CN115278054B - Method for recovering high frame rate global shutter video from rolling shutter image - Google Patents
Method for recovering high frame rate global shutter video from rolling shutter image Download PDFInfo
- Publication number
- CN115278054B CN115278054B CN202210726715.XA CN202210726715A CN115278054B CN 115278054 B CN115278054 B CN 115278054B CN 202210726715 A CN202210726715 A CN 202210726715A CN 115278054 B CN115278054 B CN 115278054B
- Authority
- CN
- China
- Prior art keywords
- rolling shutter
- flow
- image
- shutter
- scan line
- 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.)
- Active
Links
- 238000005096 rolling process Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000003287 optical effect Effects 0.000 claims abstract description 17
- 230000000644 propagated effect Effects 0.000 claims description 3
- 231100000462 teratogen Toxicity 0.000 claims description 3
- 239000003439 teratogenic agent Substances 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 abstract description 5
- 238000013135 deep learning Methods 0.000 abstract description 3
- 238000006243 chemical reaction Methods 0.000 abstract description 2
- 230000007547 defect Effects 0.000 abstract description 2
- 230000002457 bidirectional effect Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 7
- 238000003702 image correction Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 229910044991 metal oxide Inorganic materials 0.000 description 2
- 150000004706 metal oxides Chemical class 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000001444 catalytic combustion detection Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005669 field effect Effects 0.000 description 1
- 230000020169 heat generation Effects 0.000 description 1
- 235000015110 jellies Nutrition 0.000 description 1
- 239000008274 jelly Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000002601 radiography Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention relates to a method for recovering a high-frame-rate global shutter video from a rolling shutter image, which aims to overcome the defects of the prior art and provides an internal geometric model for correcting the rolling shutter image, wherein the internal geometric model comprises the following steps: modeling a bidirectional rolling shutter de-distortion flow for removing rolling shutter distortion under a uniform motion model; then, a geometrical relation between the optical flow between the successive frames and the de-distortion flow corresponding to any scan line is established by a simple scaling operation; furthermore, a mechanism of mutual conversion between different de-distortion flows corresponding to different scan lines is established. The method is based on continuous two-frame rolling shutter image data acquired by the CMOS camera, can reverse a rolling shutter imaging mechanism by organically fusing the geometric model into a deep learning network, and simultaneously realizes rolling artifact removal and image time super-resolution, and is relatively suitable for practical applications such as mobile phone shooting, unmanned aerial vehicle shooting, computational shooting and the like by adopting the CMOS camera.
Description
Technical Field
The invention belongs to the technical field of image and video synthesis, relates to a method for recovering a high-frame-rate global shutter video from a rolling shutter image, and particularly relates to an efficient high-frame-rate global shutter video recovery method.
Background
Image sensors can be generally classified into two general types, i.e., a CCD (Charge-Coupled Device) image sensor and a CMOS (Complementary Metal Oxide Semiconductor, complementary metal oxide field effect transistor) image sensor, according to the type of structure thereof. Since the 90 s, CMOS image sensing technology has gained attention and a great deal of research and development resources. In recent years, CMOS image sensors have taken an absolute predominance in the market, and the replacement of CCD image sensors is basically realized, and camera devices using CMOS image sensors have been increasingly popularized in the fields of unmanned driving, unmanned aerial vehicles, virtual reality, augmented reality, computed photography, and the like. The main advantages of CMOS cameras can be summarized in three ways: 1) In terms of cost, the CMOS image sensor chip generally adopts a standard flow process suitable for mass production, and the unit cost is far lower than that of a CCD during mass production; 2) In terms of size, the CMOS sensor can integrate the image acquisition unit and the signal processing unit on the same substrate, so that the volume is greatly reduced, and the CMOS sensor is very suitable for mobile equipment and various miniaturized equipment; 3) CMOS sensors also retain the advantages of low power consumption and low heat generation over CCDs in terms of power consumption.
