CN103747189A - Digital image processing method - Google Patents
Digital image processing method Download PDFInfo
- Publication number
- CN103747189A CN103747189A CN201310722906.XA CN201310722906A CN103747189A CN 103747189 A CN103747189 A CN 103747189A CN 201310722906 A CN201310722906 A CN 201310722906A CN 103747189 A CN103747189 A CN 103747189A
- Authority
- CN
- China
- Prior art keywords
- video
- image
- sample
- training
- extract
- 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
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 16
- 238000004458 analytical method Methods 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 45
- 238000000034 method Methods 0.000 claims description 30
- 239000000284 extract Substances 0.000 claims description 17
- 230000002708 enhancing effect Effects 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 3
- 206010039203 Road traffic accident Diseases 0.000 claims description 2
- 239000000654 additive Substances 0.000 claims description 2
- 230000000996 additive effect Effects 0.000 claims description 2
- 238000005286 illumination Methods 0.000 claims description 2
- 230000008520 organization Effects 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 3
- 230000002123 temporal effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 5
- 238000013507 mapping Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005316 response function Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
本发明涉及一种数字图像处理方法，包括以下步骤：对摄像机采集到高速公路视频图像依次进行视频增强、视频分析以及视频理解。视频增强是将同一场景多幅具有不同曝光参数的低分辨率图像重建为具有高亮度动态范围和高分辨率的高质量图像，为视频分析层提供高质量的视频图像，从而提高视频处理结果的可靠性；视频分析是通过运动目标进行检测，运动估计和目标跟踪视频分析算法来提取视频底层和中层的时空对象特征，为高层视频处理中的事件识别提供推断依据；所述视频理解通过分析和理解视频分析层提供的时空对象底层特征来完成对监控视频事件的识别。本发明实现使得检测事件的类型更广，精度更高，事件的正确识别确保了高速公路的自动控制。
The invention relates to a digital image processing method, which comprises the following steps: video enhancement, video analysis and video understanding are sequentially performed on the highway video images collected by a camera. Video enhancement is to reconstruct multiple low-resolution images of the same scene with different exposure parameters into high-quality images with high brightness dynamic range and high resolution, and provide high-quality video images for the video analysis layer, thereby improving the accuracy of video processing results. Reliability; video analysis is to detect through moving targets, motion estimation and target tracking video analysis algorithms to extract the temporal and spatial object features of the bottom layer and middle layer of the video, and provide inference basis for event recognition in high-level video processing; the video understanding through analysis and Understand the underlying characteristics of spatio-temporal objects provided by the video analysis layer to complete the identification of surveillance video events. The realization of the invention makes the types of detection events wider and the accuracy higher, and the correct identification of events ensures the automatic control of expressways.
Description
Technical field
The present invention relates to a kind of digital image processing method, relate in particular to HDR and the HR image rebuilding method of a kind of highway video based on sample prediction.
Background technology
Affect a lot of because have of picture quality, as spatial resolution, luminance contrast, noise etc.High-quality image, when effectively showing high contrast scene, also should have higher spatial resolution.For the high dynamic range images demonstration of image and the Problems of Reconstruction of spatial resolution, many scholars have carried out some fruitful research work, but they independently carry out respectively substantially.Existing Super-Resolution Restoration from Image Sequences supposes that the exposure parameter of multiple image is that parameter and noise parameter constant, camera response function are known conventionally.But the image obtaining in real world is difficult to meet to above assumed condition conventionally.Therefore, in unified technological frame, rebuild high dynamic range and high-definition picture, the theoretical foundation to image co-registration and practical application all have certain value.
Video traffic event detection technology is generally divided into " Indirect Detecting Method " and " direct detecting method " two large classes.Front a kind of be the existence that indirectly judges traffic events according to the variation of traffic flow, this method, due to the complexity of data error and traffic conditions, causes the time of event detection longer, and is not suitable for using in the situation that the volume of traffic is lower.In addition, the low image collecting that causes with poor contrast of the common resolution of image of camera acquisition cannot meet the accuracy that " Indirect Detecting Method " requires at present.
