Product
Patsnap Eureka
For R&D, Patsnap Eureka makes reading and utilizing patents & technical documents easy.
Patsnap Eureka AIR
Designed for self-driven R&D workflows. Generate viable solutions, solve complex R&D challenges, empower your innovation with AI.
Patsnap Eureka Materials
Designed for material experts only. Revolutionize your material R&D, from search, analyze, to developing new materials.
Feature
TechResearch
Generate reliable direction feasibility study reports for your R&D in just a few steps.
TechSeek
Discover and master advanced knowledge NOW. Basics, ideas, possibilities, all at once.
TechMind
As an expert in R&D Theories, TechMind can generates customized viable solutions instantly.
TechRisk
Analyze your overall solution with one click, know your potential R&D risks in advance.
TechMonitor
Get weekly tech updates, stay abreast of the latest tech innovations and key insights.
Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
35 results about "Interframe correlation" patented technology
Efficacy Topic
Property
Owner
Technical Advancement
Application Domain
Technology Topic
Technology Field Word
Patent Country/Region
Patent Type
Patent Status
Application Year
Inventor
Moving object detection method based on depth learning
ActiveCN107123131AImprove accuracyAvoid missing detectionImage enhancementImage analysisNetwork modelInterframe correlation
The invention discloses a moving object detection method based on depth learning, which belongs to the technical field of video image processing. The method comprises steps: a constructed depth neural network model is trained to obtain an object neural network model with an ability of identifying a meaningful moving object; and the object neural network model is used for detecting a moving object from the currently-acquired scene image. The depth learning technique and the inter-frame correlation information are used, the method consumes small memory space, and the moving object is detected accurately with a little computational cost.
Owner:ANHUI TSINGLINK INFORMATION TECH
Target identification method based on quality evaluation
ActiveCN108765394ASolve the problem of object recognitionValid descriptionImage enhancementImage analysisImaging qualityGoal recognition
The invention provides a target identification method based on quality evaluation. The target identification method based on quality evaluation includes the steps: constructing a target identificationmodel which includes a quality evaluation network, a feature extraction network, and a feature aggregation network, wherein the target identification model is used for extracting the target feature from a video so as to characterize the overall structural information and local information of the target; training the target identification model, and adjusting the parameters of the quality evaluation network and the feature extraction network during the training process so as to enable the target identification model to output the target feature according with the preset demand; and performingtarget identification on the video through the trained target identification model. Therefore, the target identification method based on quality evaluation solves the target identification problem caused by changeable appearance and irregular image quality in a video sequence, and adds the interframe correlation information in quality evaluation so as to obtain more effective target information toenable characterization of the target to be more accurate, thus improving the identification accuracy.
Owner:SHANGHAI JIAO TONG UNIV
Subtitle detection method and device as well as background recovery method and device
InactiveCN101835011ALess detectableEffective reflectionTelevision system detailsColor television detailsPattern recognitionRecovery method
The invention relates to a video detection technology and discloses subtitle detection method and device as well as background recovery method and device, wherein the subtitle detection method comprises the following steps of: acquiring the probability of a pixel region in a video frame belonging to a smooth background; determining the probability of the pixel region belonging to a still texture by using an interframe correlation coefficient; acquiring the probability of the pixel region belonging to a dynamic texture; and determining whether the pixel region belongs to a roll subtitle regionor not according to the probability belonging to the smooth background, the probability belonging to the still texture and the probability belonging to the dynamic texture of the pixel region. By using the technical scheme provided by the embodiment of the invention, interframe information can be used for detecting subtitles of a video.
Owner:HUAWEI TECH CO LTD +1
A radar data compression method based on reference frame misalignment prediction
InactiveCN108989817ASmall amount of calculationPromote recoveryDigital video signal modificationPeak valueData reconstruction
The invention discloses a radar data compression method based on reference frame misalignment prediction. At first, that optimal reference frame with the strong data correlation among the original frames is determined, a prediction frame is constructed by performing intra-frame positional misalignment on the optimal reference frame, the difference data between the current frame and its predicted frame is compressed and decompressed by the block adaptive quantization algorithm to obtain the recovered difference data, and then the position information of the misalignment coding of the referenceframe data is transmitted to the decoding end through Huffman lossless coding, and then the recovered difference data and the predicted frame are added to obtain the reconstructed data. The inventionadopts inter-frame correlation to determine frame data reference frame, interframe misalignment prediction, block adaptive quantization algorithm, Huffman coding, frame data difference and reconstruction technology, under the premise of guaranteeing a certain amount of computation, achieves a higher peak signal-to-noise ratio and data reconstruction degree, has the characteristic of strong interframe prediction correlation, small difference process quantization error, high reconstruction data recovery degree and so on.