The CMOS camera mostly adopts a rolling shutter working mechanism, and the CCD camera adopts a global shutter mode. Global shutter cameras typically use sensors to complete the entire pixel exposure and collect light at one instant to obtain a frame of global shutter image. Rolling shutter CMOS cameras capture images by way of line-by-line exposure. The sensor scans the scene light according to the lines to cause pixel points in different lines to have different exposure time, a certain time delay exists between two adjacent lines, and the exposure of pixels in all lines is realized in a very short time so as to obtain a rolling shutter image. In practical applications, most photographic cameras adopt a rolling shutter working mode, and due to an imaging mechanism of progressive exposure, when the camera shoots a scene or a moving object in the scene in moving, the obtained rolling shutter image has rolling shutter effects (namely "jelly" effects) such as image tilting, shaking, twisting and the like. This not only causes serious degradation of imaging quality in photography, but also renders most three-dimensional vision algorithms designed for global shutter camera models ineffective. In recent years, many studies have pointed out that eliminating the rolling shutter effect is extremely important in practical applications.
Over a decade of long-standing development, researchers have achieved a number of very effective results in eliminating the rolling shutter effect. In particular, the deep neural network has been applied to a rolling shutter image correction task in recent years, and has achieved a superior de-distortion effect compared to the conventional model method. In fact, when one is observing two consecutive rolling shutter images, one can automatically infer in the mind a sequence of virtual global shutter images hidden at any instant, i.e. a high frame rate global shutter video. However, the existing deep learning method can only restore a global shutter image corresponding to a specific time from an input rolling shutter image, and does not have a strong automatic inference capability similar to a person. Therefore, reversing the rolling shutter imaging mechanism to recover high frame rate high quality global shutter video would be of great significance for practical applications such as scene understanding, computed radiography, video entertainment, video compression and editing.
Disclosure of Invention
Technical problem to be solved
In order to avoid the deficiencies of the prior art, the present invention proposes a method of recovering a high frame rate global shutter video from a rolling shutter image, providing a method for recovering a high frame rate global shutter video from two consecutive rolling shutter images.
Technical proposal
A method of recovering high frame rate global shutter video from a rolling shutter image, characterized by the steps of:
step 1: inputting two continuous rolling shutter images with the resolution of h multiplied by w multiplied by 3 into an optical flow estimation network PWC-Net, and outputting the rolling shutter images as dense optical flow f between the two rolling shutter images, wherein the resolution of f is h multiplied by w multiplied by 2;
step 2: the two shutter images and the dense light flow f between them are input together to an encoder-decoder UNet network, and output as the dense light flow f and the dense de-distortion flow u of the corresponding intermediate scan line m An association factor graph c between, where u m The resolution of (a) is h×w×2, and the resolution of c is h×w×1;
step 3: explicitly using the correlation factor value c (x) located at pixel x of the kth scan line in the rolling shutter imageCalculating de-distortion flow u corresponding to middle scanning line at pixel x m (x)：
u m (x)＝c(x)×f(x)
Wherein: c (x) represents the associated factor value of the corresponding pixel x in c, and f (x) represents the optical flow value of the corresponding pixel x in f;
step 4: usingDe-teration flow u corresponding to intermediate scanning line m (x) Explicitly propagated to the corresponding arbitrary scan line s E [1, h ]]Is used for removing teratogen s (x)；
Step 5: repeating the step 3 and the step 4 for h multiplied by w pixels x in the image to obtain the corresponding arbitrary scanning line s epsilon [1, h ]]Dense de-terating flow u s Wherein u is s The resolution of (2) is h×w×2;
step 6: according to the scan line s E [1, h ]]Dense de-terating flow u s Transforming the first shutter image using a forward warping technique to recover the scan line s e 1, h]A corresponding global shutter image;
step 7: and (3) sequentially recovering the global shutter image sequences corresponding to the continuous scanning lines s=1, 2, and the number h according to the step (6), and finally outputting the global shutter video with high frame rate.
Advantageous effects
The invention provides a method for recovering a high-frame-rate global shutter video from a rolling shutter image, which aims to overcome the defects of the prior art and provides an internal geometric model of the rolling shutter image correction problem: modeling a rolling shutter de-distortion flow for removing rolling shutter distortion under a uniform motion model; then, a geometrical relation between the optical flow between the successive frames and the de-distortion flow corresponding to any scan line is established by a simple scaling operation; furthermore, a mechanism of mutual conversion between different de-distortion flows corresponding to different scan lines is established. The method is based on continuous two-frame rolling shutter image data acquired by a CMOS camera, and the geometric model is organically fused into a deep learning network, so that the technical scheme for realizing the purpose is as follows: a method of recovering high frame rate global shutter video from rolling shutter images.