Second method is directly by the video image collecting, and by image processing techniques, finds the method that Vehicle Driving Cycle is abnormal, also can have good testing result in the situation that the volume of traffic is lower.But the method is also limited by the impact of the low and poor contrast of the common resolution of image of current camera acquisition.
Summary of the invention
One of object of digital image processing method of the present invention is: by Same Scene, several low-resolution images with different exposure parameters are redeveloped into and have high brightness dynamic range and high-resolution high quality graphic.
Two of the object of digital image processing method of the present invention is: existing highway video system is outputting high quality and Video Events accurately directly.
A kind of digital image processing method, comprise the following steps: camera acquisition is carried out to video enhancing, video analysis and video successively to highway video image and understand, wherein, it is that several low-resolution images with different exposure parameters are redeveloped into and have high brightness dynamic range and high-resolution high quality graphic by Same Scene that described video strengthens, for video analysis layer provides high-quality video image, thus the reliability of raising Video processing result; Described video analysis is to detect by moving target, and estimation and target following video analysis algorithm extract the space-time characteristics of objects in video bottom and middle level, for the Identification of events in high-rise Video processing provides deduction foundation; The space-time object low-level image feature that described video is understood by analyzing and understand video analysis layer and providing completes the identification to monitor video event.
Further, it is that HDR and HR image rebuilding method based on sample prediction realized that described video strengthens, and this method for reconstructing comprises step 1, off-line training part and step 2, rebuilds part online.
Further, described step 1, off-line training part consist of following sub-step: step 1.1, collection training sample: several that training image is Same Scene have low-resolution image and the target image corresponding to a width of different exposure parameters; In training image, extract corresponding LDR-LR and HDR-HR image information piece to as training sample; Step 1.2, tissue training's sample: adopt clustering method to carry out taxonomic organization to the sample set from Different background illumination district; Step 1.3, be a fallout predictor of each cluster sample set training; Obtain three classification fallout predictors in corresponding clear zone, dark space and moderate district.
Further, step 2, the online part of rebuilding consist of following sub-step: step 2.1, the scene brightness of input image sequence is cut apart, formed three different regions of exposure; Step 2.2, the basic layer of input image sequence is estimated; Step 2.3, according to the brightness classification results of input picture, by the classification fallout predictor training, the detailed information of each position image block is predicted, obtain the levels of detail of scene; Step 2.4, the basic layer of general and levels of detail estimated result additive fusion; Step 2.5, fused images is carried out to the constraint of low-resolution image observation model, obtain reconstructed results.
Further, step 1.1, gather training sample step and comprise: step 1.1.1, according to scene brightness difference, sample is divided into clear zone, dark space and San Ge region, moderate district; Step 1.1.2, detailed information are calculated: the detailed information that employing two-sided filter extracts every width sample image is as sample data; Step 1.1.3, sample collection: sample is the paired image information piece that input LDR-LR and target HDR-HR detail pictures correspondence position extract; According to scene brightness classification results, in the clear zone sample detail view corresponding with target image at short LDR-LR of time for exposure, extract; In the dark space sample detail view corresponding with target image at long LDR-LR of time for exposure, extract; The moderate district of brightness, selects to extract in detail view that moderate LDR-LR of time for exposure is corresponding with target image; For three brightness region, extract three corresponding training sample sets.
Further, in described video understanding, Video Events comprises output road traffic accident (comprise parking, queuing, hypervelocity and drive in the wrong direction) and traffic parameter (comprising flow, car speed, vehicle classification).
Based on HDR and the HR image rebuilding method of sample prediction, the method is divided into off-line training part and the online two parts of rebuilding; Off-line part comprises learning sample collection, tissue and classification fallout predictor training part.The online part of rebuilding is that to carry out HDR-HR be high dynamic range and high resolution image reconstruction for low-dynamic range and low-resolution image to several LDR-LR with little same exposure parameter of input.First, by the average image of input picture, carry out the background luminance classification of scene; Then, according to brightness classification results, utilize the classification fallout predictor that off-line part trains to carry out high dynamic range and the prediction of high-resolution detailed information to input picture, finally rebuild HDR-HR image.
