmitted as a second video, the server may transmit importance information for the first device as a second video.
- the importance information may be changed and transmitted to a server that manages a video conference, and may be transmitted to other electronic devices capable of using the importance information through the server.
- an electronic device that does not support AI upscale and AI downscale may generate and transmit AI data or acquire and process AI data.
- an electronic device that does not support AI downscale and AI upscale and participates in a video conference transmits video data in which the original video of the video conference is encoded as it is to the server, and decodes and restores the video conference video data acquired from the server as it is. display an image.
- an electronic device that does not support AI downscale and AI upscale may transmit importance information about the importance of an electronic device that is not supported as meta data separately from image data or may include and transmit the image data.
- first electronic device refers to an electronic device supporting AI downscaling that transmits video of a video conference to a server that participates in a video conference and manages the video conference
- second electronic device refers to an electronic device that supports video conferencing.
- a “third electronic device” refers to a server that participates in a video conference and manages the video conference. It means an electronic device that does not support AI downscaling that transmits video.
- first electronic device “second electronic device”, and “third electronic device” are for describing an embodiment, but are not limited thereto.
- first image data for classifying image data, and thus Not limited.
- 15 is a block diagram illustrating a configuration of a server that AI upscales a video conference video according to whether an electronic device of a participant supports AI upscaling and importance information during a video conference according to an embodiment.
- a server 1500 may include a receiving unit 1510, an AI decoding unit 1530, and a first encoding unit 1550.
- the AI decoder 1530 may include a parser 1531, a first decoder 1532, an AI upscaler 1533, and an AI setter 1534.
- the receiving unit 1510, the AI decoding unit 1530, and the first encoding unit 1550 shows the receiving unit 1510, the AI decoding unit 1530, and the first encoding unit 1550 as separate devices, but the receiving unit 1510, the AI decoding unit 1530, and the first encoding unit 1550 It can be implemented through a single processor. In this case, it may be implemented as a dedicated processor, or may be implemented through a combination of S/W and a general-purpose processor such as an AP, CPU, or GPU. In addition, a dedicated processor may include a memory for implementing an embodiment of the present disclosure or a memory processing unit for using an external memory.
- the receiving unit 1510, the AI decoding unit 1530, and the first encoding unit 1550 may be configured with a plurality of processors. In this case, it may be implemented with a combination of dedicated processors, or it may be implemented with a combination of S/W and a plurality of general-purpose processors such as APs, CPUs, or GPUs.
- the receiving unit 1510 and the parsing unit 1531 are implemented with a first processor
- the first decoding unit 1532 is implemented with a second processor different from the first processor
- the AI upscaler 1533 is implemented with a third processor different from the first and second processors
- the first encoder 1550 is implemented with a fourth processor different from the first processor, the second processor, and the third processor.
- the receiving unit 1510 receives AI-encoded data obtained as a result of AI encoding and importance information of the first electronic device from the first electronic device participating in the video conference.
- AI-encoded data may be a video file having a file format such as mp4 or mov.
- the AI-encoded data obtained by the receiving unit 1510 from the first electronic device is data related to the first image obtained by AI downscaling the original image by the first electronic device.
- the receiving unit 1510 may receive AI-encoded data transmitted from the first electronic device and importance information of the first electronic device through a communication network.
- the receiving unit 1510 outputs AI encoded data and importance information to the parsing unit 1531 of the AI decoding unit 1530.
- the importance information may be included in AI data of AI-encoded data and transmitted. In one embodiment, the importance information may be received from the first electronic device as meta data separately from the AI-encoded data.
- the AI-encoded data and importance information of the first electronic device are bypassed through the server 1500 and transmitted to the second electronic device. That is, the server only serves to transmit files. Whether or not the second electronic device supports AI upscaling may be determined by whether AI data is included in data transmitted from the second electronic device to the server or whether information indicating that AI upscaling is supported is received.
- the parsing unit 1531 parses the AI-encoded data and transfers the video data generated as a result of the first encoding of the video conference video to the first decoding unit 1532, and the AI data and importance information to the AI setting unit 1534. convey Since AI data is acquired from the first electronic device, it can be known that the first electronic device supports AI downscaling.
- the parsing unit 1531 may parse image data and AI data separately included in AI-encoded data.
- the parsing unit 1531 may read a header in the AI-encoded data and distinguish between AI data and image data included in the AI-encoded data.
- the parsing unit 1331 may distinguish between AI data related to AI upscaling included in AI data and importance information for a video conference video.
- AI-encoded data including AI data and image data separated from each other is omitted as described above with reference to FIG. 9 .
- the parsing unit 1531 parses video data from AI-encoded data, extracts AI data from the video data, transfers the AI data to the AI setting unit 1534, and performs first decoding on the remaining video data. It can be passed to unit 1532. That is, AI data may be included in video data. For example, AI data may be included in supplemental enhancement information (SEI), which is an additional information area of a bitstream corresponding to video data.
- SEI Supplemental Enhancement information
- the parsing unit 1531 divides a bitstream corresponding to image data into a bitstream to be processed in the first decoding unit 1532 and a bitstream corresponding to AI data, and each of the divided bitstreams It can be output to the first decoding unit 1532 and the AI setting unit 1534.
- the parsing unit 1531 acquires video data included in AI-encoded data through a predetermined codec (eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1). It can also be confirmed that it is image data. In this case, corresponding information may be transmitted to the first decoder 1532 so that the video data can be processed with the identified codec.
- a predetermined codec eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1
- the first decoding unit 1532 restores a second video corresponding to the first video obtained by AI downscaling the video conference original video based on the video data received from the parsing unit 1531 .
- the second image obtained by the first decoder 1532 is provided to the AI upscaler 1533.
- the importance information from the AI setting unit 1534 is also provided to the AI upscaling unit 1533.
- the AI upscaler 1533 provides the second image as it is to the first encoder 1550 according to the importance information or provides the third image obtained by AI upscaling the second image to the first encoder 1550. do.
- the AI upscaler 1533 provides a third image obtained by AI upscaling the second image to the first encoder 1550, .
- the second image is provided to the first encoder 1550.
- the importance information may be flag information set to 1 if a user of another electronic device is a presenter and set to 0 if the user is a listener.
- the server 1500 when starting a video conference, the server 1500 obtains and stores importance information of a first electronic device from a first electronic device participating in the video conference, and the second electronic device uses the importance information. , The server 1500 transmits the importance information to the second electronic device, and when the importance is changed, the server 1500 obtains and stores the changed importance information from the first electronic device, and the second electronic device transmits the importance information. When used, the server 1500 may transmit importance information to the second electronic device. That is, the importance information may not always be transmitted, but the changed importance information may be transmitted when a video conference starts and when there is a change in importance.
- the server 1500 acquires and stores importance information of the first electronic device transmitted when the video conference starts, performs AI upscaling based on the importance information, .
- the server 1500 transmits the received importance information of the first electronic device or the changed importance information of the first electronic device to the second electronic device when the video conference starts. .
- first decoding-related information such as prediction mode information, motion information, and quantization parameter information may be provided from the first decoding unit 1532 to the AI setting unit 1534.
- the first decoding related information may be used to obtain DNN setting information.
- the AI data provided to the AI setting unit 1534 includes information enabling AI up-scaling of the second image.
- the upscale target of the second image must correspond to the downscale target of the first DNN. Therefore, the AI data must include information capable of identifying the downscale target of the first DNN.
- the information included in the AI data there are information on the difference between the resolution of the original video and the resolution of the first video obtained by AI downscaling the original video during video conference, and information related to the first video.
- the difference information may be expressed as information about a degree of resolution conversion of the first image compared to the original image (eg, resolution conversion rate information).
- the difference information may be expressed only with the resolution information of the original image.
- the resolution information may be expressed as a horizontal/vertical screen size, or as a ratio (16:9, 4:3, etc.) and a size on one axis.
- it may be expressed in the form of an index or flag.
- the first image-related information may include information on at least one of a resolution of the first image, a bit rate of image data obtained as a result of the first encoding of the first image, and a codec type used in the first encoding of the first image.
- the AI setting unit 1534 may determine an upscale target of the second image based on at least one of difference information included in the AI data and information related to the first image.
- the upscaling target may indicate, for example, to what resolution the second image should be upscaled.
- the AI upscaler 1533 AI upscales the second image through a second DNN to obtain a third image corresponding to the upscale target.
- the AI setting unit 1534 determines an upscaling target and the AI upscaling unit 1533 AI upscaling the second video according to the upscaling target will be described.
- the AI setting unit 1534 may store a plurality of DNN setting information that can be set in the second DNN.
- the DNN configuration information may include information on at least one of the number of convolution layers included in the second DNN, the number of filter kernels for each convolution layer, and parameters of each filter kernel.
- a plurality of DNN configuration information may correspond to various upscale targets, and the second DNN may operate based on the DNN configuration information corresponding to a specific upscale target.
- the second DNN may have a different structure according to DNN configuration information.
- the second DNN may include three convolutional layers according to certain DNN configuration information, and the second DNN may include four convolutional layers according to other DNN configuration information.
- the DNN configuration information may include only parameters of a filter kernel used in the second DNN.
- the structure of the second DNN is not changed, but only parameters of an internal filter kernel may be changed according to the DNN configuration information.
- the AI setting unit 1534 may obtain DNN setting information for AI upscaling of the second image from among a plurality of DNN setting information.
- Each of the plurality of DNN setting information used herein is information for acquiring a reconstructed image having a predetermined resolution and/or a predetermined quality, and is trained in association with the first DNN.