Owner:XI AN JIAOTONG UNIV
Low illumination smog video detection method based on image correlation
InactiveCN106778488AQuick extractionImprove accuracyCharacter and pattern recognitionPhase correlationGlobal illumination
The invention provides a low illumination smog video detection method based on image correlation; the method comprises the following steps: compressing a recorded video image and converting into a gray scale image, and carrying out four-frame difference operation on the obtained gray scale image, thus extracting a motion object in the video image; selecting a property threshold to convert the interframe difference result into a binary image; carrying out median filtering on the obtained binary image, thus initially reducing image noises; calculating interframe correlation of various connecting zones of the binary image in order, removing non-smog motion areas with small image correlation, and obtaining an accurate smog area image. The method can directly observe the smog developing conditions, thus removing temperature, humidity, air pressure and smell factor influences, improving the smog detection accuracy, and keeping the false alarm rate within 10% under conservative estimation; in addition, the method can replay an extracted smog area developing process, thus accurately positioning the smog source position.
Owner:CIVIL AVIATION UNIV OF CHINA
Multi-view stereoscopic vision three-dimensional scene reconstruction method based on deep learning
PendingCN112734915ARealize 3D scene reconstructionCharacter and pattern recognitionNeural architecturesPattern recognitionFeature extraction
The invention relates to a multi-view stereoscopic vision three-dimensional scene reconstruction method based on deep learning, and aims to solve the problem that the existing reconstruction method based on deep learning generates a 3D cost body by extracting the feature of the last layer of an image, and does not well utilize shallow features, so that information of different scales is lost. Moreover, according to those methods, during depth map refinement, only the effect of the reference image on depth refinement is considered, and the contribution of the depth of the adjacent image to depth map prediction is ignored. In order to solve the problem, a multi-scale feature extraction and fusion network and a depth map refinement network based on inter-frame correlation are provided to improve the prediction precision and integrity of a scene. Compared with an existing method based on deep learning, the method of the invention has the advantages that the context features of the input image can be learned better, the shielded and missing area of the target scene can be reconstructed, the three-dimensional information of the scene can be restored more completely, and high-precision three-dimensional scene reconstruction is achieved.
Owner:BEIJING UNIV OF TECH
Quick refreshing method and device for video code processing
ActiveCN1964487AQuick refreshGood overall compressionTelevision systemsDigital video signal modificationData streamInformation transmission
The provided fast refresh method in video coding process comprises: in every video signal frame, according to frame-to-frame cycle rule, coding new non-correlative data to code video information based on interframe correlation. It also provides a relative coding device controlled by user register. This invention can minimize the video information transmission error and data interruption, and improve efficiency and quality for compression and coding.
Owner:BEIJING MICRO VISION COMM DIGITAL TECH
Source data acquisition and compression method for wireless projection
InactiveCN103248890AReduce data volumeSmooth projection screenTelevision systemsDigital video signal modificationData transmissionRegion of interest
The invention discloses a source data acquisition and compression method for wireless projection, which belongs to the field of wireless projection. The method can quickly and efficiently capture and compress screen images, transmits the compressed screen images to a service terminal for display via a wireless network, and finally achieves the function of wireless projection. The scheme adopts screenshot based on mirror image driving and an optimization algorithm based on H.264. Compared with the traditional screenshot GDI and DirectX, the method has the advantages of low occupancy of CPU, high quality of images, high efficiency of screenshot and the like. Aiming to the characteristic of the large interframe relativity of the screen images, the video compression algorithm based on the H.264 is improved, and on the basis, layered codes based on ROI (Region of Interest) is introduced, and as a result, the data transmission capacity is lower and the quality is higher.