This digital image processing method has following beneficial effect:
(1) the present invention is realized and is made the type of the event that detects wider by video enhancing, video analysis and video understanding, and precision is higher, and the automatic control of highway has been guaranteed in the correct identification of event.
(2) the present invention, by the study to example sample, sets up the mapping relations between LDR-LR (Low Dynamic Range-Low Resolution) and HDR-HR (High Dynamic Range-High Resolution).By rationalization's sample, the strategies such as classification based training learning model are without the artificial reconstruction of combining that realizes HDR-HR image mutual in the situation that.
(3) the inventive method is divided into off-line and online two parts.Off-line part mainly completes the training of collection, tissue and the classification fallout predictor of example sample; Online part trains the classification fallout predictor obtaining to complete the reconstruction of combining of image by off-line part.The reconstruction of combining of carrying out image high dynamic range and super-resolution in the framework based on sample prediction study is provided, and the method can be rebuild high dynamic range and high-resolution target image simultaneously.
Accompanying drawing explanation
Fig. 1: the flow chart of digital image processing method of the present invention;
Fig. 2: off-line training part flow chart of the present invention;
Fig. 3: the online reconstruction portion of the present invention is divided flow chart;
Fig. 4: sample extraction mode of the present invention (corresponding relation); A) b) LDR-LR detail pictures of HDR-HR detail pictures.
Embodiment
Below in conjunction with Fig. 1 to Fig. 4, the present invention will be further described:
As shown in Figure 1, a kind of digital image processing method, comprise the following steps: camera acquisition is carried out to video enhancing, video analysis and video successively to highway video image and understand, wherein, it is that several low-resolution images with different exposure parameters are redeveloped into and have high brightness dynamic range and high-resolution high quality graphic by Same Scene that described video strengthens, for video analysis layer provides high-quality video image, thus the reliability of raising Video processing result; Described video analysis is to detect by moving target, and estimation and target following video analysis algorithm extract the space-time characteristics of objects in video bottom and middle level, for the Identification of events in high-rise Video processing provides deduction foundation; The space-time object low-level image feature that described video is understood by analyzing and understand video analysis layer and providing completes the identification to monitor video event.
Put forward the methods of the present invention is divided into off-line training and the online two parts of rebuilding.Off-line training part flow chart as shown in Figure 2, comprises learning sample collection, tissue and classification fallout predictor training part.Sample collection process is divided three classes and carries out respectively according to scene brightness difference.Adopt clustering method to organize sample file.Then, by linearity or nonlinear prediction device learning method, classification fallout predictor is trained.
Online reconstruction portion divides flow chart as shown in Figure 3, and the LR-LDR image 3 width of input to different exposure parameters carries out HDR-HR reconstruction.First, by the average image of input picture, carry out the background luminance classification of scene; Then, according to brightness classification results, utilize the classification fallout predictor that off-line training part trains input picture to be carried out to the prediction of high dynamic range and high-resolution detailed information, finally rebuild high-frequency information.
Below in conjunction with example, the method is elaborated.
(1) off-line training part;
Training image is chosen many group HDR scene image compositions.Each HDR scene training plan is by the excessive I that exposes
1, moderate T exposes
0with the too small T of exposure
-1three width LDR-LR images and the target image IHDR-HR composition of HDR-HR scene corresponding to a width.When sample collection, first HDR-HR scene is carried out to background luminance classification.Background luminance classification can adopt kinds of schemes, as can be to the average image I of three width LDR-LR images
averagecarry out K mean cluster, be divided into three classes, thereby image is divided into clear zone, moderate district and San Ge region, dark space.According to scene brightness classification results, in the clear zone sample detail view corresponding with target image at short LDR-LR of time for exposure, extract; In the dark space sample detail view corresponding with target image at long LDR-LR of time for exposure, extract; The moderate district of brightness, selects to extract in detail view that moderate LDR-LR of time for exposure is corresponding with target image; Acquisition example sample in each region, forms three training sample sets.
