WO2025201291A1 - 视频鉴别信息添加的方法、装置、设备、存储介质及产品 - Google Patents

视频鉴别信息添加的方法、装置、设备、存储介质及产品

Info

Publication number
WO2025201291A1
WO2025201291A1 PCT/CN2025/084595 CN2025084595W WO2025201291A1 WO 2025201291 A1 WO2025201291 A1 WO 2025201291A1 CN 2025084595 W CN2025084595 W CN 2025084595W WO 2025201291 A1 WO2025201291 A1 WO 2025201291A1
Authority
WO
WIPO (PCT)
Prior art keywords
video
dynamic
information
static
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2025/084595
Other languages
English (en)
French (fr)
Other versions
WO2025201291A9 (zh
Inventor
邓米雪
黄敏峰
程宝平
王乐才
王茜姿
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Mobile Communications Group Co Ltd
China Mobile Hangzhou Information Technology Co Ltd
Original Assignee
China Mobile Communications Group Co Ltd
China Mobile Hangzhou Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Mobile Communications Group Co Ltd, China Mobile Hangzhou Information Technology Co Ltd filed Critical China Mobile Communications Group Co Ltd
Publication of WO2025201291A1 publication Critical patent/WO2025201291A1/zh
Publication of WO2025201291A9 publication Critical patent/WO2025201291A9/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/503Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
    • H04N19/51Motion estimation or motion compensation
    • H04N19/513Processing of motion vectors
    • H04N19/517Processing of motion vectors by encoding
    • H04N19/52Processing of motion vectors by encoding by predictive encoding
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/46Embedding additional information in the video signal during the compression process
    • H04N19/467Embedding additional information in the video signal during the compression process characterised by the embedded information being invisible, e.g. watermarking