- one piece of DNN setting information among a plurality of DNN setting information is a reconstructed image having a resolution twice as large as the resolution of the second image, for example, a third image of 1080p that is twice as large as the second image of 540p. It may include information for acquiring or information for acquiring a reconstructed image having a resolution four times larger than the resolution of the second image, for example, a third image of 1080p four times larger than the second image of 270p.
- each DNN can be trained in consideration of the resolution of an image used in a video conference. For example, when the size of the original video used in video conferencing is 1080p, the DNN for AI downscaling mainly produces a first video of 540p having a size of 1/2 of the original video or a size of 1/4 of the original video. , and the DNN for AI upscaling may be trained to output a 540p second image or a 1080p third image from a 270p second image.
- Each of the plurality of DNN setting information is created in association with the DNN setting information of the first DNN of the other electronic device, and the AI setting unit 1534 sets the plurality of DNN settings according to the scaling ratio corresponding to the scaling ratio of the DNN setting information of the first DNN. Acquire one DNN setting information among DNN setting information. To this end, the AI setting unit 1534 needs to check the information of the first DNN.
- the server 1500 receives AI data including information of the first DNN from an electronic device participating in a video conference.
- the AI setting unit 1534 uses information received from electronic devices participating in the video conference to check information targeted by the DNN setting information of the first DNN used to acquire the first video, and It is possible to obtain DNN setting information of the second DNN trained in conjunction with .
- DNN setting information for AI upscaling of the second image is obtained from among the plurality of DNN setting information
- the DNN setting information is transmitted to the AI upscaler 1533, and based on the second DNN operating according to the DNN setting information Input data can be processed.
- the AI upscaler 1533 obtains any one DNN setting information
- the first convolution layer 310 and the second convolution layer 330 of the second DNN 300 shown in FIG. ) and the third convolution layer 350 the number of filter kernels included in each layer and parameters of the filter kernels are set to values included in the obtained DNN configuration information.
- the AI upscaler 1533 sets the parameters of the 3 X 3 filter kernel used in any one convolution layer of the second DNN shown in FIG. 4 to ⁇ 1, 1, 1, 1, 1, 1 , 1, 1, 1 ⁇ , and when there is a change in the DNN configuration information, the parameters of the filter kernel included in the changed DNN configuration information are ⁇ can be replaced.
- the AI setting unit 1534 may obtain DNN setting information for upscaling the second image from among a plurality of DNN setting information based on information included in the AI data. explain in detail.
- the AI setting unit 1534 may obtain DNN setting information for upscaling the second image from among a plurality of DNN setting information based on difference information included in AI data. For example, when it is determined that the resolution of the original image (eg, 1080p) is twice as large as the resolution (eg, 540p) of the first image based on the difference information, the AI setting unit 1534 It is possible to obtain DNN setting information capable of doubling the resolution of 2 images.
- the AI setting unit 1534 may obtain DNN setting information for AI upscaling a second image among a plurality of DNN setting information based on information related to the first image included in the AI data.
- the AI setting unit 1534 may pre-determine a mapping relationship between the image-related information and the DNN setting information, and obtain DNN setting information mapped to the first image-related information.
- a plurality of DNN setting information for AI downscaling and a plurality of DNN setting information for AI upscaling may be obtained through the training process described with reference to FIGS. 11 and 12 described above.
- one DNN setting information for AI downscaling and one DNN setting information for AI upscaling may be obtained through the training process described with reference to FIGS. 11 and 12 described above. That is, a pair of DNN setting information may be acquired for AI downscaling and AI upscaling, and in this case, one DNN setting information for AI upscaling may be stored in the AI setting unit.
- the AI setting unit 1534 is one DNN setting information for AI up-scaling the second image corresponding to the downscaled first image of 540p to the third image of 1080p based on the information included in the AI data. can be obtained.
- the information included in the AI data is information indicating that the image data transmitted together is for an image obtained through AI downscaling, and the AI setting unit 1534 converts one DNN setting information stored in the AI setting unit to AI upscaling. It is provided to the scale part 1533.
- the AI upscaling unit 1533 acquires a third image by performing AI upscaling based on the DNN setting information provided from the AI setting unit 1534, and provides the third image to the first encoding unit 1550.
- the AI upscaler 1533 converts the DNN setting information provided from the AI setting unit 1534 into a 1080p third image corresponding to a 540p first image downscaled from a 1080p original image.
- One predetermined DNN setting information for AI upscaling ie, DNN setting information with a scaling factor of 2
- one predetermined DNN setting information AI upscaling is performed twice using , to obtain a third image of 1080p, and the third image is provided to the first encoder 1550.
- the reason for performing AI upscaling twice is to match the resolution of the third image to the original image.
- the AI upscaler 1533 AI upscales the second image corresponding to the downscaled first image of 540p to the third image of 1080p based on the DNN setting information provided from the AI setting unit 1534. If the first image is 270p and the original image is 1080p and the importance is low, AI upscaling is not performed and the second image of 270p is provided to the first encoder 1550. .
- the AI upscaler 1533 AI upscales the second image corresponding to the downscaled first image of 540p to the third image of 1080p based on the DNN setting information provided from the AI setting unit 1534. If the first image is 270p and the original image is 1080p and the importance is low, AI upscaling is performed once using the predetermined DNN setting information to obtain a third image of 540p. is obtained, and the third image is provided to the first encoder 1550.
- the AI upscaling unit 1533 provides the second image to the first encoding unit 1550 without performing AI upscaling based on the importance information provided from the AI setting unit 1534.
- the first encoder 1550 first encodes the second image corresponding to the downscaled image of the first electronic device participating in the video conference provided by the AI upscaler 1533 to obtain second image data; Third image data obtained by first encoding a third image corresponding to the original image provided from the AI upscaler 1533 is acquired.
- the first encoder 1550 transmits the second image data or the third image data to the second electronic device participating in the video conference.
- 16 is a block diagram illustrating a configuration of a server for managing a video conference that AI downscales a video conference video according to an exemplary embodiment
- a server 1600 may include a parsing unit 1610, a first decoding unit 1620, an AI encoding unit 1630, and a transmission unit 1640.
- the AI encoder 1630 may include an AI downscaling unit 1631, a first encoding unit 1632, a data processing unit 1633, and an AI setting unit 1634.
- the parsing unit 1610, the first decoding unit 1620, the AI encoding unit 1630, and the transmission unit 1640 may be implemented through one processor.
- it may be implemented as a dedicated processor, or may be implemented through a combination of S/W and a general-purpose processor such as an AP, CPU, or GPU.
- a dedicated processor may include a memory for implementing an embodiment of the present disclosure or a memory processing unit for using an external memory.
- the parsing unit 1610, the first decoding unit 1620, the AI encoding unit 1630, and the transmission unit 1640 may include a plurality of processors. In this case, it may be implemented with a combination of dedicated processors, or it may be implemented with a combination of S/W and a plurality of general-purpose processors such as APs, CPUs, or GPUs.
- the parsing unit 1610 is implemented by a first processor
- the first decoding unit 1620 is implemented by a second processor different from the first processor
- the AI encoding unit 1630 is implemented by the first processor and the second processor. It may be implemented with a third processor different from the second processor
- the transmitter 1640 may be implemented with a first processor, a second processor, and a fourth processor different from the third processor.
- the parsing unit 1610 obtains fourth image data and importance information from a third electronic device. Since the server 1600 does not need to perform AI downscaling when the third electronic device supports AI downscaling, the fourth image data acquired by the server 1600 from the third electronic device is an image of the original image. It is data.
- the server 1600 transmits the AI-encoded data of the first image obtained by AI downscaling the original video of the third electronic device and the importance information to the second electronic device regardless of the importance information. send.
- the server 1600 determines whether to perform AI downscaling of the original video of the third electronic device based on the importance information, and if the importance is high, the server 1600 determines whether to perform AI downscaling of the original video of the third electronic device.
- the fourth image data of or image data generated by newly first encoding the original image is transmitted to the second electronic device, and if the importance is low, the fifth image data of the image obtained by AI downscaling the original image of the third electronic device is transmitted to the second electronic device.
- the server 1500 transmits the acquired AI-encoded data acquired from the first electronic device to the second electronic device as it is, and the second electronic device performs AI upscaling. If not supported, the third image data for the third image obtained by performing AI upscaling according to the importance of the first electronic device is transmitted to the second electronic device, or AI downscaled second image data is not performed without performing AI upscaling. The second image data for the 2 images is transmitted to the second electronic device.
- the server 1600 transmits AI-encoded data obtained by AI downscaling the original video of the third electronic device to the second electronic device, or the second electronic device If AI upscaling is not supported, the fifth image data for the first image obtained by performing AI downscaling according to the importance of the third electronic device is transmitted to the second electronic device, or AI downscaling is not performed, The fourth image data corresponding to the original image is transmitted to the second electronic device.
- the server 1500 of FIG. 15 and the server 1600 of FIG. 16 are components that perform different roles in a server managing a video conference in which the first electronic device, the second electronic device, and the third electronic device participate. , and are the same server.
- the parsing unit 1610 needs to decode image data into images in order to perform AI downscaling
- the fourth image data for the original image is provided to the first decoder 1620 and importance information is provided to the AI setting unit. (1634).
- the importance information may be included in fourth image data of the original image and transmitted. Also, the importance information may be transmitted as metadata separately from the fourth image data for the original image.
- the first decoder 1620 decodes video data (fourth video data) of the original video of the third electronic device, restores the original video of the video conference, and provides it to the AI downscaling unit 1631.