Example sample is comprised of paired image information piece, i.e. HDR-HR image block and corresponding LDR-LR image block.Before sample extraction, respectively LDR-LR and HDR-HR training image are carried out to two-sided filter filtering, then original image is deducted to filtered image and obtain detailed information.On corresponding detailed information image, according to the paired example sample of the corresponding relation collection shown in Fig. 4.In Fig. 4, take sampling multiple, equal 2 as example.What corresponding sample extracted respectively is the vector of 16 dimensions.
Three training sample database that collect are carried out respectively the sample tissue based on cluster.Can adopt K mean cluster, the LDR-LR part in sample is carried out to cluster.
For each Sample Storehouse, train a classification fallout predictor.Classification fallout predictor consists of one group of sub-fallout predictor, the corresponding sub-fallout predictor of sample set of each cluster classification.In the training of sub-fallout predictor, all samples of corresponding classification are training sample.Wherein LDR-LR part is input, and HDR-HR part is target.The object of fallout predictor is to describe the mapping relations of similar sample LDR-LR and HDR-HR.This mapping relations are for instructing the HDR-HR image reconstruction of non-training sample LDR-LR image sequence.Sub-fallout predictor can adopt simple least mean-square error (Least Mean Squares, LMS) fallout predictor.
The object of off-line training part is training and the classification fallout predictor of the corresponding number of background luminance classification quantity, represents the mapping relations between the concentrated LDR-LR of different training samples and HDR-HR.Classification fallout predictor is for the detailed information prediction of online process of reconstruction.
(2) rebuild online part.
The input picture that does not belong to training image take three width is example, I
--1be a shorter image of time for exposure, I
1the image of growing for the time for exposure, I
0for time for exposure normal picture.In order to keep scene overall brightness dynamic range, select the average image of three width experiment input pictures as LDR-LR initial pictures, initial pictures is amplified to target image size through bilinear interpolation, as basic tomographic image.To I
0gray level image carry out K mean cluster and obtain scene brightness classification, be partitioned into the moderate district of clear zone, dark space and brightness.
To I
--1, I
0and T
1carry out respectively detailed information extraction, be about to the difference image of former figure and the filtered smoothed image of two-sided filter as LDR-LR detail pictures.
According to brightness classification results, in the process of rebuilding in the pixel in what region, adopt respectively the classification fallout predictor of corresponding classification to predict.During prediction, the code book that first input data produce by sample classification process is encoded, i.e. classification; Then according to its classification, select corresponding sub-fallout predictor to carry out detailed information prediction.
Corresponding to three width input pictures, for clear zone, I
-1the corresponding details of image is often for dark space, I
-1the corresponding details of epigraph is many.Corresponding normal region I
0the corresponding details of image is many.Prisoner this, when being used for carrying out the prediction of high-frequency information with classification fallout predictor, correspond respectively to different luminance areas, adopt different input pictures to instruct the prediction of detailed information, estimate the stacked HDR-HR image that forms fusion in initial estimation image that is added to of the detailed information obtaining most.
Finally, by image observation model, the basic tomographic image that adopts the value of taking out to amplify carries out model constrained to fused images, by iteration optimization, obtain reconstructed results image.
The high dynamic range and the super-resolution associating method for reconstructing that the present invention is based on study, can high contrast scene carry out effective imaging, reaches the target of simultaneously rebuilding high-resolution and high dynamic range images.Off-line training process can once be trained, repeatedly application.Online reconstruction is effective, fast operation.
By reference to the accompanying drawings the present invention has been carried out to exemplary description above; obvious realization of the present invention is not subject to the restrictions described above; as long as the various improvement that adopted method design of the present invention and technical scheme to carry out; or without improving, design of the present invention and technical scheme are directly applied to other occasion, all in protection scope of the present invention.
Claims (6)
1. a digital image processing method, comprise the following steps: camera acquisition is carried out to video enhancing, video analysis and video successively to highway video image and understand, wherein, it is that several low-resolution images with different exposure parameters are redeveloped into and have high brightness dynamic range and high-resolution high quality graphic by Same Scene that described video strengthens, for video analysis layer provides high-quality video image, thus the reliability of raising Video processing result; Described video analysis is to detect by moving target, and estimation and target following video analysis algorithm extract the space-time characteristics of objects in video bottom and middle level, for the Identification of events in high-rise Video processing provides deduction foundation; The space-time object low-level image feature that described video is understood by analyzing and understand video analysis layer and providing completes the identification to monitor video event.