Definitions

  • the present application relates to the field of network security technology, and in particular to a method, apparatus, device, storage medium and product for adding video identification information.
  • identification information such as invisible watermarks
  • identification information has important applications in copyright protection, content authentication, and source tracking. For example, adding an invisible watermark to a video does not affect the user's viewing experience.
  • real-time video services such as live broadcasts, video conferencing, and real-time home security monitoring are longer than traditional videos and contain enormous amounts of data.
  • Embedding an invisible watermark requires calculations for each frame, consuming computing resources and time, resulting in a significant increase in computing power costs. This also fails to meet the real-time requirements of the service, making invisible watermarking inefficient.
  • the image group in the video is divided into a dynamic image group and a static image group;
  • the method before obtaining dynamic and static classification information of the image group in the video, the method further includes:
  • the above-mentioned implementation mode of the present application classifies the dynamic situation of the video by directly obtaining the motion vector information of the image group in the video during the encoding process.
  • the process of processing information is faster and more direct, which can speed up the processing of information.
  • the above-mentioned embodiment of the present application compares the positional relationship between each different image block in the image group and the image blocks in the reference image group to determine whether there is a motion relationship, and then obtains the motion vector information corresponding to each image group.
  • the motion vector information can be accurately obtained.
  • the motion vector information is input into a dynamic and static video classification model, and the dynamic and static video classification model is used to classify the image groups in the video according to the degree of motion represented by the motion vector information to obtain dynamic and static classification information corresponding to the image groups.
  • the above embodiment of the present application uses a pre-trained static and dynamic video classification model to classify the image groups in the video into static and dynamic based on motion vector information, thereby obtaining classification results quickly and accurately.
  • inputting the motion vector information into a dynamic and static video classification model, classifying the image groups in the video according to the degree of motion represented by the motion vector information using the dynamic and static video classification model, and obtaining dynamic and static classification information corresponding to the image groups includes:
  • the dynamic and static classification sub-model in the dynamic and static video classification model is used to perform dynamic and static classification on the image group in the video based on the total number of the motion vectors, the average value of the displacement, the maximum value of the displacement in the motion vector information, and the minimum value of the displacement in the motion vector information to obtain the dynamic and static classification information corresponding to the image group.
  • the above-described embodiment of the present application obtains the total number of motion vectors and the average value of each motion displacement value, and then classifies the video into static and dynamic based on four features: the total number of motion vectors, the average value of the displacement, the maximum displacement value in the motion vector information, and the minimum displacement value in the motion vector information.
  • the total number of motion vectors the average value of the displacement
  • the maximum displacement value in the motion vector information the maximum displacement value in the motion vector information
  • the minimum displacement value in the motion vector information By using more comprehensive motion vector information, dynamic and static classification information corresponding to the image group can be more accurately obtained.
  • the total number of motion vectors, the maximum value of the displacement, the minimum value of the displacement and the average value of the displacement of each image group are used as input data, and the initial support vector machine is trained according to the corresponding labels of the image group to obtain a dynamic and static classification sub-model.
  • the above-mentioned embodiment of the present application obtains historical video data and annotates the dynamic image group or static image group in each image group in the historical video data, and then obtains the total number of corresponding motion vectors, the maximum displacement, the minimum displacement, and the average displacement based on the motion vector information of each image group. These are then used to train an initial support vector machine, thereby obtaining a dynamic and static classification sub-model.
  • the initial support vector machine can be trained using historical video data to obtain an accurate dynamic and static classification sub-model.
  • the preset dynamic image adding frequency includes adding identification information in each frame whose interval number from the reference frame is an integer multiple of a first value
  • the preset static image adding frequency includes adding identification information in each frame whose interval number with the reference frame is an integer multiple of a second value, and the second value is greater than the first value.
  • the above embodiment of the present application adds identification information to frames that are integer multiples of the interval between the dynamic image group and the static image group, which can make the identification information more evenly distributed in the video.
  • the method before obtaining the dynamic and static classification information of the image group in the video, the method further includes:
  • the video information including motion vector information generated during encoding of a group of images in the original video, the motion vector information representing a degree of motion of the group of images;
  • the above-mentioned embodiment of the present application adds the dynamic and static classification information to the supplementary enhancement information frame and then adds the original video information, which makes the dynamic and static classification information easier to obtain and use.
  • the present application provides a device for adding video identification information, the device comprising:
  • an acquisition module configured to acquire dynamic and static classification information of an image group in a video, wherein the dynamic and static classification information is obtained by classifying the image group according to a degree of motion in the video;
  • a classification module configured to divide the image group in the video into a dynamic image group and a static image group according to the dynamic and static classification information
  • the present application provides a device for adding video identification information, the device comprising: a processor, and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the method for adding video identification information as described above.
  • the present application provides a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the method for adding video identification information as described above is implemented.
  • the present application provides a computer program product.
  • the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the method for adding video identification information as described above.
  • the above-described embodiment of the present application obtains dynamic and static classification information of image groups in a video and then divides the video information into dynamic image groups and static image groups based on the dynamic and static classification information. Then, based on the dynamic image and static image addition frequencies, identification information is added to specific frames of the dynamic image group and the static image group, respectively, to obtain a video with the identification information added. Because the preset dynamic image addition frequency and the preset static image addition frequency are less than the number of video frames, the number of added frames is reduced, thereby reducing the computing power consumption of adding identification information and the time required to add identification information. This can improve the efficiency of adding identification information and meet the high real-time requirements of the business.
  • FIG1 is a flow chart of a method for adding video identification information provided by one embodiment of the present application.
  • FIG2 is a schematic diagram of a motion vector in one embodiment of the present application.
  • FIG3 is a flow chart of another method for adding video identification information provided by an embodiment of the present application.
  • FIG4 is a schematic structural diagram of an apparatus for adding video identification information provided by one embodiment of the present application.
  • FIG5 is a schematic diagram of the hardware structure of a device for adding video identification information provided in an embodiment of the present application.
  • Supplemental Enhancement Information defined in the video bitstream, providing a method for adding information to the video bitstream.
  • the method for adding video identification information of the present application may be executed by the above-mentioned apparatus for adding video identification information, and the method includes the following steps S101 to S103 .
  • Step S101 obtaining dynamic and static classification information of an image group in a video, wherein the dynamic and static classification information is obtained by dividing the image group in the video according to the degree of motion.
  • a GOP is a collection of images that are closely connected in time, typically a continuous sequence of frames in a video. During video encoding, the video is typically divided into multiple GOPs.
  • a decoding operation is performed to convert the video file into a series of image groups.
  • the corresponding dynamic and static classification information for each image group is then obtained.
  • custom data containing dynamic and static classification information can be inserted into the video information.
  • the dynamic and static classification information can be extracted.
  • Step S102 dividing the image group in the video into a dynamic image group and a static image group according to the dynamic and static classification information.
  • the acquired dynamic and static classification information can be used to classify each image group into dynamic and static images.
  • the image group with a large motion amplitude is classified as a dynamic image group, and the image group with a small motion amplitude is classified as a static image group.
  • Step S103 adding identification information to specific frames of the dynamic image group according to a preset dynamic image adding frequency; adding identification information to specific frames of the static image group according to a preset static image adding frequency, to obtain a video with added identification information.
  • the identification information can be a specific mark, watermark or other form of identification code.
  • specific frames are selected from the dynamic image group according to a preset dynamic image addition frequency, and then the identification information is added to these selected frames.
  • a specific selection method can be to randomly select several frames from the image group according to a probability value.
  • the video contains sound information
  • the sound information of the video can be obtained, and the watermark is preferably added to the image corresponding to the sound information.
  • specific frames are selected from the static image group according to a preset static image addition frequency, and identification information is then added to these selected frames.
  • a sequence of video frames after the addition of identification information is obtained, and the video with the added identification information is reconstructed.
  • the preset dynamic image addition frequency and the preset static image addition frequency are both smaller than the number of video frames, eliminating the need to add identification information to every frame of the video.
  • the above-described embodiment of the present application obtains dynamic and static classification information of image groups in a video and then divides the video information into dynamic image groups and static image groups based on the dynamic and static classification information. Then, based on the dynamic image and static image addition frequencies, identification information is added to specific frames of the dynamic image group and the static image group, respectively, to obtain a video with the identification information added. Because the preset dynamic image addition frequency and the preset static image addition frequency are less than the number of video frames, the number of added frames is reduced, thereby reducing the computing power consumption of adding identification information and the time required to add identification information. This can improve the efficiency of adding identification information and meet the high real-time requirements of the business.
  • Video information of a video is obtained, where the video information includes motion vector information generated during encoding of a group of images in the video, where the motion vector information represents a degree of motion of the group of images.
  • video information is encoded, and during this process, motion estimation is performed on each image frame. By comparing the current frame with adjacent frames, the motion vector information of the image group is obtained.
  • Common methods include optical flow and global motion estimation.
  • the image groups in the video are divided according to the motion vector information to obtain dynamic and static classification information of the image groups in the video.
  • an appropriate motion vector threshold is set based on the application requirements and actual conditions. This threshold can be determined based on the size, direction, or other characteristics of the motion vector to distinguish between "moving" and “still” image groups.
  • a threshold value can be applied to the motion vector of each image group, classifying image groups with motion vectors greater than the threshold as dynamic image groups, while image groups with motion vectors less than the threshold as static image groups.
  • a difference calculation can be performed on the motion vectors of adjacent frames in an image group. If the difference value is greater than a threshold, the image group is classified as a dynamic image group; otherwise, it is classified as a static image group.
  • the above-mentioned implementation mode of the present application classifies the dynamic situation of the video by directly obtaining the motion vector information of the image group in the video during the encoding process.
  • the process of processing information is faster and more direct, which can speed up the processing of information.
  • obtaining video information of a video includes:
  • a reference image group is selected for each image group.
  • This reference image group can be the previous frame, an image group multiple frames ago, or a frame in the current image group. Based on the differences between the image group and the corresponding reference image group, their positional relationship is determined.
  • the position relationship it can be determined whether the position difference between the image group and the corresponding reference image group exceeds a preset threshold. If the position difference exceeds the preset threshold, it is determined that there is a motion relationship between it and the reference image group.
  • motion vector information corresponding to each image group is obtained according to the corresponding positional relationship.
  • feature points can be extracted from each image group.
  • the feature points of each image group are then matched with those of the corresponding reference image group, for example using a descriptor-based matching algorithm.
  • the feature point matching information is then used to calculate the motion transformation between the image group and the corresponding reference image group using a motion estimation algorithm.
  • Motion transformations include, but are not limited to, translation and rotation.
  • motion vector information corresponding to each image group can be calculated, such as the magnitude and direction of the displacement vector and the rotation angle.
  • dividing the image groups in the video according to the motion vector information to obtain the image groups in the video includes:
  • the motion vector information is input into a dynamic and static video classification model, and the dynamic and static video classification model is used to classify the image groups in the video according to the degree of motion represented by the motion vector information to obtain dynamic and static classification information corresponding to the image groups.
  • the dynamic and static video classification model can be a machine learning algorithm or a deep learning algorithm, which is used to classify image groups in a video according to their dynamic or static characteristics.
  • Machine learning algorithms such as support vector machines, decision trees, and random forests
  • deep learning algorithms such as convolutional neural networks
  • Cross-validation and regularization techniques can be used during training to improve model performance.