- the AI encoder 1630 performs AI down-scaling of the original video of the third electronic device participating in the video conference and AI down-scaled first image. First encoding is performed, and the AI-encoded data and importance information are delivered to the transmission unit 1640.
- the AI encoder 1630 may use the third electronic device participating in the video conference based on the importance information received from the parsing unit 1610 when the second electronic device participating in the video conference does not support AI upscaling. Determines whether to perform AI downscaling of the original video of , performs first encoding of the original video or the AI downscaled first video according to the importance information, and transmits the AI-encoded data and the importance information to the transmission unit 1640 .
- the AI-encoded data is fourth image data of the original image or fifth image data of the AI downscaled first image, and does not include AI data related to AI downscaling. This is because AI data is not required because the second electronic device does not support AI upscaling.
- the transmission unit 1640 transmits the AI-encoded data and the importance information to the second electronic device participating in the video conference.
- the fourth image data includes data obtained as a result of first encoding the original image
- the fifth image data includes data obtained as a result of first encoding the first image.
- the fourth image data or the fifth image data is data obtained based on pixel values in the original image or the first image, for example, residual data that is a difference between the original image and prediction data of the original image or the first image and the first image data. It may include residual data that is a difference between the predicted data of the first image.
- the fourth image data or the fifth image data includes information used in the first encoding process of the original image or the first image, respectively.
- the fourth image data or the fifth image data may include prediction mode information, motion information, and a quantization parameter used for first encoding the original image or the first image, respectively. It may include related information, etc.
- the AI data includes information enabling the AI upscaling unit 1333 of the electronic device 1300 to AI upscale the second image to an upscaling target corresponding to the downscaling target of the first DNN.
- the AI data may include difference information between the original image and the first image.
- the AI data may include information related to the first image.
- the first image related information may include information on at least one of a resolution of the first image, a bit rate of image data obtained as a result of the first encoding of the first image, and a codec type used in the first encoding of the first image.
- the AI data may include an identifier of mutually promised DNN setting information so that the second image can be AI upscaled to an upscale target corresponding to a downscale target of the first DNN.
- the AI data may include DNN setting information that can be set in the second DNN.
- the AI downscaling unit 1631 may obtain an AI downscaled first image from the original image through the first DNN regardless of importance information.
- the AI downscaling unit 1631 may perform AI downscaling of the original video using DNN setting information provided from the AI setting unit 1634.
- the AI downscaling unit 1631 determines whether to perform AI downscaling based on the importance information provided from the AI setting unit 1634 when the second electronic device does not support AI upscaling.
- the AI downscaling unit 1631 provides the original video to the first encoding unit 1632 without performing AI downscaling.
- the AI downscaler 1631 may obtain an AI downscaled first image from the original image through the first DNN.
- the AI downscaling unit 1631 may perform AI downscaling of the original video using DNN setting information provided from the AI setting unit 1634.
- Whether or not the second electronic device supports AI upscaling may be determined based on whether the server 1600 acquires AI data from the second electronic device or AI upscaling support information obtained from the second electronic device.
- the AI setting unit 1634 may determine a downscale target of the original video based on a predetermined criterion.
- the AI setting unit 1634 may provide the importance information received from the parsing unit 1610 to the AI downscaling unit 1631 to determine whether to perform AI downscaling.
- the AI setting unit 1634 may store a plurality of DNN setting information that can be set in the first DNN.
- the AI setting unit 1634 obtains DNN setting information corresponding to a downscale target from among a plurality of DNN setting information, and provides the obtained DNN setting information to the AI downscaling unit 1631.
- Each of the plurality of DNN setting information may be trained to acquire a first image having a predetermined resolution and/or a predetermined quality.
- any one DNN setting information among a plurality of DNN setting information is a first image having a resolution 1/2 smaller than the resolution of the original image), for example, 540p 1/2 smaller than the original image of 1080p. It may include information for acquiring the first image of , or information for acquiring the first image of 270p, which is 1/4 times smaller than the original image of 1080p.
- the AI setting unit 1634 may provide the AI downscaling unit 1631 with DNN setting information obtained by combining selected parts of the lookup table values according to the downscaling target.
- the AI configuration unit 1634 may determine the structure of the DNN corresponding to the downscale target and obtain DNN configuration information corresponding to the determined structure of the DNN, for example, filter kernel parameters.
- the plurality of DNN setting information for AI downscaling of the original video may have an optimized value by performing joint training of the first DNN and the second DNN.
- each DNN configuration information includes at least one of the number of convolution layers included in the first DNN, the number of filter kernels for each convolution layer, and parameters of each filter kernel.
- the AI downscaling unit 1631 may set the first DNN with DNN setting information determined for AI downscaling of the original video, and obtain a first video having a predetermined resolution and/or a predetermined quality through the first DNN.
- DNN setting information for AI downscaling of an original video is obtained from among a plurality of DNN setting information
- each layer in the first DNN may process input data based on information included in the DNN setting information.
- the AI setting unit 1634 may store one DNN setting information that can be set in the first DNN to acquire the first image. For example, the AI setting unit 1634 obtains one DNN setting information for AI downscaling a 1080p original video to a first 540p video, and the AI downscaling unit 1631 uses the acquired DNN setting information.
- the downscale target may indicate, for example, how much the first image whose resolution is reduced from the original image should be acquired.
- the AI setting unit 1634 obtains one or more pieces of input information.
- the target resolution of the first image, the target bitrate of the image data, the bitrate type of the image data eg, variable bitrate type, constant bitrate type or average bitrate type, etc.
- AI downscaling is applied At least one of a color format (luminance component, chrominance component, red component, green component, blue component, etc.), codec type for first encoding of the first image, compression history information, original image resolution, and original image type can do.
- One or more pieces of input information may include information previously stored in the server 1600 or input from a user.
- the AI setting unit 1634 controls the operation of the AI downscaling unit 1631 based on the input information.
- the AI setting unit 1634 may determine a downscale target according to input information and provide DNN setting information corresponding to the determined downscale target to the AI downscaling unit 1631 .
- the AI setting unit 1634 transfers at least a portion of the input information to the first encoding unit 1632 so that the first encoding unit 1632 has a bit rate of a specific value, a bit rate of a specific type, and a specific codec.
- the first image may be subjected to first encoding.
- the AI setting unit 1634 may include compression rate (eg, resolution difference between the original video and the first video, target bitrate), compression quality (eg, bitrate type), compression history information, and A downscale target may be determined based on at least one type of original video.
- the AI setting unit 1634 may determine a downscale target based on a compression rate or compression quality that is preset or input from a user.
- the AI setting unit 1634 may determine a downscale target using compression history information stored in the server 1600 .
- the encoding quality or compression rate preferred by the participants in the video conference may be determined, and the downscale target may be set according to the encoding quality determined based on the compression history information.
- the resolution and quality of the first image may be determined according to the most frequently used encoding quality according to the compression history information.
- the AI setting unit 1634 determines the encoding quality that has been used more than a predetermined threshold value (eg, the average quality of encoding qualities that have been used more than a predetermined threshold value) according to the compression history information. Based on this, a downscale target may be determined.
- a predetermined threshold value eg, the average quality of encoding qualities that have been used more than a predetermined threshold value
- the AI setting unit 1634 may determine a downscale target based on the resolution and type (eg, file format) of the original video.
- the AI setting unit 1634 independently acquires DNN setting information for each of a predetermined number of frames, and the independently acquired DNN setting information is used by the AI downscaling unit 1631 ) can also be provided.
- the AI setting unit 1634 may divide the frames constituting the original video into a predetermined number of groups, and independently obtain DNN setting information for each group.
- the same or different DNN setting information may be obtained for each group.
- the number of frames included in the groups may be the same or different for each group.
- the AI setting unit 1634 may independently determine DNN setting information for each frame constituting the original video. The same or different DNN configuration information may be obtained for each frame.
- the exemplary structure of the first DNN which is the basis of AI downscaling, is omitted as described above in FIG. 8 .
- the AI downscaling unit 1631 may include at least one ALU for convolution operation and activation layer operation.
- An ALU may be implemented as a processor.
- the ALU may include a multiplier for performing a multiplication operation between sample values of a feature map output from an original image or a previous layer and sample values of a filter kernel, and an adder for adding multiplication result values.
- the ALU is a multiplier that multiplies input sample values with weights used in a predetermined sigmoid function, Tanh function, or ReLU function, and compares the multiplication result with a predetermined value to obtain an input sample value.
- a comparator for determining whether to transfer to the next layer may be included.
- the AI downscaling unit 1631 AI downscales the original video based on one DNN setting information for AI downscaling the 1080p original video provided by the AI setting unit 1634 to the first 540p video. do.
- the AI downscaling unit 1631 receives additional importance information from the AI setting unit 1634 and converts the 1080p original video into a 540p first video in order to further reduce data usage in the case of an original video having low importance.
- a first image of 270p may be obtained by performing AI downscaling twice using one DNN setting information for AI downscaling with .
- the AI downscaling unit 1631 AI downscales the 1080p original video provided by the AI setting unit 1634 to the first 540p video regardless of the importance, based on one DNN setting information. AI downscales 2 times. In this case, data usage may be further reduced by AI downscaling the 1080p original video to the 270p first video.
- the AI setting unit 1634 transfers AI data to the data processing unit 1613 .
- the AI data includes information enabling the AI upscaling unit 1333 of the electronic device to AI upscale the second image to an upscale target corresponding to the downscale target of the first DNN.