2. digital image processing method according to claim 1, is characterized in that: it is that HDR and HR image rebuilding method based on sample prediction realized that described video strengthens, and this method for reconstructing comprises step 1, off-line training part and step 2, rebuilds part online.
3. digital image processing method according to claim 2, is characterized in that: described step 1, off-line training part consist of following sub-step: step 1.1, gather training sample: several that training image is Same Scene have low-resolution image and the target image corresponding to a width of different exposure parameters; In training image, extract corresponding LDR-LR and HDR-HR image information piece to as training sample; Step 1.2, tissue training's sample: adopt clustering method to carry out taxonomic organization to the sample set from Different background illumination district; Step 1.3, be a fallout predictor of each cluster sample set training; Obtain three classification fallout predictors in corresponding clear zone, dark space and moderate district.
4. digital image processing method according to claim 3, is characterized in that: step 2, the online part of rebuilding consist of following sub-step: step 2.1, the scene brightness of input image sequence is cut apart, formed three different regions of exposure; Step 2.2, the basic layer of input image sequence is estimated; Step 2.3, according to the brightness classification results of input picture, by the classification fallout predictor training, the detailed information of each position image block is predicted, obtain the levels of detail of scene; Step 2.4, the basic layer of general and levels of detail estimated result additive fusion; Step 2.5, fused images is carried out to the constraint of low-resolution image observation model, obtain reconstructed results.
5. according to digital image processing method described in claim 3 or 4, it is characterized in that: step 1.1, gather training sample step and comprise: step 1.1.1, according to scene brightness difference, sample is divided into clear zone, dark space and San Ge region, moderate district; Step 1.1.2, detailed information are calculated: the detailed information that employing two-sided filter extracts every width sample image is as sample data; Step 1.1.3, sample collection: sample is the paired image information piece that input LDR-LR and target HDR-HR detail pictures correspondence position extract; According to scene brightness classification results, in the clear zone sample detail view corresponding with target image at short LDR-LR of time for exposure, extract; In the dark space sample detail view corresponding with target image at long LDR-LR of time for exposure, extract; The moderate district of brightness, selects to extract in detail view that moderate LDR-LR of time for exposure is corresponding with target image; For three brightness region, extract three corresponding training sample sets.
6. according to digital image processing method described in any one in claim 1 to 5, it is characterized in that: in described video understanding, Video Events comprises that output road traffic accident (comprise parking, queuing, hypervelocity and drive in the wrong direction) and traffic parameter (comprise flow, car speed, vehicle classification).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310722906.XA CN103747189A (en) | 2013-11-27 | 2013-12-25 | Digital image processing method |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310609601 | 2013-11-27 | ||
| CN201310609601.8 | 2013-11-27 | ||
| CN201310722906.XA CN103747189A (en) | 2013-11-27 | 2013-12-25 | Digital image processing method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103747189A true CN103747189A (en) | 2014-04-23 |
Family
ID=50504175
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310722906.XA Pending CN103747189A (en) | 2013-11-27 | 2013-12-25 | Digital image processing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103747189A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570849A (en) * | 2016-10-12 | 2017-04-19 | 成都西纬科技有限公司 | Image optimization method |