  • the above embodiment of the present application uses a pre-trained static and dynamic video classification model to classify the image groups in the video into static and dynamic based on motion vector information, thereby obtaining classification results quickly and accurately.
  • inputting the motion vector information into a dynamic and static video classification model, classifying the image groups in the video according to the degree of motion represented by the motion vector information using the dynamic and static video classification model, and obtaining dynamic and static classification information corresponding to the image groups includes:
  • the number of motion vectors in each image group may be accumulated to obtain the total number of motion vectors.
  • each image group its motion vector information is extracted, including the magnitude (i.e., displacement value) and direction of the motion vector.
  • the magnitude i.e., displacement value
  • direction of the motion vector For each image group's motion vector information, all its displacement values are accumulated and then divided by the total number of displacement values to obtain the average motion displacement value.
  • the motion vector information is used to calculate the required features, including but not limited to the total number of motion vectors, the average displacement, and the maximum and minimum displacement values. These features can be used to describe the dynamic characteristics of the image group.
  • the trained dynamic and static video classification model is applied to the new image group. The image group features are input into the trained classification model, which then outputs the corresponding dynamic and static classification information.
  • the above-described embodiment of the present application obtains the total number of motion vectors and the average value of each motion displacement value, and then classifies the video into static and dynamic based on four features: the total number of motion vectors, the average value of the displacement, the maximum displacement value in the motion vector information, and the minimum displacement value in the motion vector information.
  • the total number of motion vectors the average value of the displacement
  • the maximum displacement value in the motion vector information the maximum displacement value in the motion vector information
  • the minimum displacement value in the motion vector information By using more comprehensive motion vector information, dynamic and static classification information corresponding to the image group can be more accurately obtained.
  • the method before using the dynamic and static classification submodel in the dynamic and static video classification model to perform dynamic and static classification on the image group in the video based on the total number of motion vectors, the average value of the displacement, the maximum value of the displacement in the motion vector information, and the minimum value of the displacement in the motion vector information to obtain the dynamic and static classification information corresponding to the image group, the method further includes:
  • the historical video data is divided into a plurality of image groups, and labels of the plurality of image groups are obtained, where the labels are dynamic image groups or static image groups.
  • the total number of motion vectors, the maximum value of the displacement, the minimum value of the displacement and the average value of the displacement of each image group are used as input data, and the initial support vector machine is trained according to the corresponding labels of the image group to obtain a dynamic and static classification sub-model.
  • the above-mentioned embodiment of the present application obtains historical video data and annotates the dynamic image group or static image group in each image group in the historical video data, and then obtains the total number of corresponding motion vectors, the maximum displacement, the minimum displacement, and the average displacement based on the motion vector information of each image group. These are then used to train an initial support vector machine, thereby obtaining a dynamic and static classification sub-model.
  • the initial support vector machine can be trained using historical video data to obtain an accurate dynamic and static classification sub-model.
  • the preset static image adding frequency includes adding identification information in each frame whose interval number with the reference frame is an integer multiple of a second value, and the second value is greater than the first value.
  • the second value is greater than the first value, that is, the second value corresponding to the static image is larger, and the frequency of adding identification information is correspondingly greater, then the frequency of adding dynamic images is less than the frequency of adding static images.
  • the above embodiment of the present application adds identification information to frames that are integer multiples of the interval between the dynamic image group and the static image group, which can make the identification information more evenly distributed in the video.
  • the method before obtaining the dynamic and static classification information of the image group in the video, the method further includes:
  • Video information of an original video is obtained, where the video information includes motion vector information generated during encoding of a group of images in the original video, where the motion vector information represents a degree of motion of the group of images.
  • motion vector information between images is extracted for each image group.
  • the image groups in the original video are divided according to the motion vector information to obtain initial dynamic and static classification information of the image groups in the original video.
  • the image groups in the original video can be divided according to a set motion vector threshold.
  • the image group is considered to belong to the dynamic image group, otherwise it is considered to belong to the static image group.
  • the initial dynamic and static classification information is added to the supplementary enhancement information frame.
  • each image group of the original video may be traversed, and the initial dynamic and static classification information of the image group may be embedded in the corresponding supplementary enhancement information frame.
  • the initial supplementary enhancement information frame is added to the original video information to obtain the video, and the initial dynamic and static classification information is the dynamic and static information.
  • the supplementary enhancement information frame embedded with the classification information is added to the corresponding position of the original video. In this way, the original video information processed by the supplementary enhancement information frame is merged to generate a final video containing dynamic and static classification information.
  • Motion estimation is the process of searching for a matching block in a reference image during encoding, using a motion vector to represent the positional relationship between the reference block and the current block. If there is no motion relationship between the reference block and the current block, the motion vector is empty. If there is a motion relationship between the reference block and the current block, a motion vector is present.
  • HEVC High Efficiency Video Coding
  • FIG. 2 a motion vector diagram is shown.
  • the grid in the figure represents the result of matching block division, and the arrow in the figure represents the motion vector.
  • Motion vectors only exist when there are moving blocks in the figure. Therefore, by determining the motion vector information within a video frame, it can be determined whether the current frame is a static frame.
  • Training is then performed using a support vector machine with a Gaussian radial basis function as the kernel function to obtain the final model.
  • motion vector information for a complete image group is collected.
  • the total number of motion vectors, maximum displacement, minimum displacement, and average displacement are extracted as features.
  • the previously trained support vector machine model is then loaded to determine the status of the current image group: 1 indicates a dynamic image group, and 0 indicates a static image group.
  • the results are then encapsulated in an SEI frame in JSON format, using the format ⁇ svmResult:0/1 ⁇ .
  • the SEI frame is then added to the encoded bitstream, and the video is saved.
  • the watermark server pulls the codestream from the cloud storage server, parses the information within the codestream's SEI frames, and determines whether the current image group is dynamic or static. If it is a dynamic image group, invisible watermark frames are added every M frames; if it is a static image group, invisible watermark frames are added every N frames. For details, see Figure 3, where the frequency of adding invisible watermark frames is higher for dynamic image groups, that is, within the same frame, dynamic image groups have more invisible watermark frames. This is because when watching security videos, users focus on dynamic videos. For example, when no one is active at night, security videos are static (similar to a picture). Most users will skip this kind of video.
  • this embodiment utilizes adaptive video hidden watermark frame addition technology to hierarchically process dynamic and static videos, reducing the time and computing power required to add hidden watermark frames.
  • this embodiment When judging whether a video is dynamic or static, this embodiment directly uses the motion vector information in the encoding stage as a feature, obtains the result through support vector machine classification, and judges the dynamic or static state of the GOP without the need to extract other features, thereby reducing the amount of calculation.
  • an apparatus 400 for adding video identification information includes the following modules:
  • the above-described embodiment of the present application obtains dynamic and static classification information of image groups in a video and then divides the video information into dynamic image groups and static image groups based on the dynamic and static classification information. Then, based on the dynamic image and static image addition frequencies, identification information is added to specific frames of the dynamic image group and the static image group, respectively, to obtain a video with the identification information added. Because the preset dynamic image addition frequency and the preset static image addition frequency are less than the number of video frames, the number of added frames is reduced, thereby reducing the computing power consumption of adding identification information and the time required to add identification information. This can improve the efficiency of adding identification information and meet the high real-time requirements of the business.
  • the division module is configured to divide the image groups in the video according to the motion vector information to obtain dynamic and static classification information of the image groups in the video.
  • the apparatus 400 for adding video identification information further includes:
  • the acquisition module is configured to acquire the positional relationship between each image group and the corresponding reference image group in the video information.
  • the acquisition module is further configured to obtain motion vector information corresponding to each image group according to the corresponding positional relationship when there is a motion relationship.
  • the above-mentioned embodiment of the present application compares the positional relationship between each different image block in the image group and the image blocks in the reference image group to determine whether there is a motion relationship, and then obtains the motion vector information corresponding to each image group.
  • the motion vector information can be accurately obtained.
  • the apparatus 400 for adding video identification information further includes:
  • the classification module is configured to input the motion vector information into a dynamic and static video classification model, and classify the image group in the video according to the degree of motion represented by the motion vector information through the dynamic and static video classification model to obtain dynamic and static classification information corresponding to the image group.
  • the above embodiment of the present application uses a pre-trained static and dynamic video classification model to classify the image groups in the video into static and dynamic based on motion vector information, thereby obtaining classification results quickly and accurately.
  • the determination module is configured to determine the total number of motion vectors in the motion vector information through a dynamic and static video classification model.
  • the calculation module is configured to calculate an average value of each motion displacement value in the motion vector information.
  • the classification module is configured to use the dynamic and static classification sub-model in the dynamic and static video classification model to perform dynamic and static classification on the image group in the video based on the total number of the motion vectors, the average value of the displacement, the maximum value of the displacement in the motion vector information, and the minimum value of the displacement in the motion vector information, so as to obtain dynamic and static classification information corresponding to the image group.
  • the above-described embodiment of the present application obtains the total number of motion vectors and the average value of each motion displacement value, and then classifies the video into static and dynamic based on four features: the total number of motion vectors, the average value of the displacement, the maximum displacement value in the motion vector information, and the minimum displacement value in the motion vector information.
  • the total number of motion vectors the average value of the displacement
  • the maximum displacement value in the motion vector information the maximum displacement value in the motion vector information
  • the minimum displacement value in the motion vector information By using more comprehensive motion vector information, dynamic and static classification information corresponding to the image group can be more accurately obtained.
  • the apparatus 400 for adding video identification information further includes:
  • the acquisition module is configured to acquire historical video data.
  • the segmentation module is configured to segment the historical video data into a plurality of image groups and obtain labels of the plurality of image groups, wherein the labels are dynamic image groups or static image groups.
  • the calculation module is configured to obtain the total number of corresponding motion vectors, the maximum value of displacement, the minimum value of displacement and the average value of displacement according to the motion vector information of each of the image groups.
  • the above-mentioned embodiment of the present application obtains historical video data and annotates the dynamic image group or static image group in each image group in the historical video data, and then obtains the total number of corresponding motion vectors, the maximum displacement, the minimum displacement, and the average displacement based on the motion vector information of each image group. These are then used to train an initial support vector machine, thereby obtaining a dynamic and static classification sub-model.
  • the initial support vector machine can be trained using historical video data to obtain an accurate dynamic and static classification sub-model.
  • the apparatus 400 for adding video identification information further includes:
  • the above-mentioned embodiment of the present application adds the dynamic and static classification information to the supplementary enhancement information frame and then adds the original video information, which makes the dynamic and static classification information easier to obtain and use.
  • the various modules in the apparatus for adding video identification information provided in the embodiment of the present application can implement the various steps in the above-mentioned method for adding video identification information and achieve the corresponding effects. For the sake of brevity, they will not be repeated here.
  • the device for adding video identification information may include a processor 501 and a memory 502 storing computer program instructions.
  • the above-mentioned processor 501 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
  • CPU central processing unit
  • ASIC application specific integrated circuit
  • the memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical/tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method for adding video identification information according to any one embodiment of the present disclosure.
  • the processor 501 reads and executes computer program instructions stored in the memory 502 to implement any one of the methods for adding video identification information in the above embodiments.
  • the device for adding video identification information may further include a communication interface 503 and a bus 510. As shown in FIG5, the processor 501, the memory 502, and the communication interface 503 are connected via the bus 510 and communicate with each other.
  • embodiments of the present application may provide a computer storage medium for implementation.
  • the computer storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any of the methods for adding video identification information in the above embodiments is implemented.
  • machine-readable media examples include electronic circuits, semiconductor memory devices, ROMs, flash memories, erasable ROMs (EROMs), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc.
  • the code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Image Analysis (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