Owner:陈日清
Turbo balancing, inter-frame, intro-frame relevant prediction method and Turbo balancing, inter-frame, intro-frame relevant prediction device
ActiveCN103188040AImprove decoding performanceImprove performanceCode conversionCoding detailsAlgorithmChannel decoder
The invention relates to a Turbo balancing, inter-frame, intra-frame relevant prediction method and a Turbo balancing, inter-frame, intro-frame relevant prediction device. The Turbo balancing method includes the following steps. Channel decoding is performed according to balanced soft information after de-interleaver, soft information of information source bit and soft information of an information channel coding output bit sequence are output. When channel decoding is performed, prior knowledge, obtained by inter-frame relevant prediction, of a frame information source bit is used for decoding operation, namely addition or reduction, in the first iteration. Apart from the first iteration, prior knowledge, obtained by intra-frame relevant prediction, of a frame information source bit is used for decoding operation in other iterations. The invention further discloses an inter-frame relevant prediction method, an intra-frame relevant prediction method in the channel decoding and a corresponding intra-frame relevant prediction device and an intra-frame relevant prediction device. The scheme modifies the inter-frame relevant prediction method and the intra-frame relevant prediction method and uses the inter-frame relevant prediction method and the intra-frame relevant prediction method for balancing Turbo, and decoding performance of the channel decoder is improved.
Owner:ZTE CORP
Reducing latency in video encoding and decoding
Techniques and tools for reducing latency in video encoding and decoding by constraining latency due to reordering of video frames, and by indicating the constraint on frame reordering latency with one or more syntax elements that accompany encoded data for the video frames. For example, a real-time communication tool with a video encoder sets a syntax element that indicates a constraint on frame reordering latency, which is consistent with inter-frame dependencies between multiple frames of a video sequence, then outputs the syntax element. A corresponding real-time communication tool with a video decoder receives the syntax element that indicates the constraint on frame reordering latency, determines the constraint on frame reordering latency based on the syntax element, and uses the constraint on frame reordering latency to determine when a reconstructed frame is ready for output (in terms of output order).
Owner:MICROSOFT TECH LICENSING LLC
Video fingerprint detection and video sequence matching method and system based on visual features
ActiveCN103593464BVideo sequence detection results are goodSimple designCharacter and pattern recognitionSpecial data processing applicationsFeature vectorVision based
The invention proposes a video fingerprint detection method based on visual features, specifically: segment the video sequence according to the inter-frame correlation, extract key frames in the segments; extract video features in each key frame; use pixel feature dictionary Classify the pixels; divide each key frame into multiple different numbers of blocks, count the occurrence times of each element of the pixel feature dictionary in each sub-block according to the classification results of the pixels, and obtain the feature vector of the sub-block; The eigenvectors of all sub-blocks are concatenated to obtain the high-dimensional video fingerprints of the key frames; the dimensionality of each high-dimensional video fingerprint is reduced; the low-dimensional video fingerprints of the key frames of each video segment are connected in chronological order into a key frame video fingerprint string. The present invention also provides a video matching method based on the above-mentioned fingerprint detection method. The invention effectively describes the key information of the video content, greatly reduces the complexity of the algorithm and effectively improves the detection efficiency without affecting the matching rate.
Owner:HUAZHONG UNIV OF SCI & TECH
Method and system for filtering pseudo-star targets of large-view-field star sensor
ActiveCN112595312AAchieve filteringConsistentNavigation by astronomical meansControl systemInterframe correlation
The invention relates to a method and system for filtering pseudo-star targets of a large-view-field star sensor. The method comprises the steps of respectively calculating the angular distance between any two targets in a current frame and a former frame, in the current frame and a former two frame, and in the former one frame and the former two frame and the gray deviation between the two targets, and judging whether all targets in the current frame and the former two frames have inter-frame correlation or not; for the targets with inter-frame correlation, carrying out primary filtering on the pseudo-star targets based on the characteristic that the inter-frame angular distance of the real star targets is stable and invariable; and for the target result after the primary filtering, performing secondary filtering on the pseudo-star targets based on the characteristics of the relative concentration of the inter-frame real star target motion angular distance, and outputting the remaining targets as real star targets. According to the method, inter-frame correlation processing is carried out in combination with motion characteristics of the targets, effective filtering of the pseudo-star targets is realized, the attitude output efficiency of the star sensor is ensured, and the use requirements of a control system are met.