| CN108960084A (en) * | 2018-06-19 | 2018-12-07 | 清华大学深圳研究生院 | Target tracking method, system, readable storage medium storing program for executing and electronic equipment |
| CN109360163A (en) * | 2018-09-26 | 2019-02-19 | 深圳积木易搭科技技术有限公司 | A kind of fusion method and emerging system of high dynamic range images |
| CN110475072A (en) * | 2017-11-13 | 2019-11-19 | Oppo广东移动通信有限公司 | Shoot method, apparatus, terminal and the storage medium of image |
| CN110620857A (en) * | 2018-06-20 | 2019-12-27 | Zkw集团有限责任公司 | Method and apparatus for creating high contrast images |
| WO2020093694A1 (en) * | 2018-11-07 | 2020-05-14 | 华为技术有限公司 | Method for generating video analysis model, and video analysis system |
| CN113228660A (en) * | 2018-12-18 | 2021-08-06 | 杜比实验室特许公司 | Machine learning based dynamic synthesis in enhanced standard dynamic range video (SDR +) |
| CN113273180A (en) * | 2019-02-27 | 2021-08-17 | 华为技术有限公司 | Image processing apparatus and method |
| CN113920040A (en) * | 2020-07-09 | 2022-01-11 | 阿里巴巴集团控股有限公司 | Video processing method and model construction method |
| WO2024179510A1 (en) * | 2023-02-28 | 2024-09-06 | 华为技术有限公司 | Image processing method and related device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6724941B1 (en) * | 1998-09-30 | 2004-04-20 | Fuji Photo Film Co., Ltd. | Image processing method, image processing device, and recording medium |
| CN102945603A (en) * | 2012-10-26 | 2013-02-27 | 青岛海信网络科技股份有限公司 | Method for detecting traffic event and electronic police device |
| CN103208184A (en) * | 2013-04-03 | 2013-07-17 | 昆明联诚科技有限公司 | Traffic incident video detection method for highway |
-
2013
- 2013-12-25 CN CN201310722906.XA patent/CN103747189A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6724941B1 (en) * | 1998-09-30 | 2004-04-20 | Fuji Photo Film Co., Ltd. | Image processing method, image processing device, and recording medium |
| CN102945603A (en) * | 2012-10-26 | 2013-02-27 | 青岛海信网络科技股份有限公司 | Method for detecting traffic event and electronic police device |
| CN103208184A (en) * | 2013-04-03 | 2013-07-17 | 昆明联诚科技有限公司 | Traffic incident video detection method for highway |
Non-Patent Citations (1)
| Title |
|---|
| 李晓光: "高分辨率与高动态范围图像联合重建研究进展", 《测控技术》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570849A (en) * | 2016-10-12 | 2017-04-19 | 成都西纬科技有限公司 | Image optimization method |
| CN110475072B (en) * | 2017-11-13 | 2021-03-09 | Oppo广东移动通信有限公司 | Method, device, terminal and storage medium for capturing images |
| US11412153B2 (en) | 2017-11-13 | 2022-08-09 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Model-based method for capturing images, terminal, and storage medium |
| CN110475072A (en) * | 2017-11-13 | 2019-11-19 | Oppo广东移动通信有限公司 | Shoot method, apparatus, terminal and the storage medium of image |
| CN108960084A (en) * | 2018-06-19 | 2018-12-07 | 清华大学深圳研究生院 | Target tracking method, system, readable storage medium storing program for executing and electronic equipment |
| CN110620857A (en) * | 2018-06-20 | 2019-12-27 | Zkw集团有限责任公司 | Method and apparatus for creating high contrast images |
| CN109360163A (en) * | 2018-09-26 | 2019-02-19 | 深圳积木易搭科技技术有限公司 | A kind of fusion method and emerging system of high dynamic range images |
| WO2020093694A1 (en) * | 2018-11-07 | 2020-05-14 | 华为技术有限公司 | Method for generating video analysis model, and video analysis system |
| CN113228660B (en) * | 2018-12-18 | 2023-12-12 | 杜比实验室特许公司 | Machine learning based dynamic synthesis in enhanced standard dynamic range video (SDR+) |
| CN113228660A (en) * | 2018-12-18 | 2021-08-06 | 杜比实验室特许公司 | Machine learning based dynamic synthesis in enhanced standard dynamic range video (SDR +) |
| US12086969B2 (en) | 2018-12-18 | 2024-09-10 | Dolby Laboratories Licensing Corporation | Machine learning based dynamic composing in enhanced standard dynamic range video (SDR+) |
| CN113273180A (en) * | 2019-02-27 | 2021-08-17 | 华为技术有限公司 | Image processing apparatus and method |
| US12182963B2 (en) | 2019-02-27 | 2024-12-31 | Huawei Technologies Co., Ltd. | Image processing apparatus and method implementing a neural network to pre-process a CFM image for further process by ISP pipeline |