本申请提供一种视频鉴别信息添加的方法、装置、设备、存储介质及产品。该方法包括:获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的;根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组;根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。

Description

视频鉴别信息添加的方法、装置、设备、存储介质及产品
相关申请的交叉引用
本申请基于申请号202410340204.3、申请日为2024年03月25日,申请名称为“视频鉴别信息添加的方法、装置、设备、存储介质及产品”的中国专利申请提出,并要求该中国专利申请的优先权,该中国专利申请的全部内容在此以引入方式结合在本申请中。
技术领域
本申请涉及网络安全技术领域,尤其涉及一种视频鉴别信息添加的方法、装置、设备、存储介质及产品。
背景技术
随着网络技术的发展和网络基础设施的完善,直播、视频会议和家庭实时安防监控等业务也越来越完善。然而,随之而来的数据泄露的风险也日益增加,进而可能导致各种信息隐私和版权保护相关的问题。
目前的视频会在视频中嵌入鉴别信息,比如隐形水印,以此追溯来源,解决数据泄露问题。鉴别信息在版权保护、内容认证和来源追踪等方面具有重要应用价值。如在视频中添加隐形水印,不会影响用户的观看体验。但是,直播、视频会议和家庭实时安防监控这类实时视频业务的视频与传统视频相比时长更长,其数据量巨大,而嵌入隐形水印需要针对每一帧图像进行运算,消耗计算资源和时间,导致算力成本急剧增加,而且不能满足业务的实时性要求,添加隐形水印效率低。
因此,现有技术中添加鉴别信息不能满足业务实时性高的业务需求,且添加鉴别信息的效率比较低。
发明内容
本申请提供一种视频鉴别信息添加的方法、装置、设备、存储介质及产品,可以提高添加鉴别信息的效率,满足业务的实时性需求。
第一方面,本申请提供一种视频鉴别信息添加的方法,所述方法包括:
获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的;
根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组;
根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频,其中,所述预设动态图像添加频率和所述预设静态图像添加频率小于视频帧的数量。
本申请的上述实施方式通过获取到视频中图像组的动静态分类信息之后,根据动静态分类信息将视频信息分为动态图像组和静态图像组。然后按照动态图像与静态图像添加频率,分别向动态图像组以及静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。由于所述预设动态图像添加频率和所述预设静态图像添加频率小于视频帧的数量,因此减少了添加帧的数量,从而降低添加鉴别信息的算力消耗,减少添加鉴别信息时间,能够提升添加鉴别信息的效率,满足业务的高实时性需求。
在一些可能的实现方式中,在所述获取视频中图像组的动静态分类信息之前,所述方法还包括:
获取视频的视频信息,所述视频信息包括所述视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度;
根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组的动静态分类信息。
本申请的上述实施方式通过直接获取视频中图像组在编码过程中的运动矢量信息来对视频的动态情况进行分类,处理信息的过程更加快速和直接,能够加快信息的处理速度。
在一些可能的实现方式中,所述获取视频的视频信息,包括:
获取视频信息中各个图像组和对应参考图像组的位置关系;
根据所述位置关系确定各个图像组与对应参考图像组之间是否存在运动关系;
在存在运动关系的情况下,根据相应的所述位置关系,得到各个图像组对应的运动矢量信息。
本申请的上述实施方式通过将图像组中各个不同图像分块,与参考图像组中的图像分块的位置关系进行对比,从而判断是否存在运动关系,进而得到各个图像组对应的运动矢量信息。能够准确的得到运动矢量信息。
在一些可能的实现方式中,所述根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组,包括:
将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息。
本申请的上述实施方式通过使用预先训练好的动静态视频分类模型来通过运动矢量信息对视频中图像组进行动静分类。可以快速准确的得到分类结果。
在一些可能的实现方式中,所述将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息,包括:
通过动静态视频分类模型确定运动矢量信息中运动矢量的总数量;
计算所述运动矢量信息中各运动位移值的平均值;
采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息。
本申请的上述实施方式通过获取运动矢量的总数量和各运动位移值的平均值,再基于运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值这四个特征进行视频的动静态分类。采用了更全面的运动矢量信息,能够更准确的得到所述图像组对应的动静态分类信息。
在一些可能的实现方式中,在所述采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息之前,所述方法还包括:
获取历史视频数据;
将所述历史视频数据分割为多个图像组,获取所述多个图像组的标注,所述标注为动态图像组或静态图像组;
根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值;
将各个所述图像组的所述运动矢量的总数量、所述位移的最大值、所述位移的最小值和所述位移的平均值作为输入数据,根据图像组对应的标注对初始支持向量机进行训练,得到动静态分类子模型。
本申请的上述实施方式通过获取历史视频数据,以及历史视频数据中各个图像组中的动态图像组或静态图像组标注,再根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值,进而使用这些进行对初始支持向量机进行训练,从而得到动静态分类子模型。能够通过历史视频数据对初始支持向量机进行训练,得到判断准确的动静态分类子模型。
在一些可能的实现方式中,所述预设动态图像添加频率包括在每个与基准帧的间隔帧数为第一数值整数倍的帧中添加鉴别信息;
所述预设静态图像添加频率包括在每个与基准帧的间隔帧数为第二数值整数倍的帧中添加鉴别信息,所述第二数值大于所述第一数值。
本申请的上述实施方式通过再动态图像组与静态图像组的间隔整数倍的帧中添加鉴别信息。能够使得鉴别信息再视频中分布的更加均匀。
在一些可能的实现方式中,所述获取视频中图像组的动静态分类信息之前,所述方法还包括:
获取原始视频的视频信息,所述视频信息包括所述原始视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度;
根据所述运动矢量信息对所述原始视频中的图像组进行划分,得到所述原始视频中图像组的初始动静态分类信息;
将所述初始动静态分类信息加入补充增强信息帧;
将所述初始补充增强信息帧加入所述原始视频信息,得到所述视频,所述初始动静态分类信息为所述动静态信息。
本申请的上述实施方式通过将动静态分类信息加入补充增强信息帧,进而添加进行原始视频信息。能够让动静态分类信息更易被获取和使用。
第二方面,本申请提供一种视频鉴别信息添加的装置,所述装置包括:
获取模块,配置为获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的;
分类模块,配置为根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组;
添加模块,配置为根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频,其中,所述预设动态图像添加频率和所述预设静态图像添加频率小于视频帧的数量。
第三方面,本申请提供一种视频鉴别信息添加的设备,所述设备包括:处理器,以及存储有计算机程序指令的存储器;所述处理器读取并执行所述计算机程序指令,以实现如上文描述的视频鉴别信息添加的方法。
第四方面,本申请提供一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机程序指令,所述计算机程序指令被处理器执行时实现如上文描述的视频鉴别信息添加的方法。
第五方面,本申请提供一种计算机程序产品,所述计算机程序产品中的指令由电子设备的处理器执行时,使得所述电子设备执行如上文描述的视频鉴别信息添加的方法。
本申请的上述实施方式通过获取到视频中图像组的动静态分类信息之后,根据动静态分类信息将视频信息分为动态图像组和静态图像组。然后按照动态图像与静态图像添加频率,分别向动态图像组以及静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。由于所述预设动态图像添加频率和所述预设静态图像添加频率小于视频帧的数量,因此减少了添加帧的数量,从而降低添加鉴别信息的算力消耗,减少添加鉴别信息时间,能够提升添加鉴别信息的效率,满足业务的高实时性需求。
附图说明
从下面结合附图对本申请的具体实施方式的描述中可以更好地理解本申请,其中:
通过阅读以下参照附图对非限制性实施例所作的详细描述,本申请的其它特征、目的和优点将会变得更明显,其中,相同或相似的附图标记表示相同或相似的特征。
图1是本申请一个实施例提供的视频鉴别信息添加的方法的流程示意图;
图2是本申请一个实施例中运动矢量的示意图;
图3是本申请一个实施例提供的又一视频鉴别信息添加的方法的流程示意图;
图4是本申请一个实施例提供的视频鉴别信息添加的装置的结构示意图;
图5是本申请实施例提供的视频鉴别信息添加的设备的硬件结构示意图。
具体实施方式
下面将详细描述本申请的各个方面的特征和示例性实施例。在下面的详细描述中,提出了许多具体细节,以便提供对本申请的全面理解。但是,对于本领域技术人员来说很明显的是,本申请可以在不需要这些具体细节中的一些细节的情况下实施。下面对实施例的描述仅仅是为了通过示出本申请的示例来提供对本申请的更好的理解。本申请决不限于下面所提出的任何具体配置和算法,而是在不脱离本申请的精神的前提下覆盖了元素、部件和算法的任何修改、替换和改进。在附图和下面的描述中,没有示出公知的结构和技术,以便避免对本申请造成不必要的模糊。
补充增强信息(Supplemental Enhancement Information,SEI):定义在视频码流内,提供了在视频码流中添加信息的方法。
随着技术的迅猛发展和人们需要的增加,当前家庭安防视频云存规模巨大,但数据泄露和篡改的风险也日益增加,海量的视频给维护视频数据安全带来挑战。安防视频流出的事件频发,这导致用户对家庭安防安全性存在质疑。
为了解决这些问题,可以在视频中嵌入不可见信息并追溯来源,这种视频隐形水印技术在版权保护、内容认证和来源追踪等方面具有重要应用价值。并且这种不可见的视频水印,不会影响用户的观看体验。然而由于安防视频规模巨大,嵌入水印需要消耗计算资源和时间,算力成本急剧增加,这也会导致用户播放时视频卡顿,下载或分享视频时等待时间过长。
另外,安防视频的下载和回看需要满足实时性的要求,然而增加视频隐形水印耗时耗算力,目前的现有技术一般着眼于隐形水印算法,并未考虑视频播放时的实时性问题。
针对上述问题研究发现,在部分监控场景中,静态视频所占比重很大,由于静态视频往往是一个静止的画面,提供的信息量较少,所以出于对节约算力和实时性的角度,对动态视频和静态视频进行区分。添加水印的目的是为了溯源,作为信息量较低的静态视频,流传的可能性很低,因此没有必要消耗资源大量的添加水印。而且,作为视频,一般不会出现单帧图像单独流传的情况,所以,没有必要在每一帧中都添加水印。
在本申请视频鉴别信息添加的方法的一个实施例中,可以参照图1,在本实施例中,方法可以由上述的视频鉴别信息添加的装置执行,方法包括以下步骤S101至S103。
步骤S101:获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的。
图像组可以是在时间上紧密相连的多张图像,通常是指视频中一段连续的图像帧。通常在视频编码过程中,即会将视频分为多个图像组。
具体来说,执行解码操作,将视频文件转化为一系列图像组。进而获取每一个图像组对应的动静态分类信息。举例来说,在视频信息中插入自定义的数据信息,其中包含有动静态分类信息,在接收到视频后,就可以提取其中的动静态分类信息。