- the first encoder 1632 receiving the first image from the AI downscaler 1631 first encodes the first image according to a frequency conversion-based image compression method to reduce the amount of information of the first image. .
- a predetermined codec eg, MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1
- Image data is obtained according to the rules of a predetermined codec, that is, syntax.
- the image data may include residual data that is a difference between a first image and predicted data of the first image, prediction mode information used for first encoding the first image, motion information, and used for first encoding the first image. quantization parameter related information, etc. may be included.
- Image data obtained as a result of the first encoding of the first encoder 1632 is provided to the data processor 1633 .
- the AI setting unit 1634 obtains importance information indicating whether the user of the third electronic device is a presenter or a listener and transmits it to the data processing unit 1633 .
- the AI setting unit 1634 may transmit importance information by including it in AI data.
- the AI setting unit 1634 acquires the importance information as metadata separate from the AI-encoded data and transmits it to the data processing unit 1633.
- the first encoder 1632 obtains a first image by AI downscaling the original image using a DNN for downscaling, regardless of the importance of the third electronic device. and transmits fifth image data obtained by first encoding the first image to the data processing unit 1633.
- the first encoder 1632 converts fourth image data or image data generated by newly encoding the original image.
- data processing unit 1633 and if the second electronic device does not support AI upscaling and the importance of the third electronic device indicates a listener, the original video is AI downscaled using the downscale DNN to obtain the first video is obtained, and the fifth image data obtained by first encoding the first image is transmitted to the data processing unit 1633.
- the data processing unit 1633 may perform AI-encoded data including the fifth image data received from the first encoding unit 1632 and the AI data received from the AI setting unit 1634. generate Also, if the second electronic device does not support AI upscaling, the data processor 1633 does not include AI data and includes one of the fourth image data or the fifth image data received from the first encoder 1632. Generates AI-encoded data that
- the data processing unit 1633 may generate AI-encoded data including image data and AI data in a separated state.
- AI data may be included in a Vendor Specific InfoFrame (VSIF) within an HDMI stream.
- VSIF Vendor Specific InfoFrame
- the data processing unit 1633 may include AI data in image data obtained as a result of the first encoding by the first encoding unit 1632 and generate AI-encoded data including the corresponding image data. .
- the data processing unit 1633 may combine a bitstream corresponding to image data and a bitstream corresponding to AI data to generate image data in the form of a bitstream.
- the data processor 1413 may express AI data as bits having a value of 0 or 1, that is, as a bitstream.
- the data processor 1633 may include a bitstream corresponding to AI data in supplemental enhancement information (SEI), which is an additional information area of the bitstream obtained as a result of the first encoding.
- SEI Supplemental Enhancement Information
- AI-encoded data is transmitted to the transmitter 1640.
- the transmission unit 1640 transmits AI-encoded data obtained as a result of AI-encoding through a network.
- the AI encoded data is magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. medium) and the like may be stored in a data storage medium.
- the data processing unit 1633 transmits AI-encoded data based on AI data including importance information to the transmission unit 1640.
- the data processing unit 1633 transmits the importance information to the transmission unit 1640 together with the AI-encoded data.
- the transmission unit 1640 transmits AI-encoded data obtained as a result of AI-encoding and importance information through a network.
- the importance information is generated when a third electronic device first participates in a video conference, transmitted to and stored in the server 1600, and the server 1600 determines that the stored importance information can be used by the second electronic device. It can be transmitted to the second electronic device.
- the changed importance information is transmitted to and stored in the server 1600, and the server 1600 allows the second electronic device to use the changed importance information. It can be transmitted to the second electronic device.
- the AI setting unit 1634 does not transmit importance information and AI data to the data processing unit 1633 if the second electronic device does not support AI upscaling.
- the second electronic device only decodes and displays the image data transmitted from the server, and there is no need to perform any operation based on the importance and AI data.
- the data processing unit 1633 transmits AI-encoded data based on image data not including AI data to the transmission unit 1640 .
- the transmission unit 1640 transmits the fourth image data of the original image or the fifth image data of the first image to the second electronic device through the network.
- 17 is a diagram for explaining a data transmission relationship between an electronic device and a server in a conventional video conference.
- an electronic device 1710 of participant (presenter) A an electronic device 1720 of participant (listener) B, and an electronic device 1730 of participant (listener) C
- AI downscaling and AI upscaling of data are not used, and the video conference video of each electronic device is transmitted to other electronic devices through the server at a fixed bit rate and resolution. is played
- the electronic device of presenter A transmits the video data of the 1080p resolution video of presenter A to the server at 10 Mbps
- the server transmits the video data of the 1080p resolution video of presenter A to the electronic devices of listeners B and listener C.
- the data is transmitted at 10 Mbps
- the electronic device of listener B transmits the video data of the 1080p resolution video of listener B to the server at 10 Mbps
- the server transmits the video data of the 1080p resolution video of listener B to the electronic devices of presenter A and listener C.
- the video data of the video of the listener C is transmitted at 10 Mbps, and the electronic device of the listener C transmits the video data of the video of the 1080p resolution of the listener C to the server at 10 Mbps. Transmits video data of 1080p resolution video at 10Mbps.
- a plurality of electronic devices transmit and receive video conference images at a fixed bit rate and resolution through communication with a server.
- 18 is a diagram for explaining a data transmission relationship between an electronic device supporting AI downscaling and AI upscaling in a videoconference and a server according to an embodiment.
- an electronic device 1810 of participant (presenter) A, an electronic device 1820 of participant (listener) B, and an electronic device 1830 of participant (listener) C are Since both support AI downscaling and AI upscaling, the video conference video of each participant is AI downscaled and transmitted, and the AI downscaled video is AI upscaled and displayed as needed. Accordingly, the usage of uploading and downloading data of electronic devices participating in the video conference is reduced and image quality is maintained.
- Electronic devices of the participants in the video conference AI downscale the video conference video and transmit it to the server. Since the electronic devices support both AI downscaling and AI upscaling, the server transmits the AI downscaled image of each electronic device to other electronic devices as it is.
- participant A Since participant A is a presenter, the quality of the video is more important than that of the other participants B and C, who are listeners, and since the video of speaker A may contain a lot of important information, the electronic devices of other listeners B and C are the electronic devices of presenter A.
- the AI downscaled video conference video of presenter A which was transmitted by bypass through the server from A, is AI upscaled, restored to the original picture quality, and displayed.
- the AI downscaled video conference video of listeners B and C is transmitted from the electronic devices of listeners B and C through the server as a bypass from the electronic devices of presenter A. is displayed as it is without AI upscaling.
- the electronic devices of listeners B and C respectively display AI downscaled video conference images of C and B as they are without AI upscaling.
- the presenter A's electronic device 1810 applies AI downscaling to the 1080p and 10Mbps video conference video, transmits video data and AI data for the 540p and 5 Mbps video conference video to the server 1850
- B's electronic device 1820 applies AI downscaling to 1080p and 10Mbps video conference video, transmits video data and AI data for 540p and 5 Mbps video conference video to the server 1850
- listener C's electronic device (1830) applies AI downscaling to 1080p and 10 Mbps video conference video and transmits video data and AI data for 540p and 5 Mbps video to the server 1850.
- the server 1850 transmits image data and AI data for AI downscaled images received from each electronic device to other electronic devices as they are. That is, the server 1850 transmits video data and AI data for a 540p and 5Mbps video of listener B and video data and AI data for a 540p and 5Mbps video of listener C to the electronic device 1810 of presenter A, , Video data and AI data for the 540p and 5Mbps video of the presenter A and video data and AI data for the 540p and 5Mbps video of the listener C are transmitted to the electronic device 1820 of listener B, and the electronic device of listener C In (1830), video data and AI data for 540p and 5Mbps video of presenter A and video data and AI data for 540p and 5Mbps video of listener B are transmitted.
- the electronic device 1810 of presenter A transmits the transmitted 540p and 5 Mbps video of listener B and 540p of listener C.
- the importance of the electronic device may be changed by an input of the electronic device.
- participant B executes an input that activates the raise hand function (eg, clicking the raise hand button) while presenter A is presenting
- participant B's importance is changed to indicate that participant B is the presenter
- the presenter The importance of A is changed to indicate that speaker A is a listener.
- the electronic device 1820 of participant B who became the presenter displays the 540p video conference video of participant A who was the presenter as it is without AI upscaling
- the electronic device 1810 of listener A and the electronic device of listener C ( 1830) AI upscales the 540p video conference video of participant B, who has been changed to the presenter, according to the change in importance, and displays the 1080p video conference video.
- the electronic device 1830 of listener B displays the 540p video conference video of participant C, who has been changed to the presenter, as a 1080p video conference video by AI upscaling according to the importance change.
- participant B executes an input granting the right to present to another participant B
- participant B is changed to a presenter
- participant A's electronic device 1810 and participant C's electronic device 1830 are presenter B
- the AI downscaled video of is upscaled by AI and displayed as a 1080p video conference video.
- a plurality of electronic devices set as presenters may be provided.
- all videos set as the presenter can be AI upscaled to original videos.
- the importance of another electronic device is determined through importance information obtained from another electronic device through a server.
- the importance information may be included in AI data obtained from other electronic devices and may be transmitted separately.
- the importance information is transmitted through the server and stored in each electronic device for use.
- the changed importance information is transmitted again through the server and stored in each electronic device for use. It can be.
- the importance information is set to 1 if the user of the electronic device is a presenter and set to 0 if the user is a listener, so that if the importance information indicates 1, AI upscaling is performed, and if the importance information indicates 0, AI upscaling is not performed.