Drawings
Fig. 1 is a flow chart of a method of the present invention for recovering high frame rate global shutter video from rolling shutter images.
Fig. 2 is a schematic diagram of a rolling shutter camera and global shutter camera exposure mechanism.
Fig. 3 is a coder-decoder (UNet) network architecture used to estimate the correlation factor.
Fig. 4 is a sequential two-frame raw rolling shutter image acquired by a CMOS camera.
Fig. 5 is a global shutter video image sequence extracted from the rolling shutter image of fig. 4 using the method of the present invention.
Detailed Description
The invention will now be further described with reference to examples, figures:
the technical proposal of the invention is as follows: a method of recovering high frame rate global shutter video from rolling shutter images. Aiming at rolling shutter images shot by a CMOS camera, the effect of recovering high-frame-rate global shutter videos from two continuous rolling shutter images is achieved, and a method for recovering the high-frame-rate global shutter videos from the rolling shutter images is provided by establishing an internal geometric model of the rolling shutter image correction problem.
The invention will be further described with reference to the drawings and experimental test results.
By pixel x= (x, y) T (corresponding depth is Z) for example, assuming that the camera experiences constant motion (v, ω) during imaging, the optical flow at pixel x f= [ f u ,f v ] T Can be expressed as:
where gamma is the camera read-out rate, satisfying 0 < gamma.ltoreq.1, f is the camera focal length, h is the number of horizontal scan lines of the image, where
Representing a potential rolling shutter spatiotemporal geometry model depends on camera motion, camera parameters, scene depth, and scan line position. Eliminating the vertical optical flow component f in (1) v It is possible to further obtain:
then, in order to obtain a de-distortion flow u= [ u ] for converting pixels x of a kth scanning line of the rolling shutter image into a global shutter image corresponding to an s-th scanning line u ,u v ] T The invention builds its model as follows:
next, based on equations (3) and (4), a geometric correlation method between the de-distortion flow and the optical flow can be established by a simple scaling operation:
where c.epsilon. (-1, 1) is the correlation factor, i.e
Since the solution of c involves a complex rolling shutter spatiotemporal geometric model, we implicitly estimate it by designing an efficient deep neural network in the present invention. Finally, based on equation (4), corresponding to scan line s 1 And scan line s 2 The mechanism of the reciprocal transformation between the two rolling shutter de-distortion flows is as follows:
since equation (7) includes only a simple matrix operation, efficient propagation efficiency can be ensured. Moreover, it can be seen from equation (7) that the de-teration flow has a strong row-dependent characteristic, both its size and direction being closely related to the target scan line to be corrected.
To sum up, equations (5) and (7) reveal the intrinsic geometric model of the rolling shutter image correction problem, i.e., equation (5) establishes the geometric relationship between the optical flow between successive frames and the de-distortion flow corresponding to any scan line by a simple scaling operation; equation (7) establishes a mechanism of interconversion between different de-distortion flows corresponding to different scan lines. Through these geometric models, theoretical support can be provided for extracting high frame rate global shutter video from continuous two-frame rolling shutter images. As shown in fig. 1, the technical scheme of the invention mainly comprises the following steps:
step 1: inputting two continuous rolling shutter images with the resolution of h multiplied by w multiplied by 3 into an optical flow estimation network PWC-Net, and outputting the rolling shutter images as dense optical flow f between the two rolling shutter images, wherein the resolution of f is h multiplied by w multiplied by 2;
step 2: the two shutter images and the dense light flow f between them are input together to an encoder-decoder UNet network, and output as the dense light flow f and the dense de-distortion flow u of the corresponding intermediate scan line m An association factor graph c between, where u m The resolution of (a) is h×w×2, and the resolution of c is h×w×1;
step 3: explicitly calculating a de-distortion flow u corresponding to an intermediate scan line at pixel x using an associated factor value c (x) at pixel x of a kth scan line in a rolling shutter image m (x)：
u m (x)＝c(x)×f(x)
Wherein: c (x) represents the associated factor value of the corresponding pixel x in c, and f (x) represents the optical flow value of the corresponding pixel x in f;
step 4: usingDe-teration flow u corresponding to intermediate scanning line m (x) Explicitly propagated to the corresponding arbitrary scan line s E [1, h ]]Is used for removing teratogen s (x)；
Step 5: repeating the step 3 and the step 4 for h multiplied by w pixels x in the image to obtain the corresponding arbitrary scanning line s epsilon [1, h ]]Dense de-terating flow u s Wherein u is s The resolution of (2) is h×w×2;
step 6: according to the scan line s E [1, h ]]Dense de-terating flow u s Transforming the first shutter image using a forward warping technique to recover the scan line s e 1, h]A corresponding global shutter image;
step 7: and (3) sequentially recovering the global shutter image sequences corresponding to the continuous scanning lines s=1, 2, and the number h according to the step (6), and finally outputting the global shutter video with high frame rate.