步骤S102:根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组。
具体来说,通过获取到的动静态分类信息,就可以对每个图像组进行动静分类。运动幅度较大的图像组被归类为动态图像组,运动幅度较小的图像组被归类为静态图像组。
步骤S103:根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。
鉴别信息可以是一个特定的标记、水印或者其他形式的识别码。
具体来说,按照预设的动态图像添加频率,在动态图像组中选择特定的帧,然后向这些选定的帧添加鉴别信息。具体的选择方法可以是按照概率值随机选择图像组中的若干帧。特别的,考虑到若视频包含有声音信息,可以获取视频的声音信息,优先选择向声音信息对应的图像添加水印。
类似地,按照预设的静态图像添加频率,在静态图像组中选择特定的帧,然后向这些选定的帧添加鉴别信息。获取添加鉴别信息后的视频帧序列,重组得到添加了鉴别信息的视频。具体来说,预设动态图像添加频率和预设静态图像添加频率都要小于视频帧的数量,使其不需要在视频的每一帧中都进行鉴别信息的添加。
本申请的上述实施方式通过获取到视频中图像组的动静态分类信息之后,根据动静态分类信息将视频信息分为动态图像组和静态图像组。然后按照动态图像与静态图像添加频率,分别向动态图像组以及静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。由于预设动态图像添加频率和预设静态图像添加频率小于视频帧的数量,因此减少了添加帧的数量,从而降低添加鉴别信息的算力消耗,减少添加鉴别信息时间,能够提升添加鉴别信息的效率,满足业务的高实时性需求。
在一些实施方式中,在所述获取视频中图像组的动静态分类信息之前,所述方法还包括:
获取视频的视频信息,所述视频信息包括所述视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度。
运动矢量信息可以是在视频编码中,用来表示相邻两帧图像之间的运动关系的向量数据。在视频编码中,为了减小视频数据的冗余性,通常采用基于运动补偿的压缩方法,即对于当前帧图像,使用前一帧或多帧图像进行预测,然后对预测误差进行编码和传输。为了准确地描述当前帧图像与参考帧之间的差异,需要计算当前帧图像相对于参考帧的运动矢量,并将其编码和传输。
具体来说,视频信息会经过编码,在过程中,会对每个图像帧进行运动估计操作,通过比较当前帧与相邻帧之间的差异,得到图像组的运动矢量信息。常用的方法有光流法、整体运动估计等。
根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组的动静态分类信息。
具体来说,得到各帧图像的运动信息之后,根据应用需求和实际情况,设置一个适当的运动矢量阈值。该阈值可根据运动矢量的大小、方向或其他特征来确定,用于区分“动”和“静”图像组。
举例来说,可以对每个图像组的运动矢量进行阈值判断,将运动矢量大于阈值的图像组划分为动态图像组,而小于阈值的图像组划分为静态图像组。还可以对图像组中相邻帧之间的运动矢量进行差分计算,如果差分值大于阈值,则将该图像组划分为动态图像组,反之则划分为静态图像组。
本申请的上述实施方式通过直接获取视频中图像组在编码过程中的运动矢量信息来对视频的动态情况进行分类,处理信息的过程更加快速和直接,能够加快信息的处理速度。
在一些实施方式中,所述获取视频的视频信息,包括:
获取视频信息中各个图像组和对应参考图像组的位置关系。
具体来说,对于每个图像组,选择一个参考图像组。参考图像组可以是前一帧、多帧之前的图像组,也可以是当前图像组中的某一帧作为参考。根据图像组与对应参考图像组画面的不同,确定它们的位置关系。
根据所述位置关系确定各个图像组与对应参考图像组之间是否存在运动关系。
具体来说,基于位置关系,可以判断图像组与对应参考图像组之间的位置差异幅度是否超过了预设阈值,在位置差异幅度超过了预设阈值的情况下,判断其与参考图像组之间存在运动关系。
在存在运动关系的情况下,根据相应的所述位置关系,得到各个图像组对应的运动矢量信息。
具体来说,可以对每个图像组进行特征点提取,然后将每个图像组的特征点和对应参考图像组的特征点进行匹配,例如可以使用基于描述子的匹配算法。然后利用特征点的匹配信息,采用运动估计算法来计算图像组与对应参考图像组之间的运动变换,运动变换包括但不限于平移、旋转等。根据运动估计得到的运动变换参数,可以计算出每个图像组对应的运动矢量信息,例如位移矢量的大小和方向、旋转角度等。
本申请的上述实施方式通过将图像组中各个不同图像分块,与参考图像组中的图像分块的位置关系进行对比,从而判断是否存在运动关系,进而得到各个图像组对应的运动矢量信息。能够准确的得到运动矢量信息。
在一些实施方式中,所述根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组,包括:
将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息。
动静态视频分类模型可以是一种机器学习算法或深度学习算法,用于将视频中的图像组按照其动态或静态特性进行分类。
具体来说,对于每个图像组,可以利用运动矢量信息计算出相应的特征向量。这些特征向量可以包括但不限于运动矢量的平均值、方差等。
然后构建一个适合的动静态视频分类模型,可以使用机器学习算法(如支持向量机、决策树、随机森林),再比如深度学习算法(如卷积神经网络)来训练模型。使用标注好的训练数据集,利用特征向量作为输入,动静态分类标签作为输出,对分类模型进行训练。训练过程中可以采用交叉验证、正则化等技术来提高模型的性能。
将图像组的特征向量输入已经训练好的分类模型,根据模型的输出结果来判断图像组的动静态分类信息。具体的分类方法可以是基于阈值、概率或多类别分类。
本申请的上述实施方式通过使用预先训练好的动静态视频分类模型来通过运动矢量信息对视频中图像组进行动静分类。可以快速准确的得到分类结果。
在一些实施方式中,所述将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息,包括:
通过动静态视频分类模型确定运动矢量信息中运动矢量的总数量。
具体来说,可以是对每个图像组中的运动矢量个数进行累加,得到运动矢量的总数量。
计算所述运动矢量信息中各运动位移值的平均值。
具体来说,对于每个图像组,将其运动矢量信息提取出来,包括运动矢量的大小(即位移值)和方向。对于每个图像组的运动矢量信息,将其所有位移值进行累加,然后除以位移值的总数量,即可得到运动位移值的平均值。
采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息。
具体来说,对于每个图像组,利用运动矢量信息计算出所需的特征,包括但不限于运动矢量的总数量、位移的平均值、位移的最大值和最小值。这些特征可以用于描述图像组的动态特性。将训练好的动静态视频分类模型应用于新的图像组。将图像组的特征输入已训练好的分类模型,模型将输出对应的动静态分类信息。
本申请的上述实施方式通过获取运动矢量的总数量和各运动位移值的平均值,再基于运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值这四个特征进行视频的动静态分类。采用了更全面的运动矢量信息,能够更准确的得到所述图像组对应的动静态分类信息。
在一些实施方式中,在所述采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息之前,所述方法还包括:
获取历史视频数据。
将所述历史视频数据分割为多个图像组,获取所述多个图像组的标注,所述标注为动态图像组或静态图像组。
具体来说,首先从历史视频数据源中获取所需的视频数据,并将其转换为图像序列。对于每个视频图像序列,将其分割为多个图像组,每个图像组包含一定数量的连续图像帧,通常情况下每个图像组包含的图像帧数量是固定的。针对分割得到的每个的图像组,需要获取相应的动态或静态标注,其可以通过人工标注来完成。
根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值。
将各个所述图像组的所述运动矢量的总数量、所述位移的最大值、所述位移的最小值和所述位移的平均值作为输入数据,根据图像组对应的标注对初始支持向量机进行训练,得到动静态分类子模型。
具体来说,将每个图像组的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值以及对应的标注组成训练数据集。将这些特征作为输入数据,将标注作为输出数据。根据训练数据集,初始化一个支持向量机分类器。使用训练数据集对初始化的支持向量机进行训练。训练过程中,支持向量机会根据输入数据的特征和标注信息调整其模型参数,以达到更好的分类效果。训练完成后,得到了一个训练好的支持向量机模型,该模型可以将输入数据(运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值)与相应的标注进行分类。这个训练好的支持向量机模型可以看作是动静态分类子模型。
本申请的上述实施方式通过获取历史视频数据,以及历史视频数据中各个图像组中的动态图像组或静态图像组标注,再根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值,进而使用这些进行对初始支持向量机进行训练,从而得到动静态分类子模型。能够通过历史视频数据对初始支持向量机进行训练,得到判断准确的动静态分类子模型。
在一些实施方式中,所述预设动态图像添加频率包括在每个与基准帧的间隔帧数为第一数值整数倍的帧中添加鉴别信息。
具体来说,首先选择一个基准帧作为参考帧。这通常是视频序列中的某一帧,通常选择第一帧或者其他关键帧作为基准帧。对于每一帧图像,将其与基准帧之间的帧数进行计算,得到帧之间的间隔帧数。根据预设的动态图像添加频率,确定在哪些帧中需要添加鉴别信息。如果间隔帧数是第一数值的整数倍,就在该帧中添加鉴别信息。
所述预设静态图像添加频率包括在每个与基准帧的间隔帧数为第二数值整数倍的帧中添加鉴别信息,所述第二数值大于所述第一数值。
本步骤的具体执行过程与上述类似,在此不再赘述,而具体来说,所述第二数值大于所述第一数值,也就是静态图像对应的第二数值更大,添加鉴别信息的频率也就相应的更大,则动态图像添加频率小于静态图像添加频率。
本申请的上述实施方式通过再动态图像组与静态图像组的间隔整数倍的帧中添加鉴别信息。能够使得鉴别信息再视频中分布的更加均匀。
在一些实施方式中,所述获取视频中图像组的动静态分类信息之前,所述方法还包括:
获取原始视频的视频信息,所述视频信息包括所述原始视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度。
在具体实现中,对于每个图像组提取图像间的运动矢量信息。
根据所述运动矢量信息对所述原始视频中的图像组进行划分,得到所述原始视频中图像组的初始动静态分类信息。
在具体实现中,可以根据设定的运动矢量阈值,对原始视频中的图像组进行划分。一般来说,如果某个图像组中的运动矢量超过了设定的阈值,则认为该图像组属于动态图像组,否则认为属于静态图像组。
将所述初始动静态分类信息加入补充增强信息帧。
在具体实现中,可以遍历原始视频的每个图像组,在对应的补充增强信息帧中嵌入该图像组的初始动静态分类信息。
将所述初始补充增强信息帧加入所述原始视频信息,得到所述视频,所述初始动静态分类信息为所述动静态信息。
在具体实现中,将嵌有分类信息的补充增强信息帧添加到原始视频中的相应位置。这样,经过补充增强信息帧处理的原始视频信息合并,生成包含动静态分类信息的最终视频。
本申请的上述实施方式通过将动静态分类信息加入补充增强信息帧,进而添加进行原始视频信息。能够让动静态分类信息更易被获取和使用。
作为本申请的另一实现方式,采集视频之后通过常规的编码器,例如高效视频编码(High Efficiency Video Coding,HEVC)进行编码,在编码过程中,对运动估计得到的运动矢量信息进行存储。运动估计是编码时当前块在参考图像中搜索找到匹配块的过程,利用运动矢量表示参考块与当前块的位置关系。如果参考块与当前块无运动关系,则运动矢量为空,如果参考块与当前块有运动关系,则存在运动矢量。
如图2所示,图2为运动矢量示意图,图中的格子为匹配块划分的结果,图中箭头为运动矢量,只有图中有运动块时,才会存在运动矢量。因此,可以通过判断视频帧内的运动矢量信息,来确认当前帧是否为静态帧。
为了进行动静态的分类,需要进行机器学习模型的训练,在前期训练阶段,利用家庭安防内的视频数据库,将视频数据分割成各个图像组(Group of Pictures,GOP),人工对数据进行标注,标注1为动态图像组,0为静态图像组。在训练之前对n个图像组进行特征提取,以此组成特征向量[a,b,c,d]*n,其中,“a,b,c,d”表示“运动矢量的总数量、位移的最大值、位移的最小值、位移的平均值”。根据实际情况,还可以添加更多的特征,本实施例考虑到有些视频只有几个当前块是有运动矢量,还有一些轻微的运动,当前块运动很小,但是人肉眼看还是静态的,这种情况其实都是可以算做静态视频,所以选取了运动矢量相关的特征。之后利用以高斯径向基函数为核函数支持向量机,进行训练,得到最终的模型。