| CN113920040A (en) * | 2020-07-09 | 2022-01-11 | 阿里巴巴集团控股有限公司 | Video processing method and model construction method |
| WO2024179510A1 (en) * | 2023-02-28 | 2024-09-06 | 华为技术有限公司 | Image processing method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103747189A (en) | Digital image processing method | |
| Scheerlinck et al. | CED: Color event camera dataset | |
| CN111539273B (en) | Traffic video background modeling method and system | |
| CN103413285A (en) | HDR and HR image reconstruction method based on sample prediction | |
| CN114494050B (en) | A self-supervised video deblurring and image interpolation method based on event camera | |
| CN103413286B (en) | United reestablishing method of high dynamic range and high-definition pictures based on learning | |
| DE102020214863A1 (en) | SELF-MONITORED PROCEDURE AND SYSTEM FOR DEPTH ESTIMATION | |
| CN102682442B (en) | Motion target super-resolution image reconstruction method based on optical flow field | |
| CN108182388A (en) | A kind of motion target tracking method based on image | |
| CN111079507B (en) | Behavior recognition method and device, computer device and readable storage medium | |
| WO2018168539A1 (en) | Learning method and program | |
| CN112084928B (en) | Road traffic accident detection method based on visual attention mechanism and ConvLSTM network | |
| Jiang et al. | Event-based low-illumination image enhancement | |
| CN108520238A (en) | A Scene Prediction Method for Night Vision Images Based on Depth Predictive Coding Network | |
| Scott et al. | Kalman filter based video background estimation | |
| CN104700405A (en) | Foreground detection method and system | |
| Pini et al. | Video synthesis from intensity and event frames | |
| Zhiwei et al. | New method of background update for video-based vehicle detection | |
| CN117830178A (en) | A high dynamic remote sensing image reconstruction method and system based on event camera | |
| CN114067172B (en) | A simulation image generation method, a simulation image generation device and an electronic device | |
| Zhang et al. | GlocalFuse-Depth: Fusing transformers and CNNs for all-day self-supervised monocular depth estimation | |
| CN109493370A (en) | A kind of method for tracking target based on spatial offset study | |
| CN118628663A (en) | Three-dimensional digital modeling system and method based on real scene video | |
| CN113409321A (en) | Cell nucleus image segmentation method based on pixel classification and distance regression | |
| CN113808408A (en) | Roadside parking detection method, device and equipment and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| ASS | Succession or assignment of patent right |
Free format text: FORMER OWNER: YANG YANYAN LIU WENJIE Effective date: 20140530 Owner name: NANYANG INSTITUTE OF TECHNOLOGY Free format text: FORMER OWNER: YANG XINFENG Effective date: 20140530 |
|
| C41 | Transfer of patent application or patent right or utility model | ||
| C53 | Correction of patent of invention or patent application | ||
| CB03 | Change of inventor or designer information |
Inventor after: Yang Xinfeng Inventor after: Yang Yanyan Inventor after: Liu Wenjie Inventor after: Lu Yingying Inventor after: Quan Shangke Inventor after: Liu Xiaohui Inventor before: Yang Xinfeng Inventor before: Yang Yanyan Inventor before: Liu Wenjie |
|
| COR | Change of bibliographic data |
Free format text: CORRECT: INVENTOR; FROM: YANG XINFENG YANG YANYAN LIU WENJIE TO: YANG XINFENG YANG YANYAN LIU WENJIE LU YINGYING QUAN SHANGKE LIU XIAOHUI |
|
| TA01 | Transfer of patent application right |
Effective date of registration: 20140530 Address after: Changjiang Road, Nanyang city Henan province 473004 Wancheng District No. 80 Applicant after: Nanyang Science Technology College Address before: Changjiang Road, Nanyang city Henan province 473004 Wancheng District No. 80 Applicant before: Yang Xinfeng Applicant before: Yang Yanyan Applicant before: Liu Wenjie |
|
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140423 |
|
| RJ01 | Rejection of invention patent application after publication |