The effectiveness of the invention can be further illustrated by the following experiments:
(1) Network model training
The invention develops experiments on NVIDIA GeForce RTX 2080Ti GPU based on PyTorch and uses an Adam optimizer to perform network optimization. The present invention employs a current, more sophisticated optical flow estimation network (i.e., PWC-Net) to estimate optical flow between two frames of input, while using a simple encoder-decoder (i.e., UNet) structure to estimate correlation factors. The UNet network architecture is shown in fig. 3. The global shutter image corresponding to the middle scanning line is only used for supervision in the training stage, and the global shutter image corresponding to any scanning line can be effectively transmitted in the testing stage, namely the rolling shutter reversal is realized.
(2) Simulation experiment result analysis
Fig. 4 shows two sets of input consecutive two-frame raw rolling shutter images. It can be seen that the rolling shutter image exhibits significant rolling shutter distortion such as graphic tilting, shaking, warping, etc. With the global shutter video restoration method provided by the present invention, the extracted 6-frame global shutter video image sequence is shown from top to bottom in fig. 5. Tests show that for two continuous frames of rolling shutter images with the input resolution of 640 multiplied by 480, the method can efficiently recover 960 global shutter video image sequences in 1.8 seconds. In conclusion, the effectiveness and the high efficiency of the method are fully verified, and meanwhile, the practical application requirements can be met.
Claims (1)
1. A method of recovering high frame rate global shutter video from a rolling shutter image, characterized by the steps of:
step 1: inputting two continuous rolling shutter images with the resolution of h multiplied by w multiplied by 3 into an optical flow estimation network PWC-Net, and outputting the rolling shutter images as dense optical flow f between the two rolling shutter images, wherein the resolution of f is h multiplied by w multiplied by 2;
step 2: the two shutter images and the dense light flow f between them are input together to an encoder-decoder UNet network, and output as the dense light flow f and the dense de-distortion flow u of the corresponding intermediate scan line m An association factor graph c between, where u m The resolution of (a) is h×w×2, and the resolution of c is h×w×1;
step 3: explicitly calculating a de-distortion flow u corresponding to an intermediate scan line at pixel x using an associated factor value c (x) at pixel x of a kth scan line in a rolling shutter image m (x)：
u m (x)＝c(x)×f(x)
Wherein: c (x) represents the associated factor value of the corresponding pixel x in c, and f (x) represents the optical flow value of the corresponding pixel x in f;
step 4: usingDe-teration flow u corresponding to intermediate scanning line m (x) Explicitly propagated to the corresponding arbitrary scan line s E [1, h ]]Is used for removing teratogen s (x)；
Step 5: repeating the step 3 and the step 4 for h multiplied by w pixels x in the image to obtain the corresponding arbitrary scanning line s epsilon [1, h ]]Dense de-terating flow u s Wherein u is s The resolution of (2) is h×w×2;
step 6: according to the scan line s E [1, h ]]Is thick of (3)Dense de-teration flow u s Transforming the first shutter image using a forward warping technique to recover the scan line s e 1, h]A corresponding global shutter image;
step 7: and (3) sequentially recovering the global shutter image sequences corresponding to the continuous scanning lines s=1, 2, and the number h according to the step (6), and finally outputting the global shutter video with high frame rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210726715.XA CN115278054B (en) | 2022-06-23 | 2022-06-23 | Method for recovering high frame rate global shutter video from rolling shutter image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210726715.XA CN115278054B (en) | 2022-06-23 | 2022-06-23 | Method for recovering high frame rate global shutter video from rolling shutter image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115278054A CN115278054A (en) | 2022-11-01 |
| CN115278054B true CN115278054B (en) | 2023-08-08 |