这样,在视频编码过程中,对完整的一个图像组运动矢量信息进行收集,提取运动矢量的总数量、位移的最大值、位移的最小值、位移的平均值作为特征,加载前期训练得到的支持向量机模型,对当前图像组的状态进行判断,1为动态图像组,0为静态图像组。得到结果之后,将结果以JSON形式封装进SEI帧,格式具体为{svmResult:0/1}。将SEI帧添加进编码后的码流内,将视频进行保存。
在用户选择播放、下载或分享视频的情况下,水印服务器拉取云存储服务器内的码流,解析码流SEI帧内的信息,判断当前图像组为动态或静态,若是动态图像组,则间隔M帧添加隐形水印帧;若是静态图像组,则间隔N帧添加隐形水印帧。具体可参考图3,其中,动态图像组的添加频率要更高,也就是相同帧的图像中,动态图像组的添加隐形水印帧要更多。这是因为在通过安防视频回看的时候,用户关注点是在动态视频上面的,比如晚上没有人活动的时候,安防视频都是静态的(类似一张图片),这种视频大部分用户会跳过,在安防场景,这些静态视频是大片的,占比很高,因为就是一个画面,如果添加过多水印帧,浪费算力,耗时也长,而且实际应用不大,所以选择自适应视频隐形水印帧添加技术,动态的事件视频,多添加一些水印帧,在算力允许的情况下,视频帧全部加上水印也可以,静态视频少添加一些水印帧,甚至不添加也可以,结合实际情况可以自由选择。
由于安防视频规模巨大,若对每一帧的视频嵌入隐形水印需要消耗大量计算资源,算力成本急剧增加。因此本实施例利用自适应视频隐形水印帧添加技术,对动态和静态视频分级处理,减少隐形水印帧添加的耗时,减少算力成本。
在对视频动态或静态判断时,本实施例直接利用的是编码阶段的运动矢量的信息作为特征,通过支持向量机分类获得结果,判GOP的动静态,不需要额外提取其他特征,减少了计算量。
本实施例将算法结果以JSON形式封装进SEI帧,添加进码流内,解码端只需要解析SEI帧信息,即可获得当前GOP的动静态结果,往往对视频的处理都是在解码端,本实施例将对视频的处理分析放在了编码端,减轻了解码端计算量。
基于上述实施例提供的视频鉴别信息添加的方法,相应地,本申请还提供了视频鉴别信息添加的装置的具体实现方式。请参见以下实施例。
首先参见图4,本申请实施例提供的视频鉴别信息添加的装置400,包括以下模块:
获取模块401,配置为获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的。
分类模块402,配置为根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组。
添加模块403,配置为根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。
本申请的上述实施方式通过获取到视频中图像组的动静态分类信息之后,根据动静态分类信息将视频信息分为动态图像组和静态图像组。然后按照动态图像与静态图像添加频率,分别向动态图像组以及静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。由于所述预设动态图像添加频率和所述预设静态图像添加频率小于视频帧的数量,因此减少了添加帧的数量,从而降低添加鉴别信息的算力消耗,减少添加鉴别信息时间,能够提升添加鉴别信息的效率,满足业务的高实时性需求。
作为本申请的一种实现方式,视频鉴别信息添加的装置400,还包括:
获取模块,配置为获取视频的视频信息,所述视频信息包括所述视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度;
划分模块,配置为根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组的动静态分类信息。
本申请的上述实施方式通过直接获取视频中图像组在编码过程中的运动矢量信息来对视频的动态情况进行分类,处理信息的过程更加快速和直接,能够加快信息的处理速度。
作为本申请的一种实现方式,视频鉴别信息添加的装置400,还包括:
获取模块,配置为获取视频信息中各个图像组和对应参考图像组的位置关系。
确定模块,配置为根据所述位置关系确定各个图像组与对应参考图像组之间是否存在运动关系。
获取模块,还配置为在存在运动关系的情况下,根据相应的所述位置关系,得到各个图像组对应的运动矢量信息。
本申请的上述实施方式通过将图像组中各个不同图像分块,与参考图像组中的图像分块的位置关系进行对比,从而判断是否存在运动关系,进而得到各个图像组对应的运动矢量信息。能够准确的得到运动矢量信息。
作为本申请的一种实现方式,视频鉴别信息添加的装置400,还包括:
分类模块,配置为将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息。
本申请的上述实施方式通过使用预先训练好的动静态视频分类模型来通过运动矢量信息对视频中图像组进行动静分类。可以快速准确的得到分类结果。
作为本申请的一种实现方式,视频鉴别信息添加的装置400,还包括:
确定模块,配置为通过动静态视频分类模型确定运动矢量信息中运动矢量的总数量。
计算模块,配置为计算所述运动矢量信息中各运动位移值的平均值。
分类模块,配置为采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息。
本申请的上述实施方式通过获取运动矢量的总数量和各运动位移值的平均值,再基于运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值这四个特征进行视频的动静态分类。采用了更全面的运动矢量信息,能够更准确的得到所述图像组对应的动静态分类信息。
作为本申请的一种实现方式,视频鉴别信息添加的装置400,还包括:
获取模块,配置为获取历史视频数据。
分割模块,配置为将所述历史视频数据分割为多个图像组,获取所述多个图像组的标注,所述标注为动态图像组或静态图像组。
计算模块,配置为根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值。
训练模块,配置为将各个所述图像组的所述运动矢量的总数量、所述位移的最大值、所述位移的最小值和所述位移的平均值作为输入数据,根据图像组对应的标注对初始支持向量机进行训练,得到动静态分类子模型。
本申请的上述实施方式通过获取历史视频数据,以及历史视频数据中各个图像组中的动态图像组或静态图像组标注,再根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值,进而使用这些进行对初始支持向量机进行训练,从而得到动静态分类子模型。能够通过历史视频数据对初始支持向量机进行训练,得到判断准确的动静态分类子模型。
作为本申请的一种实现方式,视频鉴别信息添加的装置400,还包括:
获取模块,配置为获取原始视频的视频信息,所述视频信息包括所述原始视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度。
划分模块,配置为根据所述运动矢量信息对所述原始视频中的图像组进行划分,得到所述原始视频中图像组的初始动静态分类信息。
加入模块,配置为将所述初始动静态分类信息加入补充增强信息帧。
加入模块,还配置为将所述初始补充增强信息帧加入所述原始视频信息,得到所述视频,所述初始动静态分类信息为所述动静态信息。
本申请的上述实施方式通过将动静态分类信息加入补充增强信息帧,进而添加进行原始视频信息。能够让动静态分类信息更易被获取和使用。
本申请实施例提供的视频鉴别信息添加的装置中的各个模块可以实现上述视频鉴别信息添加的方法中的各个步骤,并达到相应效果,为简洁描述,在此不再赘述。
图5示出了本申请实施例提供的视频鉴别信息添加的硬件的结构示意图。
在视频鉴别信息添加的设备可以包括处理器501以及存储有计算机程序指令的存储器502。
具体地,上述处理器501可以包括中央处理器(CPU),或者特定集成电路(Application Specific Integrated Circuit,ASIC),或者可以被配置成实施本申请实施例的一个或多个集成电路。
存储器502可以包括配置为数据或指令的大容量存储器。举例来说而非限制,存储器502可包括硬盘驱动器(Hard Disk Drive,HDD)、软盘驱动器、闪存、光盘、磁光盘、磁带或通用串行总线(Universal Serial Bus,USB)驱动器或者两个或更多个以上这些的组合。在合适的情况下,存储器502可包括可移除或不可移除(或固定)的介质。在合适的情况下,存储器502可在综合网关容灾设备的内部或外部。在特定实施例中,存储器502是非易失性固态存储器。
存储器可包括只读存储器(ROM),随机存取存储器(RAM),磁盘存储介质设备,光存储介质设备,闪存设备,电气、光学或其他物理/有形的存储器存储设备。因此,通常,存储器包括一个或多个编码有包括计算机可执行指令的软件的有形(非暂态)计算机可读存储介质(例如,存储器设备),并且当该软件被执行(例如,由一个或多个处理器)时,其可操作来执行参考根据本公开的任意一个实施例的视频鉴别信息添加的方法所描述的操作。
处理器501通过读取并执行存储器502中存储的计算机程序指令,以实现上述实施例中的任意一种视频鉴别信息添加的方法。
在一个示例中,视频鉴别信息添加的设备还可包括通信接口503和总线510。其中,如图5所示,处理器501、存储器502、通信接口503通过总线510连接并完成相互间的通信。
通信接口503,主要配置为实现本申请实施例中各模块、装置、单元和/或设备之间的通信。
总线510包括硬件、软件或两者,将在线数据流量计费设备的部件彼此耦接在一起。举例来说而非限制,总线可包括加速图形端口(AGP)或其他图形总线、增强工业标准架构(EISA)总线、前端总线(FSB)、超传输(HT)互连、工业标准架构(ISA)总线、无限带宽互连、低引脚数(LPC)总线、存储器总线、微信道架构(MCA)总线、外围组件互连(PCI)总线、PCI-Express(PCI-X)总线、串行高级技术附件(SATA)总线、视频电子标准协会局部(VLB)总线或其他合适的总线或者两个或更多个以上这些的组合。在合适的情况下,总线510可包括一个或多个总线。尽管本申请实施例描述和示出了特定的总线,但本申请考虑任何合适的总线或互连。
另外,结合上述实施例中的视频鉴别信息添加的方法,本申请实施例可提供一种计算机存储介质来实现。该计算机存储介质上存储有计算机程序指令;该计算机程序指令被处理器执行时实现上述实施例中的任意一种视频鉴别信息添加的方法。
但是,需要明确,本申请并不局限于上文所描述并在图中示出的特定配置和处理。并且,为了简明起见,这里省略对已知方法技术的详细描述。在上述实施例中,描述和示出了若干具体的步骤作为示例。但是,本申请的方法过程并不限于所描述和示出的具体步骤,本领域的技术人员可以在领会本申请的精神之后,作出各种改变、修改和添加,或者改变步骤之间的顺序。
以上所述的结构框图中所示的功能块可以实现为硬件、软件、固件或者它们的组合。当以硬件方式实现时,其可以例如是电子电路、专用集成电路(ASIC)、适当的固件、插件、功能卡等等。当以软件方式实现时,本申请的元素是被用于执行所需任务的程序或者代码段。程序或者代码段可以存储在机器可读介质中,或者通过载波中携带的数据信号在传输介质或者通信链路上传送。“机器可读介质”可以包括能够存储或传输信息的任何介质。机器可读介质的例子包括电子电路、半导体存储器设备、ROM、闪存、可擦除ROM(EROM)、软盘、CD-ROM、光盘、硬盘、光纤介质、射频(RF)链路,等等。代码段可以经由诸如因特网、内联网等的计算机网络被下载。
本申请可以以其他的具体形式实现,而不脱离其精神和本质特征。例如,特定实施例中所描述的算法可以被修改,而系统体系结构并不脱离本申请的基本精神。因此,当前的实施例在所有方面都被看作是示例性的而非限定性的,本申请的范围由所附权利要求而非上述描述定义,并且,落入权利要求的含义和等同物的范围内的全部改变从而都被包括在本申请的范围之中。
本领域技术人员应能理解,上述实施例均是示例性而非限制性的。在不同实施例中出现的不同技术特征可以进行组合,以取得有益效果。本领域技术人员在研究附图、说明书及权利要求书的基础上,应能理解并实现所揭示的实施例的其他变化的实施例。在权利要求书中,术语“包括”并不排除其他装置或步骤;不定冠词“一个”不排除多个;术语“第一”、“第二”用于标示名称而非用于表示任何特定的顺序。权利要求中的任何附图标记均不应被理解为对保护范围的限制。权利要求中出现的多个部分的功能可以由一个单独的硬件或软件模块来实现。某些技术特征出现在不同的从属权利要求中并不意味着不能将这些技术特征进行组合以取得有益效果。