- 19 is an electronic device that supports AI downscale and AI upscale in a video conference, an electronic device that does not support AI downscale and AI upscale, and a server that supports AI downscale and AI upscale according to an embodiment. It is a diagram for explaining the data transmission relationship between
- the electronic device 1920 of participant (listener) B and the electronic device 1930 of participant (listener) C do not support AI downscale and AI upscale, and the electronic device 1910 of participant (presenter) A is assumed to support AI downscale and AI upscale.
- the electronic device 1920 of participant (listener) B and the electronic device 1930 of participant (listener) C do not support AI downscaling, so the electronic device of listener B (1920) and the electronic device (1930) of the listener C transmit the original video conference video to the server (1950) as it is.
- the electronic device 1910 of participant (presenter) A supports AI downscaling
- the electronic device 1910 of presenter A transmits the AI downscaled video to the server 1950 .
- the server 1950 also supports AI downscaling, the server 1950 performs AI downscaling of the original images of the electronic device 1920 of listener B and the electronic device 1930 of listener C, which are transmitted.
- participant A Since participant A is a presenter, the video conference video of the presenter A's electronic device is of high importance, and participants B and C, who are listeners, do not support AI upscaling, so the server 1950 transmits the AI transmitted from the presenter A's electronic device 1910.
- the AI upscaling of the downscaled video and the reconstructed video are transmitted to the electronic device 1920 of listener B and the electronic device 1930 of listener C that do not support AI upscaling.
- the server 1950 AI downscales the original video obtained from the electronic device 1920 of listener B and transmits the AI downscaled video to the electronic device 1910 of presenter A and the electronic device 1930 of listener C.
- the server 1950 AI downscales the original video acquired from the electronic device 1930 of listener C and transmits the AI downscaled video to the electronic device 1910 of presenter A and the electronic device 1920 of listener B. do.
- the electronic device 1910 of presenter A transmits image data and AI data for a first image of 540p and 5Mbps obtained by applying AI downscale to an original image of 1080p and 10Mbps to the server 1950
- the electronic device 1920 of listener B transmits video data for the original video of 1080p and 10 Mbps to the server 1950
- the electronic device 1930 of listener C transmits video data of the original video of 1080p and 10 Mbps to the server ( 1950) to be transmitted.
- the server 1950 transmits image data and AI data for the first image obtained by AI downscaling the original video of 1080p and 10 Mbps from the electronic device of listener B to the first image of 540p and 5 Mbps by AI downscaling the original video of 1080p and 10 Mbps from the electronic device of speaker A (1910 ), and the video data and AI data for the first video obtained by AI downscaling the original video of 1080p and 10Mbps of listener C to the first video of 540p and 5Mbps are transmitted to the electronic device 1910 of presenter A. do.
- the server 1950 transmits image data for a third image obtained by AI upscaling a 540p and 5Mbps first video of participant A, a presenter with a high importance, to 1080p and 10Mbps by the server and 1080p and 10Mbps of a listener C's electronic device.
- the video data for the first video obtained by AI downscaling the original video of 540p and 5 Mbps to the first video is transmitted to the electronic device 1920 of listener B, and the server 1950 is presenter A, who is a presenter of high importance.
- Image data for the third image obtained by AI upscaling the first video of 540p and 5Mbps to 1080p and 10Mbps and the original video of 1080p and 10Mbps of the listener B's electronic device AI downscaled to the first video of 540p and 5Mbps Image data of the first image obtained by scaling is transmitted to the electronic device 1930 of listener C.
- the presenter A's electronic device 1910 transmits 540p and 5 Mbps of the transmitted listener B based on the importance information. 1 video and the first video of 540p and 5 Mbps of listener C are displayed without AI upscaling.
- the electronic device 1920 of listener B does not support AI upscaling.
- the server 1950 receives and displays video data of a third video of 1080p and 10Mbps obtained by AI upscaling the first video of 540p and 5Mbps of presenter A, and displays the video data of the third video of 540p and 5Mbps of listener C. Image data of the first image is received and displayed.
- the electronic device 1930 of listener C does not support AI upscaling.
- the server 1950 receives and displays the video data of the third video of 1080p and 10Mbps obtained by AI upscaling the first video of 540p and 5Mbps of presenter A, and displays the video data of the third video of 540p and 5Mbps of listener B. Image data of the first image is received and displayed.
- the electronic device 1920 of the participant B who has been changed to the presenter does not support AI upscaling, so the original video of presenter B is 1080p.
- the server 1930 transmits the video data of the original video as it is without performing AI downscaling, and to the electronic device 1910 of listener A, the server 1950 transmits the original video of 1080p. Image data and AI data of the 540p first image obtained by AI downscaling are transmitted.
- the electronic device 1930 of listener C restores the original video data and displays the original 1080p video
- the electronic device 1910 of listener A AI upscales the AI downscaled first video based on the importance and displays the restored video.
- a third image of 1080p is displayed. A specific example for this case will be described later with reference to FIG. 21 .
- 20 is an electronic device supporting AI downscaling and AI upscaling in a video conference, an electronic device not supporting AI downscaling and AI upscaling, and a server supporting AI downscaling and AI upscaling according to another embodiment. It is a diagram for explaining the data transmission relationship between
- the electronic device 2020 of participant (listener) B and the electronic device 2030 of participant (listener) C do not support AI downscale and AI upscale, and the electronic device 2010 of participant (presenter) A is assumed to support AI downscale and AI upscale.
- the electronic device 2010 of the participant (presenter) A and the server 2050 that manages the video conference support AI downscaling and AI upscaling, and the remaining participants (listeners) ) Since the electronic device 2020 of B and the electronic device 2030 of participant (listener) C do not support AI downscale and AI upscale, the electronic device 2020 of listener B and the electronic device 2030 of listener C transmits the original video to the server 2050 as it is. However, the presenter A's electronic device 2010 transmits the AI downscaled video to the server 2050, and the server 2050 also transmits the received listener B's electronic device 2020 and listener C's electronic device 2030.
- AI downscales the original videos to obtain AI downscaled videos and transmits them to other participants, reducing the amount of uploading data of the electronic device of participant A, the presenter, and downloading data of the electronic devices of other participants B and C. usage is reduced.
- AI downscales only by 1/2, and the remaining listeners B and C have relatively low importance, so AI downscales by 1/4. Accordingly, images with relatively high importance are relatively less AI downscaled, so important information is better restored with less loss of information, and images with relatively lower importance are downscaled by AI more, resulting in higher data usage. can be reduced
- the electronic device 2010 of presenter A transmits video data and AI data for a first video of 540p and 5 Mbps to which AI downscale is applied to the original video of 1080p and 10 Mbps to the server 2050, and
- the electronic device 1620 transmits video data for the original video of 1080p and 10Mbps to the server 2050, and the electronic device 1630 of listener C transmits video data of the original video of 1080p and 10Mbps to the server 2050. send.
- the server 2050 transmits the video data and AI data of the first video obtained by AI downscaling the original video of 1080p and 10 Mbps of listener B to the first video of 270p and 2.5 Mbps to the electronic device 2010 of presenter A. and AI downscales the original video of 1080p and 10Mbps of listener C to the first video of 270p and 2.5Mbps, and transmits video data and AI data obtained to presenter A's electronic device 2010.
- the server 2050 transmits the image data of the third image obtained by AI upscaling the 540p and 5Mbps video of participant A, a presenter with high importance, to 1080p and 10Mbps, and the first video of 270p and 2.5Mbps of listener C. Transmits the video data to the electronic device 2020 of listener B, and the server 2050 AI upscales the first video of 540p and 5Mbps to 1080p and 10Mbps of participant A, a presenter with high importance, The video data and the video data of the first video of 270p and 2.5 Mbps of the listener B are transmitted to the electronic device 2030 of the listener C.
- presenter A's electronic device 2010 Since presenter A's electronic device 2010 has relatively low importance to the images of listeners B and C, who are not presenters, the presenter A's electronic device 2010 transmits 270p and 2.5 Mbps of listener B based on the importance information.
- the first video and the first video of 270p and 2.5 Mbps of listener C are displayed without AI upscaling, and in the electronic device 2020 of listener B, the video of participant A, the presenter, has a high importance, and the video of listener C has a high importance. Since is low, the electronic device 2020 of listener B, which does not support AI upscaling, is obtained by AI upscaling the first video of 540p and 5 Mbps of participant A by the server 2050 instead of the electronic device 2020 of listener B.
- the video data of the third video of 1080p and 10Mbps is received and displayed, the video data of the first video of 270p and 2.5Mbps of the received listener C is received and displayed, and the electronic device 2030 of the listener C receives and displays the video data of the listener C. Since the video of participant A has a high importance and the video of listener B has a low importance, the electronic device 2030 of listener C, which does not support AI upscaling, uses the server 2050 as the presenter instead of the electronic device 2030 of listener C.
- the video data of the third video of 1080p and 10Mbps obtained by AI upscaling the first video of 540p and 5Mbps of A is received and displayed, and the video data of the first video of 270p and 2.5Mbps of listener B transmitted is transmitted.
- the degree of AI downscaling may vary according to importance. That is, the presenter's video with high importance is AI downscaled by 1/2, and the listener's video with low importance is AI downscaled by 1/4. In this case, the video of the presenter having a high importance may be restored as an original video, and the video of the listener having a low importance may be restored as an image of 1/2 resolution, which is twice as large.