Owner:HUAZHONG PHOTOELECTRIC TECH INST (CHINA SHIPBUILDING IND CORP THE NO 717 INST)
Tampered face video detection method
ActiveCN111144314AGood effectCharacter and pattern recognitionNeural architecturesFeature extractionRadiology
The invention discloses a tampered face video detection method, which comprises the following steps of: decoding face video data into a group of continuous frame images, intercepting a face area of each frame image, and correspondingly storing the face areas as face pictures according to frame serial numbers; extracting each face picture through a feature extractor to obtain a corresponding feature map; and the feature maps of the two continuous frames are input to an inter-frame correlation classifier at the same time, the inter-frame correlation classifier adopts an attention mechanism to fuse the feature maps of the two frames together and then classifies the feature maps, and the classification result is the probability that the two input frames are tampered. According to the method, the information of the frame picture and the inter-frame relationship between the frame picture and the adjacent frame are utilized at the same time, and a good effect can be achieved. Meanwhile, detection is automatically completed, and the method can be suitable for large-scale video platforms and social platforms.
Owner:BEIJING RES INST UNIV OF SCI & TECH OF CHINA +1
Power measurement method and device, chip, equipment and computer readable storage medium
PendingCN112153729AHigh precisionPower managementBaseband system detailsInterframe correlationPower measure
The embodiment of the invention discloses a power measurement method and device, a chip, equipment and a computer readable storage medium, and the method comprises the steps: receiving a communicationsignal comprising a plurality of subframes, calculating intra-subframe related results and inter-subframe related results corresponding to the plurality of subframes, and calculating the sum of the intra-subframe related results and the inter-subframe related results; obtaining the correlation result summation, and determining the receiving power of the communication signal according to the powervalue corresponding to the correlation result summation, so that the power measurement precision can be improved.
Owner:BEIJING SIGBEAN INFORMATION TECH CO LTD
Video automatic motion compensation method and device, equipment and storage medium
InactiveCN112788337AImprove experienceTelevision system detailsColor television detailsComputer graphics (images)Engineering
The invention discloses a video automatic motion compensation method, device and equipment and a storage medium, wherein the video automatic motion compensation method is used for video playing equipment. The method comprises the following steps: acquiring a video played by the video playing equipment; detecting the frame rate of the video, and if the frame rate is greater than a first set threshold value, judging whether scene change occurs based on the inter-frame relevancy between two adjacent frames in each base group of the video, wherein the base group is composed of video frame images contained per second; and if the scene change occurs, obtaining the inter-frame movement speed between the adjacent frames in which the scene change occurs in the base group, and performing motion compensation on the video based on the inter-frame movement speed and the inter-frame correlation degree. The invention aims to solve the problem that an existing motion compensation method cannot be automatically adapted in some pictures, so that the user experience is influenced.
Owner:SHENZHEN SKYWORTH RGB ELECTRONICS CO LTD
Target detection and recognition method and system in low-altitude and sea-surface environments
ActiveCN108710127BEfficient detectionSuppression of strong sea clutterRadio wave reradiation/reflectionInterframe correlationComputer vision
The invention provides a target detection identification method and system in a low altitude and sea surface environment. The method comprises steps of establishing a space virtual reference coordinate system and carrying out self-calibration; for different measurement conditions, using corresponding coordinate conversation methods, allowing data of measurement points and targets to be establishedin the space virtual reference coordinate system; calculating all geometrical parameters of each target and coordinates of the target; through two radars in different heights, carrying out self-calibration and blind supplementation; by use of a de-noising method of interframe correlation, processing radar signals; and carrying out information fusion processing on multi-source target data in the radars and carrying out target detection and tracking. According to the invention, by use of dynamic self-calibration technology, and adaptive detection and recognition technology, strong sea clustersare inhibited; positioning precision and recognition rate of the targets are improved; target detection ability for observing sea area during day and night in a severe weather environment is improved;and effective detection, tracking, recognition and law enforcement monitoring evidence obtaining for small targets on the sea surface and low-altitude small targets like unmanned aerial vehicles areachieved.