Claims (12)

  1. 一种视频鉴别信息添加的方法,所述方法包括:
    获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的;
    根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组;
    根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。
  2. 根据权利要求1所述的视频鉴别信息添加的方法,其中,在所述获取视频中图像组的动静态分类信息之前,所述方法还包括:
    获取视频的视频信息,所述视频信息包括所述视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度;
    根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组的动静态分类信息。
  3. 根据权利要求2所述的视频鉴别信息添加的方法,其中,所述获取视频的视频信息,包括:
    获取视频信息中各个图像组和对应参考图像组的位置关系;
    根据所述位置关系确定各个图像组与对应参考图像组之间是否存在运动关系;
    在存在运动关系的情况下,根据相应的所述位置关系,得到各个图像组对应的运动矢量信息。
  4. 根据权利要求2所述的视频鉴别信息添加的方法,其中,所述根据所述运动矢量信息对所述视频中的图像组进行划分,得到所述视频中图像组,包括:
    将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息。
  5. 根据权利要求4所述的视频鉴别信息添加的方法,其中,所述将所述运动矢量信息输入动静态视频分类模型,通过所述动静态视频分类模型按照所述运动矢量信息表征的运动程度对所述视频中图像组进行分类,得到所述图像组对应的动静态分类信息,包括:
    通过动静态视频分类模型确定运动矢量信息中运动矢量的总数量;
    计算所述运动矢量信息中各运动位移值的平均值;
    采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息。
  6. 根据权利要求5所述的视频鉴别信息添加的方法,其中,在所述采用所述动静态视频分类模型中动静态分类子模型,基于所述运动矢量的总数量、所述位移的平均值、所述运动矢量信息中位移的最大值和所述运动矢量信息中位移的最小值对所述视频中的图像组进行动静态分类,得到所述图像组对应的动静态分类信息之前,所述方法还包括:
    获取历史视频数据;
    将所述历史视频数据分割为多个图像组,获取所述多个图像组的标注,所述标注为动态图像组或静态图像组;
    根据各个所述图像组的运动矢量信息,得到对应的运动矢量的总数量、位移的最大值、位移的最小值和位移的平均值;
    将各个所述图像组的所述运动矢量的总数量、所述位移的最大值、所述位移的最小值和所述位移的平均值作为输入数据,根据图像组对应的标注对初始支持向量机进行训练,得到动静态分类子模型。
  7. 根据权利要求1至6任一项所述的视频鉴别信息添加的方法,其中,所述预设动态图像添加频率包括在每个与基准帧的间隔帧数为第一数值整数倍的帧中添加鉴别信息;
    所述预设静态图像添加频率包括在每个与基准帧的间隔帧数为第二数值整数倍的帧中添加鉴别信息,所述第二数值大于所述第一数值。
  8. 根据权利要求2所述的视频鉴别信息添加的方法,其中,所述获取视频中图像组的动静态分类信息之前,所述方法还包括:
    获取原始视频的视频信息,所述视频信息包括原始视频中图像组在编码过程中产生的运动矢量信息,所述运动矢量信息表征图像组的运动程度;
    根据所述运动矢量信息对所述原始视频中的图像组进行划分,得到所述原始视频中图像组的初始动静态分类信息;
    将所述初始动静态分类信息加入补充增强信息帧;
    将所述初始补充增强信息帧加入原始视频信息,得到所述视频,所述初始动静态分类信息为动静态信息。
  9. 一种视频鉴别信息添加的装置,所述装置包括:
    获取模块,配置为获取视频中图像组的动静态分类信息,所述动静态分类信息是按照视频中图像组运动的程度划分得到的;
    分类模块,配置为根据所述动静态分类信息,将所述视频中的图像组分为动态图像组和静态图像组;
    添加模块,配置为根据预设动态图像添加频率,向所述动态图像组的特定帧添加鉴别信息;根据预设静态图像添加频率,向所述静态图像组的特定帧添加鉴别信息,得到添加鉴别信息的视频。
  10. 一种视频鉴别信息添加的设备,所述设备包括:处理器,以及存储有计算机程序指令的存储器;所述处理器读取并执行所述计算机程序指令,以实现如权利要求1-8任意一项所述的视频鉴别信息添加的方法。
  11. 一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机程序指令,所述计算机程序指令被处理器执行时实现如权利要求1-8任意一项所述的视频鉴别信息添加的方法。
  12. 一种计算机程序产品,所述计算机程序产品中的指令由电子设备的处理器执行时,使得所述电子设备执行如权利要求1-8任意一项所述的视频鉴别信息添加的方法。
PCT/CN2025/084595 2024-03-25 2025-03-25 视频鉴别信息添加的方法、装置、设备、存储介质及产品 Pending WO2025201291A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202410340204.3A CN118828011A (zh) 2024-03-25 2024-03-25 视频鉴别信息添加的方法、装置、设备、存储介质及产品
CN202410340204.3 2024-03-25