- 21 is an electronic device supporting AI downscaling and AI upscaling in a video conference, an electronic device not supporting AI downscaling and AI upscaling, and a server supporting AI downscaling and AI upscaling according to another embodiment; It is a diagram for explaining the data transmission relationship between
- the electronic device 2120 of participant (presenter) B and the electronic device 2130 of participant (listener) C do not support AI downscaling and AI upscaling, and the electronic device 2110 of participant (listener) A is assumed to support AI downscale and AI upscale.
- the electronic device 2110 of participant (listener) A and the server 2150 used for the video conference support AI downscaling and AI upscaling, and the remaining participants (presenters, ) Since the electronic device 2120 of B and the electronic device 2130 of participant (listener) C do not support AI downscale and AI upscale, the electronic device 2120 of presenter B and the electronic device 1730 of listener C transmits the original video to the server 2150 as it is. However, the electronic device 2110 of listener A transmits the AI downscaled video to the server 2150, and the server 2150 also transmits the electronic device 2120 of presenter B and the electronic device 2130 of listener C.
- the AI downscaled images are obtained by AI downscaling the original images and transmitted to other participants, so that the amount of uploading data of participant A's electronic device is reduced and the amount of downloading data of other participants B and C's electronic devices is reduced. this is reduced
- the video conference video of listener A is AI downscaled in the electronic device 2110 of listener A, and the AI downscaled video is sent to the server 2150. Since the electronic device 2120 of presenter B and the electronic device 2130 of listener C do not support AI downscaling, the original video of presenter B and the original video of listener C are transmitted to the server 2150.
- the server 2150 transmits the AI downscaled video transmitted from the electronic device 2110 of listener A to the electronic device 2120 of presenter B and the electronic device 2130 of listener C, and the server 2150 sends the electronic device 2130 of listener C.
- the server 2150 AI downscales the original video transmitted from the electronic device 2130 and transmits the first video to the electronic device 2110 of listener A and the electronic device 2120 of presenter B. Since participant B is a presenter, the importance is high. Therefore, the server 2150 transmits an AI downscaled image of the original video transmitted from the presenter B's electronic device 2120 to an electronic video of listener A that supports AI upscaling.
- the server 2150 transmits the original video transmitted from the presenter B's electronic device 2120 to the listener C's electronic device 1730 that does not support AI upscaling.
- the electronic device 2110 of the listener A transmits video data and AI data for a first video of 540p and 5 Mbps to which AI downscale is applied to the original video of 1080p and 10 Mbps to the server 2150, and
- the electronic device 2120 transmits video data for the original video of 1080p and 10Mbps to the server 1750, and the electronic device 2130 of presenter C transmits the video data of the original video of 1080p and 10Mbps to the server 1750. send.
- the server 2150 transmits video data and AI data of the first video obtained by AI downscaling the original video of 1080p and 10Mbps of presenter B to the first video of 540p and 5Mbps to the electronic device 2110 of listener A, , Image data and AI data of the first image obtained by AI downscaling the original video of 1080p and 10Mbps of listener C to the first video of 540p and 5Mbps are transmitted to the electronic device 2110 of listener A.
- the server 2150 transmits the video data of the first video of 540p and 5Mbps of listener A and the video data of the first video of 540p and 5Mbps of listener C to the electronic device 2120 of presenter B, and the server 2150 transmits The video data of the first video of 540p and 5Mbps of listener A and the video data of the original video of 1080p and 10Mbps of participant B, a presenter with high importance, are transmitted to the electronic device 1730 of listener C without AI downscaling.
- the first video of 540p and 5Mbps of presenter B transmitted by AI downscaling by the server 2150 is AI upscaled 1080p and 10Mbps based on the importance information. Based on the importance information, the first video that is not upscaled by AI is displayed based on the importance information of the first video of 540p and 5 Mbps of listener C transmitted by AI downscaling and transmission by the server 2150.
- the electronic device 2110 of listener A AI downscales and transmits the first video of listener A at 540p and 5Mbps and the server 2150 transmits the AI downscaled video.
- the first video of 540p and 5Mbps of listener C transmitted after being scaled is displayed.
- the electronic device 2120 of listener C does not support AI upscaling, the first video of 540p and 5 Mbps of participant A and the first video of 1080p and 10 Mbps of presenter B transmitted by AI downscaling of the electronic device 2110 of listener A The original video is displayed.
- the aforementioned electronic devices may include various types of devices capable of reproducing images, such as smart phones, tablet PCs, wearable devices, laptop computers, and desktop PCs.
- 22 is a diagram for explaining arrangement and importance of video conference images displayed on electronic devices participating in a video conference.
- a presenter when there are three electronic devices participating in a video conference and the type of video conference is a presentation, a presenter is determined to have a high importance in the video conference, and the electronic devices of the video conference participants are determined to be of high importance.
- the presenter's image is displayed in the largest area 2201, and the remaining listeners' images are displayed in the remaining areas 2202 and 2203.
- video conference images are displayed in equivalent areas 2251, 2252, 2253, and 2254.
- a presenter is determined to have a high importance in the video conference, and the electronic device 2250 participating in the video conference video conference
- the video conference video is displayed in equal areas 2251, 2252, 2253, and 2254, and the video conference video quality of the presenter is displayed higher than that of other listeners.
- FIG. 23 is a flowchart illustrating a method of AI upscaling a video conference image according to the importance of other electronic devices by an electronic device participating in a video conference during a video conference according to an embodiment.
- step S2310 image data generated as a result of a first encoding of a first image related to another electronic device participating in the video conference and AI data related to AI downscaling from the original image to the first image are acquired from the server. .
- the first image may be one of an image obtained by AI downscaling from an original image by a server or an image obtained by AI downscaling from an original image by another electronic device.
- step S2320 a second image corresponding to the first image is obtained by first decoding the image data.
- step S2330 it is determined whether to perform AI upscaling on the second image based on the importance of other electronic devices.
- importance may be identified from AI data.
- it may be determined to perform AI upscaling if the importance indicates a presenter, and if the importance is changed to a listener during a video conference, it may be determined not to perform AI upscaling.
- step S2340 if it is determined that AI upscaling is to be performed, the second image is AI upscaled through the upscaling DNN to obtain a third image, and the third image is provided to the display.
- step S2350 if it is determined not to perform AI upscaling, a second image is provided to the display.
- the importance of the user of the electronic device who first establishes the video conference may be initially set as the presenter.
- the importance may be set to presenters when the importance is high, and to listeners when the importance is low.
- the importance level may be changed according to an input received by an electronic device participating in a video conference or another electronic device.
- the input received by the other electronic device is one of an input activating a raising hand function of the other electronic device or an input requesting permission to speak from the other electronic device, and the input received by the electronic device is another electronic device. It may be an input for enlarging a video conference video for .
- an input received by the electronic device or another electronic device may be an input for changing the type of video conference.
- the importance may be determined according to at least one of the number of video conference participants and the type of video conference.
- the type of video conference may be one of presentation and discussion.
- the size of the original image may be 1080p, and the size of the first image may be 540p.
- the size of the original image may be 1080p, and the size of the first image may be 270p.
- the size of the original video is 1080p
- the size of the first video is 540p if the importance indicates that the user of the electronic device is a presenter
- the size of the first video is 270p if the importance indicates that the user is a listener.
- DNN setting information of the downscaling DNN may be obtained through joint training of the downscaling DNN and the upscaling DNN for AI upscaling of the first image.
- DNN setting information for AI upscaling of the second image among a plurality of DNN setting information is obtained based on AI data, and the DNN for upscaling may operate with the acquired DNN setting information.
- one DNN setting information for AI upscaling of the second image stored in the electronic device is acquired based on AI data, and the DNN for upscaling may operate with the acquired DNN setting information.
- FIG. 24 illustrates a video conference video by a server managing a video conference, acquiring an AI downscaled image from an electronic device participating in the video conference, and determining whether to support AI upscaling according to importance information of the electronic device. It is a flowchart for explaining a method of AI up-scaling and transmitting to another electronic device.
- step S2410 the server 1500 transmits first image data generated as a result of first encoding the first image from the first electronic device participating in the video conference and AI data related to AI downscaling from the original image to the first image.
- the first electronic device transmits AI data related to AI downscaling, it can be known that the first electronic device is an electronic device supporting AI downscaling.
- step S2420 the server 1500 first decodes the first image data to obtain a second image corresponding to the first image.
- step S2430 if the importance indicates that the user of the first electronic device is a listener, the server 1500 transmits second image data obtained by first encoding the second image to the second electronic device.
- step S2430 if the importance level indicates that the user of the first electronic device is a presenter, the server 1500 AI upscales the second image through the DNN for upscaling to obtain a third image, first encodes the third image, and The acquired third image data is transmitted to the second electronic device.
- the second electronic device is an electronic device that does not support AI upscaling because the server 1500 performs AI upscaling by determining whether to perform AI upscaling instead.
- the importance of the first electronic device may be identified from the AI data.
- the importance of the user of the electronic device that established the video conference may be initially set to the presenter.
- the importance level may be changed according to an input received by the first electronic device or the second electronic device participating in the video conference.
- the input received by the second electronic device is one of an input activating a raising hand function of the second electronic device or an input requesting permission to present by the second electronic device
- the first electronic device receives
- the input for enlarging the video conference image of the second electronic device may be an input for enlarging the video conference image.
- an input received by the first electronic device or the second electronic device may be an operation of changing a video conference type.
- the importance may be determined according to at least one of the number of video conference participants and the type of video conference.
- the type of video conference may be one of presentation and discussion.
- the size of the original image may be 1080p, and the size of the first image may be 540p.
- the size of the original image may be 1080p, and the size of the first image may be 270p.
- the size of the first video is 540p and the importance indicates that the user is a participant
- the size of the first image is It may be 270p.
- the first electronic device may support AI downscaling and the second electronic device may not support AI upscaling.
- the server 1600 may obtain fourth image data generated as a result of first encoding an original image from a third electronic device.
- the server 1600 may obtain an original image by first decoding the fourth image data.
- the server 1600 obtains a first image by AI downscaling the original image using a DNN for downscaling, and first encodes the first image to obtain fifth image data. is transmitted to the second electronic device, and if the importance of the third electronic device indicates that the speaker is a presenter, the fourth image data or image data obtained by first encoding the original image may be transmitted to the second electronic device.
- the third electronic device does not support AI downscaling and the second electronic device does not support AI upscaling.
- the server 1600 obtains a first image by AI downscaling the original image using a DNN for downscaling, and converts the first image into a first Encoded fifth image data and AI data related to the AI downscaling may be transmitted to the second electronic device.
- DNN setting information of the downscaling DNN may be obtained through joint training of the downscaling DNN and the upscaling DNN for AI upscaling of the first image.
- DNN setting information for AI upscaling of the second image among a plurality of DNN setting information is obtained based on AI data, and the DNN for upscaling may operate with the acquired DNN setting information.
- one DNN setting information for AI upscaling of the second image stored in the server is obtained based on AI data, and the DNN for upscaling may operate with the acquired DNN setting information.
- whether or not the second electronic device supports AI upscaling may be determined based on whether AI data obtained from the second electronic device exists.
- whether or not the second electronic device supports AI upscaling may be determined based on AI downscaling support information obtained from the second electronic device.
- the above-described embodiments of the present disclosure can be written as programs or instructions that can be executed on a computer, and the written programs or instructions can be stored in a medium.
- the medium may continuously store programs or instructions executable by a computer, or temporarily store them for execution or download.
- the medium may be various recording means or storage means in the form of a single or combined hardware, but is not limited to a medium directly connected to a certain computer system, and may be distributed on a network.
- Examples of the medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROM and DVD, magneto-optical media such as floptical disks, and ROM, RAM, flash memory, etc. configured to store program instructions.
- examples of other media include recording media or storage media managed by an app store that distributes applications, a site that supplies or distributes various other software, and a server.
- the above-described DNN-related model may be implemented as a software module.
- the DNN model When implemented as a software module (eg, a program module including instructions), the DNN model may be stored in a computer-readable recording medium.
- the DNN model may be integrated in the form of a hardware chip and become a part of the above-described AI decoding apparatus 200 or AI encoding apparatus 600.
- a DNN model can be built in the form of a dedicated hardware chip for artificial intelligence, or built as part of an existing general-purpose processor (eg CPU or application processor) or graphics-only processor (eg GPU). It could be.
- the DNN model may be provided in the form of downloadable software.
- a computer program product may include a product in the form of a software program (eg, a downloadable application) that is distributed electronically by a manufacturer or through an electronic marketplace. For electronic distribution, at least a portion of the software program may be stored on a storage medium or may be temporarily created.
- the storage medium may be a storage medium of a manufacturer or a server of an electronic market or a relay server.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Graphics (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Processing (AREA)
Abstract
Description
Claims (14)
- AI(artificial intelligence)를 이용하여 화상 회의에 참여하는 전자 장치에 있어서,디스플레이;상기 전자 장치에 저장된 하나 이상의 인스트럭션들을 실행하는 프로세서를 포함하고,상기 프로세서는,상기 화상 회의에 참여 중인 다른 전자 장치와 관련된 제 1 영상에 대한 제 1 부호화 결과로 생성되는 영상 데이터, 및 원본 영상으로부터 상기 제 1 영상으로의 AI 다운스케일과 관련된 AI 데이터를 서버로부터 획득하고,상기 영상 데이터를 제 1 복호화하여 상기 제 1 영상에 대응되는 제 2 영상을 획득하고,상기 다른 전자 장치에 대한 중요도에 기초하여, 상기 제 2 영상에 대한 AI 업스케일의 수행 여부를 결정하고,상기 AI 업스케일을 수행하는 것으로 결정되면, 업스케일용 DNN을 통해 상기 제 2 영상을 AI 업스케일하여 제 3 영상을 획득하고,상기 제 3 영상을 상기 디스플레이에 제공하고,상기 AI 업스케일을 수행하지 않는 것으로 결정되면, 상기 제 2 영상을 상기 디스플레이에 제공하는, 전자 장치.
- 제 1 항에 있어서,상기 다른 전자 장치에 대한 상기 중요도는 상기 AI 다운스케일과 관련된 상기 AI 데이터로부터 확인되는, 전자 장치.
- 제 1 항에 있어서,상기 제 1 영상은, 상기 원본 영상으로부터 상기 서버에 의해 AI 다운스케일된 영상 또는 상기 원본 영상으로부터 상기 다른 전자 장치에 의해 AI 다운스케일된 영상인, 전자 장치.
- 제 1 항에 있어서,상기 다른 전자 장치에 대한 상기 중요도가 상기 다른 전자 장치의 사용자가 발표자임을 나타내면, 상기 AI 업스케일을 수행하는 것으로 결정되고,상기 다른 전자 장치에 대한 상기 중요도가 상기 다른 전자 장치의 상기 사용자가 청취자임을 나타내면, 상기 AI 업스케일을 수행하지 않는 것으로 결정되는, 전자 장치.
- 제 4 항에 있어서,상기 다른 전자 장치에 대한 상기 중요도가 상기 다른 전자 장치의 사용자가 상기 발표자임을 나타내면, 상기 AI 업스케일을 수행하는 것으로 결정되고,상기 화상 회의 도중에, 상기 다른 전자 장치에 대한 상기 중요도가 상기 다른 전자 장치의 사용자가 상기 청취자임을 나타내는 것으로 변경되면, 상기 AI 업스케일을 수행하기 않는 것으로 결정되는, 전자 장치.
- 제 1 항에 있어서,상기 화상 회의를 개설한 전자 장치의 사용자에 대한 중요도는 발표자로 초기 설정되는, 전자 장치.
- AI(artificial intelligence)를 이용하여 화상 회의에 참여하는 전자 장치의 화상 회의 영상 처리 방법에 있어서,상기 화상 회의에 참여 중인 다른 전자 장치와 관련된 제 1 영상에 대한 제 1 부호화 결과로 생성되는 영상 데이터, 및 원본 영상으로부터 상기 제 1 영상으로의 AI 다운스케일과 관련된 AI 데이터를 서버로부터 획득하는 단계;상기 영상 데이터를 제 1 복호화하여 상기 제 1 영상에 대응되는 제 2 영상을 획득하는 단계;상기 다른 전자 장치에 대한 중요도에 기초하여, 상기 제 2 영상에 대한 AI 업스케일의 수행 여부를 결정하는 단계;상기 AI 업스케일을 수행하는 것으로 결정되면, 업스케일용 DNN을 통해 상기 제 2 영상을 AI 업스케일하여 제 3 영상을 획득하고, 상기 제 3 영상을 상기 디스플레이에 제공하는 단계; 및상기 AI 업스케일을 수행하지 않는 것으로 결정되면, 상기 제 2 영상을 상기 디스플레이에 제공하는 단계를 포함하는, 화상 회의 영상 처리 방법.
- AI(artificial intelligence)를 이용하여 화상 회의를 관리하는 서버에 있어서,상기 서버에 저장된 하나 이상의 인스트럭션들을 실행하는 프로세서를 포함하고,상기 프로세서는,상기 화상 회의에 참여한 제1 전자 장치로부터 제 1 영상에 대한 제 1 부호화 결과로 생성된 제1 영상 데이터, 원본 영상으로부터 상기 제 1 영상으로의 AI 다운스케일과 관련된 AI 데이터를 획득하고,상기 제1 영상 데이터를 제 1 복호화하여 상기 제 1 영상에 대응되는 제 2 영상을 획득하고,상기 제1 전자 장치에 대한 중요도가 상기 제1 전자 장치의 사용자가 청취자임을 나타내면, 상기 제2 영상을 제1 부호화하여 획득한 제 2 영상 데이터를 제2 전자 장치에 전송하고,상기 제1 전자 장치에 대한 상기 중요도가 상기 제1 전자 장치의 상기 사용자가 발표자임을 나타내면, 업스케일용 DNN을 통해 상기 제 2 영상을 AI 업스케일하여 제 3 영상을 획득하고, 상기 제 3 영상을 제1 부호화하여 획득한 제 3 영상 데이터를 상기 제2 전자 장치에 전송하는, 서버.
- 제 8 항에 있어서,상기 제1 전자 장치에 대한 상기 중요도는 상기 AI 다운스케일과 관련된 상기 AI 데이터로부터 확인되는, 서버.
- 제 8 항에 있어서,상기 제1 전자 장치는 AI 다운스케일을 지원하는 것인, 서버.
- 제 8 항에 있어서,상기 제2 전자 장치는 AI 업스케일을 지원하지 않는 것인, 서버.