Owner:上海鹰觉科技有限公司
Method and system for fast detecting static stacking letters in online video stream
InactiveCN100515048CAvoid quality problemsGuaranteed detection speedTelevision system detailsColor television detailsText detectionComputer graphics (images)
The invention relates to an image and video processing method and system, in particular to a method and system for rapidly detecting static superimposed text in an online video stream. The static superimposed text detection method utilizes inter-frame correlation information and wavelet domain modeling and other methods, which can effectively remove moving text and background areas, and retain the static superimposed text area. The method can realize the rapid detection of the static superimposed text position in the online video stream. At the same time, an online video stream retrieval system is constructed based on the static superimposed text detection method. In this system, different parameter families are used for fast text detection on each user terminal, and the results are converted into text streams through the OCR control. After each user terminal transmits the text stream to the centralized retrieval server for integration, a multi-time granular retrieval function for each channel and a content-based quick browsing function for each channel can be provided. The present invention can realize the functions of synchronous analysis, indexing, retrieval and browsing of multiple online video streams with multiple qualities without infringing copyright, and does not need to establish special servers for various video streams.
Owner:PEKING UNIV
Video SAR moving target detection method based on image region accumulation
ActiveCN110728695AImpact of suppression detectionEasy to detectImage enhancementImage analysisPattern recognitionSynthetic aperture radar
The invention discloses a video SAR (Synthetic Aperture Radar) moving target detection method based on dynamic shadows, which mainly solves the problem that the existing video SAR moving target detection is not steady. The implementation scheme is as follows: 1) designing an accumulation window and carrying out region accumulation on a video SAR image sequence; 2) determining an accumulation threshold and carrying out image reforming on an accumulation result; 3) carrying out binary segmentation on the reformed image; 4) performing connected domain size statistics on the image after binary segmentation, and reserving connected domains of which the sizes are within the range of 0.4 time of the total number of pixel points occupied by the targets and 2 times of the total number of pixel points occupied by the targets; and 5) performing inter-frame correlation processing on the image after connected domain processing to remove non-target shadows, and completing moving target detection. According to the video SAR moving target detection method, the video SAR image is subjected to regionalized multi-frame joint detection, the false alarm probability and the missed alarm probability areeffectively restrained, the detection performance is improved, and the video SAR moving target detection method can be used for tracking and real-time detection of the video SAR moving target.
Owner:XIDIAN UNIV
Subtitle detection method and device as well as background recovery method and device
InactiveCN101835011BEffective reflectionDetection speedTelevision systemsRecovery methodPattern recognition
Owner:HUAWEI TECH CO LTD +1
Space-based infrared aerial target detection method, storage medium and computer equipment
PendingCN114820647AReduce resource requirementsTo achieve effective detectionImage enhancementImage analysisContrast levelInfrared remote sensing
The invention belongs to the technical field of infrared remote sensing and infrared space, and discloses a space-based infrared aerial target detection method, a storage medium and computer equipment, and the method comprises the steps: determining a reference frame Ib and a reference frame Ib + l, carrying out the local normalization processing of a frame b, carrying out the local normalization processing of a frame b + j, and carrying out the inter-frame correlation difference; after the inter-frame correlation difference is completed, dipole enhancement and spatial difference are carried out; multiplying the data completed by the space background suppression, the dipole enhancement and the space difference to obtain an intermediate result; and performing global normalization processing, local feature contrast and adaptive threshold segmentation on the intermediate result data, and performing result detection. According to the invention, space-based infrared aerial target detection is realized by constructing the local inter-frame matching difference model and the space local contrast model, the processing complexity of infrared aerial target detection and the resource demand of hardware realization are reduced, and the target detection efficiency is effectively improved.
Owner:SHANGHAI INST OF TECHNICAL PHYSICS - CHINESE ACAD OF SCI
Unsupervised video target segmentation method based on local and global memory mechanism
PendingCN113269021ACharacter and pattern recognitionNeural learning methodsFeature learningInterframe correlation
The invention discloses an unsupervised video target segmentation method based on a local and global memory mechanism, and belongs to the technical field of feature learning and video target segmentation. The method comprises the following steps: firstly, extracting embedding features of a pair of same videos; selecting global memory candidate frames in the video, and extracting global memory candidate features, wherein each global memory candidate feature corresponds to each node corresponding to the graph convolution network, and the enhanced global memory feature expression is carried out; extracting mutual attention information between a pair of frames through a local memory module, and alternately regarding the mutual attention information as a target and a search role in an attention mechanism for mutual attention enhancement; and finally, obtaining a predicted target mask through a decoder, calculating loss by using cross entropy loss, and updating the whole model so as to obtain a final segmentation network. According to the method, local and global memory mechanisms are considered at the same time, and reliable short-time and long-time video inter-frame correlation information is obtained at the same time, so that unsupervised video target segmentation is realized.
Owner:BEIJING UNIV OF TECH
Video sar moving target detection method based on image area accumulation
ActiveCN110728695BImpact of suppression detectionEasy to detectImage enhancementImage analysisPattern recognitionBinary segmentation
The invention discloses a video SAR moving target detection method based on image area accumulation, which mainly solves the problem of unsteady detection of the existing video SAR moving target. The realization scheme is as follows: 1) Design the accumulation window and perform regional accumulation on the video SAR image sequence; 2) Determine the accumulation threshold and perform image reconstruction on the accumulation result; 3) Binary segmentation of the reconstructed image; The image after value segmentation carries out the size statistics of the connected domain and retains the connected domain whose size is within the range of 0.4 times the total number of pixels occupied by the target and 2 times the total number of pixels occupied by the target; 5) performs inter-frame correlation on the processed image of the connected domain Processing removes non-target shadows and completes moving target detection. The invention effectively suppresses the probability of false alarms and missing alarms and improves detection performance by performing regional multi-frame joint detection on video SAR images, and can be used for tracking and real-time detection of video SAR moving targets.
Owner:XIDIAN UNIV
A moving target detection method based on deep learning
ActiveCN107123131BImprove accuracyAvoid missing detectionImage enhancementImage analysisPattern recognitionNetwork model
The invention discloses a moving target detection method based on deep learning, which belongs to the technical field of video image processing. The method includes: training a constructed deep neural network model to obtain a target neural network model capable of distinguishing meaningful moving targets; using The target neural network model detects moving targets from the currently collected scene images. The invention utilizes deep learning technology and inter-frame correlation information, so that the method consumes less memory space and accurately detects moving targets with less calculation cost.
Owner:ANHUI TSINGLINK INFORMATION TECH
A Method of Video Sequence Reconstruction Based on Compressive Sensing
ActiveCN107360426BQuality improvementReduce data volumeDigital video signal modificationPattern recognitionNoise (video)
The invention discloses a video sequence reconstruction method based on compressed sensing. An encoding end separates an original video sequence into two types of frames, i.e., key frames and non-key frames, and measurement is performed respectively by use of different sampling rates. At a decoding end, different video sequences are subjected to performance judgement of two different algorithms, i.e., motion estimation / compensation prediction and optical flow method prediction, and then through combination with a comparison result, all the non-key frames are subjected to good-performance algorithm reconstruction by use of inter-frame correlation . When a video is processed, the amount of data needing to be sampled is small and the quality of the recovered video is good. Compared with a traditional algorithm of video independent reconstruction, the peak signal-to-noise ratio at a low sampling rate is improved and the reconstruction time is shortened.
Owner:FUZHOU UNIV
Bad frame automatic detection method and device, equipment, storage medium and program product
ActiveCN113365103AImplement automatic detectionImprove detection efficiencySelective content distributionFuzzy inferenceAlgorithm
The invention provides a bad frame automatic detection method and device, equipment, a storage medium and a program product, and relates to the video processing technology, and the method comprises the steps: obtaining the information of each frame of picture in a to-be-detected video, and determining the difference between adjacent frames according to the information of each frame of picture; according to a preset fuzzy inference rule, performing correlation analysis on the difference between the adjacent frames to obtain an inter-frame correlation result; and determining a bad frame according to an inter-frame correlation result among the frames of pictures. According to the scheme provided by the invention, automatic detection of the bad frame can be realized, and the detection efficiency is improved; and moreover, correlation analysis is carried out on the difference of the adjacent frames by applying a fuzzy reasoning technology, the difference of the adjacent frames is the difference between the current frame and the front and back frames, the misjudgment that the video frame subjected to normal shot conversion in the video is mistakenly recognized as a bad frame can be effectively avoided through the mode, and then the video frame automatic detection efficiency is improved.
Owner:深圳市帧彩影视科技有限公司
A Time Domain Information Fusion Method Based on Inter-Frame Correlation and Slot Worst
ActiveCN105979266BImprove performanceReduce complexityDigital video signal modificationTime domainAlgorithm
The invention discloses an interframe relevance and time slot worst based time domain information fusion method. A prediction score of each frame of a video is calculated according to an objective video quality evaluation model; sliding window mean value treatment is performed on the prediction scores and ranking is performed on the results subjected to the sliding window treatment; and the mean value of the worst frames is taken as the predication score of the whole video sequence. According to the invention, on the basis of a prior time domain fusion method, the performance of an objective video quality evaluation algorithm is improved effectively and the method provided by the invention is a simple and effective time domain information fusion method meeting human visual characteristics. Compared with a prior mean value fusion method, the method is small in errors and meets a result of human eye practical evaluation. Considering the influence among frames, the method is good in effect. The method provided by the invention takes human eye delay effect into consideration and also takes connection among the frames into consideration. By adopting sliding window mean value method for processing data of each frame, the estimation accuracy is improved substantially.
Owner:XIDIAN UNIV
Method and computing system for reducing latency in video decoding
Techniques and tools for reducing delay in video encoding and decoding by constraining delay due to reordering of video frames and by indicating constraints on frame reordering delay with one or more syntax elements accompanying encoded data for the video frame. For example, a real-time communication tool with a video encoder sets a syntax element indicating a constraint on a frame reordering delay consistent with an inter-frame correlation between frames of a video sequence, and then the The tool outputs this syntax element. A corresponding real-time communication tool having a video decoder receives the syntax element indicating a constraint on frame reordering delay, determines the constraint on frame reordering delay based on the syntax element, and uses the constraint on frame reordering delay to determine a reconstructed frame When to prepare for output (in terms of output order).
Owner:MICROSOFT TECH LICENSING LLC
A method for detecting tampered face video
ActiveCN111144314BGood effectCharacter and pattern recognitionNeural architecturesFeature extractionMedicine
The invention discloses a face tampering video detection method, comprising: decoding the face video data into a group of continuous frame images, and intercepting the face area of each frame image, and correspondingly saving it as a face picture according to the frame number; The feature extractor extracts each face picture and obtains the corresponding feature map; the feature maps of two consecutive frames are input to the inter-frame correlation classifier at the same time, and the inter-frame correlation classifier adopts the attention mechanism to combine the two frames. The feature maps are fused together for classification, and the classification result is the probability that the two input frames are tampered with. This method utilizes the information of the frame picture itself and the inter-frame relationship with its adjacent frames at the same time, which can achieve better results. At the same time, the detection is done automatically and can be applied to large-scale video platforms and social platforms.
Owner:BEIJING RES INST UNIV OF SCI & TECH OF CHINA +1
Target Recognition Method Based on Quality Evaluation
ActiveCN108765394BSolve the problem of object recognitionValid descriptionImage enhancementImage analysisImaging qualityInterframe correlation
Owner:SHANGHAI JIAO TONG UNIV
Adaptive spectral subtraction for real-time speech enhancement
InactiveCN103594094BImprove performanceImprove intelligibilitySpeech analysisNoise spectrumBackground noise
The invention discloses a method for real-time speech enhancement by adaptive spectral subtraction. The method includes: constructing a dynamic threshold for distinguishing speech and non-speech in noisy speech, and proposing a time-varying update principle of noise spectrum according to the dynamic threshold; making full use of the relative The correlation extraction information between adjacent frames realizes the pure speech spectrum smoothing iterative estimation method; for the practical problem that the sound signal is difficult to extract under non-stationary noise and strong background noise, an adaptive spectral subtraction speech enhancement algorithm is given; The fast tracking noise algorithm performs frame-by-frame smooth update on the non-stationary noise, and can better estimate the noise spectrum; the algorithm proposed by the invention can suppress the background noise more effectively, and improve the speech quality and intelligibility after denoising. This method has the advantages of low calculation cost, simple and easy implementation, and good real-time performance. It provides a new way for strong background noise denoising and weak signal detection.
Owner:HUNAN INT ECONOMICS UNIV
Who we serve
- R&D Engineer
- R&D Manager
- IP Professional
Why Patsnap Eureka
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com