Publications (2)

Publication Number Publication Date
WO2025201291A1 true WO2025201291A1 (zh) 2025-10-02
WO2025201291A9 WO2025201291A9 (zh) 2025-11-20

Family

ID=93073393

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2025/084595 Pending WO2025201291A1 (zh) 2024-03-25 2025-03-25 视频鉴别信息添加的方法、装置、设备、存储介质及产品

Country Status (2)

Country Link
CN (1) CN118828011A (zh)
WO (1) WO2025201291A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118828011A (zh) * 2024-03-25 2024-10-22 中移(杭州)信息技术有限公司 视频鉴别信息添加的方法、装置、设备、存储介质及产品

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2348071A (en) * 1999-03-18 2000-09-20 British Broadcasting Corp Watermarking
KR20010106815A (ko) * 2000-05-23 2001-12-07 서평원 정지 영상 및 동화상 워터마킹 방법 및 장치
CN112954348A (zh) * 2021-02-20 2021-06-11 浙江商汤科技开发有限公司 视频编码方法及装置、电子设备和存储介质
CN115278314A (zh) * 2022-07-08 2022-11-01 南京大学 一种多值数字视频水印嵌入及盲提取方法
CN118828011A (zh) * 2024-03-25 2024-10-22 中移(杭州)信息技术有限公司 视频鉴别信息添加的方法、装置、设备、存储介质及产品

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2348071A (en) * 1999-03-18 2000-09-20 British Broadcasting Corp Watermarking
KR20010106815A (ko) * 2000-05-23 2001-12-07 서평원 정지 영상 및 동화상 워터마킹 방법 및 장치
CN112954348A (zh) * 2021-02-20 2021-06-11 浙江商汤科技开发有限公司 视频编码方法及装置、电子设备和存储介质
CN115278314A (zh) * 2022-07-08 2022-11-01 南京大学 一种多值数字视频水印嵌入及盲提取方法
CN118828011A (zh) * 2024-03-25 2024-10-22 中移(杭州)信息技术有限公司 视频鉴别信息添加的方法、装置、设备、存储介质及产品

Also Published As

Publication number Publication date
CN118828011A (zh) 2024-10-22
WO2025201291A9 (zh) 2025-11-20

Similar Documents

Publication Publication Date Title
CN111651636B (zh) 视频相似片段搜索方法及装置
US11847816B2 (en) Resource optimization based on video frame analysis
US9646358B2 (en) Methods for scene based video watermarking and devices thereof
US9420299B2 (en) Method for processing an image
CN112270710A (zh) 位姿确定方法、位姿确定装置、存储介质与电子设备
CN110516100A (zh) 一种图像相似度的计算方法、系统、存储介质及电子设备
CN106126617A (zh) 一种视频检测方法及服务器
Li et al. Cross-level parallel network for crowd counting
CN113704506B (zh) 一种媒体内容排重方法和相关装置
CN106503112B (zh) 视频检索方法和装置
JP7539998B2 (ja) ズームアグノスティックウォーターマーク抽出
US9305603B2 (en) Method and apparatus for indexing a video stream
CN107220652A (zh) 用于处理图片的方法和装置
CN103649955A (zh) 用于视觉搜索的图像拓扑编码
WO2025201291A1 (zh) 视频鉴别信息添加的方法、装置、设备、存储介质及产品
CN110414335A (zh) 视频识别方法、装置及计算机可读存储介质
CN115914765A (zh) 一种画质调节方法、装置、设备及介质
CN117478886A (zh) 多媒体数据编码方法、装置、电子设备及存储介质
CN113627314B (zh) 人脸图像模糊检测方法、装置、存储介质与电子设备
Guo et al. Toward scalable image feature compression: a content-adaptive and diffusion-based approach
Bozkurt et al. Detection and localization of frame duplication using binary image template
CN116320429B (zh) 视频编码方法、装置、计算机设备及计算机可读存储介质
Sebai et al. End-to-end variable-rate learning-based depth compression guided by deep correlation features
CN115115968B (zh) 视频质量评价方法、装置及计算机可读存储介质
CN120125851B (zh) 基于视觉与语义特征的角色版权侵权检测方法、系统

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: 25776343

Country of ref document: EP

Kind code of ref document: A1