- 제 8 항에 있어서,제3 전자 장치로부터 원본 영상에 대한 제 1 부호화 결과로 생성된 제 4 영상 데이터를 획득하고,상기 제 4 영상 데이터를 제 1 복호화하여 상기 원본 영상을 획득하고,상기 제3 전자 장치에 대한 중요도가 상기 제3 전자 장치의 사용자가 상기 청취자임을 나타내면, 상기 원본 영상을 다운스케일용 DNN을 이용하여 AI 다운스케일하여 제 1 영상을 획득하고, 제 1 영상을 제1 부호화한 제 5 영상데이터를 상기 제2 전자 장치에 전송하고,상기 제 3 전자 장치에 대한 상기 중요도가 상기 제3 전자 장치의 상기 사용자가 상기 발표자임을 나타내면, 상기 제 4 영상 데이터를 상기 제2 전자 장치에 전송하는, 서버.
- 제 12 항에 있어서,상기 제3 전자 장치는 AI 다운스케일을 지원하지 않는 것인, 서버.
- 제 12 항에 있어서,상기 제 2 전자 장치가 AI 업스케일을 지원하면, 상기 프로세서는 상기 원본 영상을 상기 다운스케일용 DNN을 이용하여 AI 다운스케일하여 제 1 영상을 획득하고, 상기 제 1 영상을 제1 부호화한 제4 영상 데이터 및 상기 AI 다운스케일과 관련된 상기 AI 데이터를 상기 제2 전자 장치에 전송하는, 서버.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202280057322.2A CN117859321A (zh) | 2021-08-25 | 2022-08-24 | 在视频会议期间使用人工智能(ai)进行缩小和放大的方法和设备 |
EP22861712.2A EP4362454A4 (en) | 2021-08-25 | 2022-08-24 | METHOD AND DEVICE FOR SCALE REDUCTION AND SCALE ENLARGEMENT USING ARTIFICIAL INTELLIGENCE (AI) DURING VIDEO CONFERENCE |
US17/895,561 US20230085530A1 (en) | 2021-08-25 | 2022-08-25 | Method and apparatus for artificial intelligence downscaling and upscaling during video conference |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020210112653A KR20230030435A (ko) | 2021-08-25 | 2021-08-25 | 화상회의시 AI(artificial intelligence)를 이용한 다운스케일 및 업스케일 방법 및 장치 |
KR10-2021-0112653 | 2021-08-25 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/895,561 Continuation US20230085530A1 (en) | 2021-08-25 | 2022-08-25 | Method and apparatus for artificial intelligence downscaling and upscaling during video conference |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023027503A1 true WO2023027503A1 (ko) | 2023-03-02 |
Family
ID=85321942
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2022/012646 WO2023027503A1 (ko) | 2021-08-25 | 2022-08-24 | 화상회의시 ai(artificial intelligence)를 이용한 다운스케일 및 업스케일 방법 및 장치 |
Country Status (5)
Country | Link |
---|---|
US (1) | US20230085530A1 (ko) |
EP (1) | EP4362454A4 (ko) |
KR (1) | KR20230030435A (ko) |
CN (1) | CN117859321A (ko) |
WO (1) | WO2023027503A1 (ko) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20180082672A (ko) * | 2017-01-09 | 2018-07-19 | 한국전자통신연구원 | 영상 회의 방청 서비스 제공 방법 및 장치 |
JP2020053741A (ja) * | 2018-09-25 | 2020-04-02 | 京セラドキュメントソリューションズ株式会社 | テレビ会議装置及びテレビ会議プログラム |
KR20200044666A (ko) * | 2018-10-19 | 2020-04-29 | 삼성전자주식회사 | 데이터 스트리밍 방법 및 장치 |
KR20200140368A (ko) * | 2018-05-07 | 2020-12-15 | 애플 인크. | 화상 회의를 위한 보충 콘텐츠를 이용한 비디오 스트림들의 수정 |
KR20210050186A (ko) * | 2019-10-28 | 2021-05-07 | 삼성전자주식회사 | 영상의 ai 부호화 및 ai 복호화 방법, 및 장치 |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
BR102015005929B1 (pt) * | 2015-03-17 | 2021-11-30 | Samsung Eletrônica Da Amazônia Ltda | Sistema e método para composição e compressão de vídeo com base em contexto a partir de objetos com resolução espacial normalizada |
WO2020080873A1 (en) * | 2018-10-19 | 2020-04-23 | Samsung Electronics Co., Ltd. | Method and apparatus for streaming data |
US10944996B2 (en) * | 2019-08-19 | 2021-03-09 | Intel Corporation | Visual quality optimized video compression |
KR20210067788A (ko) * | 2019-11-29 | 2021-06-08 | 삼성전자주식회사 | 전자 장치, 시스템 및 그 제어 방법 |
-
2021
- 2021-08-25 KR KR1020210112653A patent/KR20230030435A/ko unknown
-
2022
- 2022-08-24 CN CN202280057322.2A patent/CN117859321A/zh active Pending
- 2022-08-24 WO PCT/KR2022/012646 patent/WO2023027503A1/ko active Application Filing
- 2022-08-24 EP EP22861712.2A patent/EP4362454A4/en active Pending
- 2022-08-25 US US17/895,561 patent/US20230085530A1/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20180082672A (ko) * | 2017-01-09 | 2018-07-19 | 한국전자통신연구원 | 영상 회의 방청 서비스 제공 방법 및 장치 |
KR20200140368A (ko) * | 2018-05-07 | 2020-12-15 | 애플 인크. | 화상 회의를 위한 보충 콘텐츠를 이용한 비디오 스트림들의 수정 |
JP2020053741A (ja) * | 2018-09-25 | 2020-04-02 | 京セラドキュメントソリューションズ株式会社 | テレビ会議装置及びテレビ会議プログラム |
KR20200044666A (ko) * | 2018-10-19 | 2020-04-29 | 삼성전자주식회사 | 데이터 스트리밍 방법 및 장치 |
KR20210050186A (ko) * | 2019-10-28 | 2021-05-07 | 삼성전자주식회사 | 영상의 ai 부호화 및 ai 복호화 방법, 및 장치 |
Non-Patent Citations (1)
Title |
---|
See also references of EP4362454A4 * |
Also Published As
Publication number | Publication date |
---|---|
EP4362454A4 (en) | 2024-06-19 |
CN117859321A (zh) | 2024-04-09 |
US20230085530A1 (en) | 2023-03-16 |
KR20230030435A (ko) | 2023-03-06 |
EP4362454A1 (en) | 2024-05-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021086016A2 (en) | Apparatus and method for performing artificial intelligence (ai) encoding and ai decoding on image | |
WO2020080765A1 (en) | Apparatuses and methods for performing artificial intelligence encoding and artificial intelligence decoding on image | |
WO2020080873A1 (en) | Method and apparatus for streaming data | |
WO2020080827A1 (en) | Ai encoding apparatus and operation method of the same, and ai decoding apparatus and operation method of the same | |
WO2021033867A1 (en) | Decoding apparatus and operating method of the same, and artificial intelligence (ai) up-scaling apparatus and operating method of the same | |
WO2020246756A1 (en) | Apparatus and method for performing artificial intelligence encoding and artificial intelligence decoding on image | |
EP3868096A1 (en) | Artificial intelligence encoding and artificial intelligence decoding methods and apparatuses using deep neural network | |
WO2020080698A1 (ko) | 영상의 주관적 품질을 평가하는 방법 및 장치 | |
EP3811618A1 (en) | Method and apparatus for streaming data | |
WO2021251611A1 (en) | Apparatus and method for performing artificial intelligence encoding and decoding on image by using low-complexity neural network | |
WO2021107275A1 (en) | Electronic apparatus, system and controlling method thereof | |
WO2021107406A1 (en) | Electronic device, control method thereof, and system | |
WO2021177652A1 (ko) | 피쳐 양자화/역양자화를 수행하는 영상 부호화/복호화 방법, 장치 및 비트스트림을 저장하는 기록 매체 | |
WO2021172834A1 (en) | Apparatus and method for performing artificial intelligence encoding and artificial intelligence decoding on image by using pre-processing | |
WO2020080782A1 (en) | Artificial intelligence (ai) encoding device and operating method thereof and ai decoding device and operating method thereof | |
WO2021242066A1 (en) | Apparatus and method for performing artificial intelligence encoding and artificial intelligence decoding on image | |
EP3868097A1 (en) | Artificial intelligence (ai) encoding device and operating method thereof and ai decoding device and operating method thereof | |
WO2021091178A1 (en) | Artificial intelligence (ai) encoding apparatus and operating method thereof and ai decoding apparatus and operating method thereof | |
WO2021086032A1 (en) | Image encoding method and apparatus and image decoding method and apparatus | |
WO2021162446A1 (en) | Method and apparatus for streaming vr image | |
WO2020080709A1 (en) | Artificial intelligence encoding and artificial intelligence decoding methods and apparatuses using deep neural network | |
WO2020251278A1 (ko) | 크로마 양자화 파라미터 데이터 기반 영상 디코딩 방법 및 그 장치 | |
WO2019031735A1 (en) | IMAGE PROCESSING APPARATUS, IMAGE PROCESSING METHOD, AND IMAGE DISPLAY SYSTEM | |
WO2021054697A1 (ko) | 영상의 ai 부호화 방법 및 장치, 영상의 ai 복호화 방법 및 장치 | |
WO2020080751A1 (ko) | 부호화 방법 및 그 장치, 복호화 방법 및 그 장치 |
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: 22861712 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2022861712 Country of ref document: EP |
|
ENP | Entry into the national phase |
Ref document number: 2022861712 Country of ref document: EP Effective date: 20240126 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202280057322.2 Country of ref document: CN |
|
NENP | Non-entry into the national phase |
Ref country code: DE |