Family
ID=83762747
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210726715.XA Active CN115278054B (en) | 2022-06-23 | 2022-06-23 | Method for recovering high frame rate global shutter video from rolling shutter image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115278054B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006211426A (en) * | 2005-01-28 | 2006-08-10 | Kyocera Corp | Image sensing device and its image generating method |
| CN104660898A (en) * | 2013-11-25 | 2015-05-27 | 华为终端有限公司 | Global shutter processing method and device |
| CN112995529A (en) * | 2019-12-17 | 2021-06-18 | 华为技术有限公司 | Imaging method and device based on optical flow prediction |
| CN114245007A (en) * | 2021-12-06 | 2022-03-25 | 西北工业大学 | High frame rate video synthesis method, device, equipment and storage medium |
-
2022
- 2022-06-23 CN CN202210726715.XA patent/CN115278054B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006211426A (en) * | 2005-01-28 | 2006-08-10 | Kyocera Corp | Image sensing device and its image generating method |
| CN104660898A (en) * | 2013-11-25 | 2015-05-27 | 华为终端有限公司 | Global shutter processing method and device |
| CN112995529A (en) * | 2019-12-17 | 2021-06-18 | 华为技术有限公司 | Imaging method and device based on optical flow prediction |
| CN114245007A (en) * | 2021-12-06 | 2022-03-25 | 西北工业大学 | High frame rate video synthesis method, device, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| Rolling-Shutter-stereo-aware motion estimation and image correction;Yuchao Dai等;《Science Driect》;第213卷;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115278054A (en) | 2022-11-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Shen et al. | Blurry video frame interpolation | |
| Gu et al. | Coded rolling shutter photography: Flexible space-time sampling | |
| US7844134B2 (en) | Image processor and camera system for correcting image distortion | |
| US9615039B2 (en) | Systems and methods for reducing noise in video streams | |
| CN111986084B (en) | Multi-camera low-illumination image quality enhancement method based on multi-task fusion | |
| CN101872491B (en) | Free view angle relighting method and system based on photometric stereo | |
| US20140055638A1 (en) | Photographing apparatus, method of controlling the same, and computer-readable recording medium | |
| KR102314703B1 (en) | Joint dictionary generation method for image processing, interlace based high dynamic range imaging apparatus using the joint dictionary and image processing method of the same | |
| CN111901532B (en) | Video stabilization method based on recurrent neural network iteration strategy | |
| US20080316352A1 (en) | Focusing apparatus and method | |
| Zhou et al. | Evunroll: Neuromorphic events based rolling shutter image correction | |
| Zhao et al. | Reconstructing clear image for high-speed motion scene with a retina-inspired spike camera | |
| CN113066022B (en) | Video bit enhancement method based on efficient space-time information fusion | |
| CN111986106A (en) | High dynamic image reconstruction method based on neural network | |
| CN112750092A (en) | Training data acquisition method, image quality enhancement model and method and electronic equipment | |
| CN103226816A (en) | Haze image medium transmission rate estimation and optimization method based on quick gaussian filtering | |
| Zou et al. | Rawhdr: High dynamic range image reconstruction from a single raw image | |
| Shen et al. | Spatial temporal video enhancement using alternating exposures | |
| Dong et al. | 3D residual interpolation for spike camera demosaicing | |
| CN116389912B (en) | Method for reconstructing high-frame-rate high-dynamic-range video by fusing pulse camera with common camera | |
| CN115278054B (en) | Method for recovering high frame rate global shutter video from rolling shutter image | |
| CN112435165A (en) | Two-stage video super-resolution reconstruction method based on generation countermeasure network | |
| Dong et al. | Joint Demosaicing and Denoising for Spike Camera | |
| Zhang et al. | Unveiling the potential of spike streams for foreground occlusion removal from densely continuous views | |
| CN115035013A (en) | Image processing method, image processing apparatus, terminal, and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |