WO2012018282A1 - Способ суперразрешения изображений и нелинейный цифровой фильтр для его осуществления - Google Patents
Способ суперразрешения изображений и нелинейный цифровой фильтр для его осуществления Download PDFInfo
- Publication number
- WO2012018282A1 WO2012018282A1 PCT/RU2011/000329 RU2011000329W WO2012018282A1 WO 2012018282 A1 WO2012018282 A1 WO 2012018282A1 RU 2011000329 W RU2011000329 W RU 2011000329W WO 2012018282 A1 WO2012018282 A1 WO 2012018282A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- neural network
- frames
- resolution
- images
- Prior art date
Links
- 238000004519 manufacturing process Methods 0.000 title 1
- 238000013528 artificial neural network Methods 0.000 claims abstract description 51
- 238000001914 filtration Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000012986 modification Methods 0.000 claims abstract description 10
- 230000004048 modification Effects 0.000 claims abstract description 10
- 238000000034 method Methods 0.000 claims description 43
- 238000010606 normalization Methods 0.000 claims description 5
- 230000015572 biosynthetic process Effects 0.000 claims 1
- 238000002955 isolation Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 description 10
- 238000012549 training Methods 0.000 description 10
- 230000006641 stabilisation Effects 0.000 description 5
- 238000011105 stabilization Methods 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000012935 Averaging Methods 0.000 description 2
- 101000633607 Bos taurus Thrombospondin-2 Proteins 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 238000001303 quality assessment method Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- ALPWRKFXEOAUDR-GKEJWYBXSA-M sodium;[(2r)-2,3-di(octadecanoyloxy)propyl] hydrogen phosphate Chemical compound [Na+].CCCCCCCCCCCCCCCCCC(=O)OC[C@H](COP(O)([O-])=O)OC(=O)CCCCCCCCCCCCCCCCC ALPWRKFXEOAUDR-GKEJWYBXSA-M 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000012897 Levenberg–Marquardt algorithm Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/30—Transforming light or analogous information into electric information
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/40—Extracting pixel data from image sensors by controlling scanning circuits, e.g. by modifying the number of pixels sampled or to be sampled
- H04N25/44—Extracting pixel data from image sensors by controlling scanning circuits, e.g. by modifying the number of pixels sampled or to be sampled by partially reading an SSIS array
- H04N25/443—Extracting pixel data from image sensors by controlling scanning circuits, e.g. by modifying the number of pixels sampled or to be sampled by partially reading an SSIS array by reading pixels from selected 2D regions of the array, e.g. for windowing or digital zooming
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/48—Increasing resolution by shifting the sensor relative to the scene
Definitions
- the invention relates to the field of photo and video images and can be used, for example, to obtain high-quality images of visually close objects with a camera or video camera equipped with sensors with an electronic shutter.
- Modern mobile devices as a rule, are equipped with photo and video cameras, which allow receiving images of good quality.
- the mobility requirement of these devices does not allow the use of optical systems (lenses) with variable focal length (zoom, zoom, zoom lens) in view of their large size. Therefore, such mobile devices use digital zoom.
- a known method of obtaining an enlarged image with low resolution using digital zoom When using this method of visual approximation, only the central part of the sensor is used to obtain an image. Then, to obtain an image with the number of pixels equal to the total number of pixels of the sensor, the obtained reduced image from the central part of the sensor is interpolated using one of the known methods of two-dimensional interpolation (bilinear or bicubic). [en.wikipedia.org/wiki/Digital_zoom].
- the interpolated signal does not contain high-frequency components, as a result the faces look fuzzy, there are no small details.
- SUBSTITUTE SHEET (RULE 26) A known method of image enhancement [Michal Irani, Shmuel Peleg "Super Resolution From Image Sequences", ICPR, 2: 115-120, June 1990], using several frames and the presence of small spatial shifts between them to increase resolution or obtain super resolution. In this method, iteratively approximates (converges) to an optimal high-resolution image. Iterations begin with the creation of the primary (rough) version of the high-resolution image. Such a primary version is created, as a rule, by simply adding up the interpolated low-resolution images. The second stage of the iteration is the reverse acquisition of low-resolution images from this high-resolution version, comparing them with the original low-resolution images and calculating the correction. Subsequent iterations compute new versions of the high-resolution image, adjusted for the previous iteration.
- SUBSTITUTE SHEET (RULE 26) temporal IIR filter ", IEEE Transactions on Consumer Electronics, Vol. 42, No.3, August 1996], which carry more information than a single two-dimensional image.
- Subpixel level motion detection and time-axis IIR filtering for visual image enlargement are proposed with high resolution, as well as for digital image stabilization.
- Experimental results on real image sequences are demonstrated.
- the method is carried out as follows: they collect data from the sensor, align, enlarge the image, combine / filter the enlarged image with a linear filter, and the next incoming frame is summed with the previous result using different weights. After enlarging the image, but before combining, an additional convolution is made with a rectangular window, i.e. additional filtering so that the subpixel-shifted image can be directly summed (filtered by a 1IR filter) with the pixels of the previous result.
- the disadvantages of the prototype are, firstly, that the simplicity of the filter used at the output does not allow to obtain an optimally clear resulting image.
- the filter does not use the area of adjacent, neighboring pixels of the image, which does not allow to correct distortions (blurring / blurring) of the camera’s optical system. Data is collected from the sensor in a standard, low-speed, manner, which leads to blurring of the image, as well as to the doubling of unsteady objects.
- SUBSTITUTE SHEET (RULE 26) - the values of the selected pixels are transmitted to the input of a pre-trained neural network
- the neural network directly gives the pixel value to form a clear image
- the neural network produces a signal with a level proportional to the likelihood of text in this area of the image.
- a nonlinear digital filter is used for image areas with faces constructed using a neural network.
- the neural network is pre-trained using 'field', natural images.
- the input data for the interpolator are the coordinates of the region, the 'quality' of the face, the slope of the face, the value of the processed pixel and its neighboring pixels.
- the 'quality' of the face and the slope of the face are calculated based on the pixel data included in the area.
- This data is transmitted to the inputs of the neural network.
- the neural network multiplies the input data by the weights determined at the stage of preliminary training of the neural network,
- the result of the neural network is the value of the interpolated pixel.
- the neural network acts as a nonlinear filter, the coordinates of the region, the 'quality' of the face, the slope of the face, the value of the processed pixel and its neighboring pixels are transmitted directly to the inputs.
- the neural network directly yields the value of the interpolated pixel.
- the neural network is trained to recognize a predetermined, limited set of patterns (face orientation options), which entails incorrect interpolation of images that are not similar to images from a natural set;
- the aim of the invention is to create a method that would allow to obtain visually approximate images of high quality and resolution when photographing and filming both stationary and moving objects.
- the proposed technical solution is based on the application of a method for increasing the resolution of an image using several low-resolution frames to obtain one high-resolution frame (i.e., super resolution), as well as on the possibility of high-speed shooting of several image frames when scanning only part of the sensor. Moreover, the solution of the problem, i.e. achievement of the required technical effect is achieved through the use of a non-linear filter designed for this purpose.
- the essence of the claimed invention lies in the fact that for superresolution of images, in a known method of improving images, including
- SUBSTITUTE SHEET (RULE 26) exposure of several frames, obtaining source images by reading from the sensor, aligning them, generating an enlarged image, filtering it with a filter, the original image is obtained from a digital sensor in the form of a continuous sequence of frames with high-speed shooting, in which the frame rate is inversely proportional to the size of the scanned part of the photosensitive area of the sensor.
- an enlarged image is formed by combining the original low-resolution images, revealing the clearest frames, and increasing the resolution is carried out by a non-linear filter, applying it to the enlarged image.
- a non-linear digital filter is used, the input for which is the pixels of the processed image, including a neural network pre-trained using a test image.
- Modified digitized data is supplied to the neural network, and their modification includes: selection of the low-frequency component, pixel-by-pixel arrangement of pixels, subtraction of the low-frequency component from the arranged pixels, and their subsequent normalization. Then, the data at the output of the neural network is subjected to reverse normalization, and the low-frequency component is added to the value at the output of the neural network.
- a filter value only the data of the pixel to be filtered and pixels spaced no more than 3 points in the horizontal and vertical directions from the enlarged image are used.
- Figure 1 shows the standard case of shooting frames in video mode using a sensor with an electronic shutter. There are pauses between frame exposures. Vertical lines 1; 2; 3 on the timeline indicate the time the exposure begins, with the distance between them corresponding to the frame rate. The shaded areas are the actual exposure time of the sensor lines (since a sensor with an electronic shutter is used, the actual time of the beginning and end of the exposure of individual lines is somewhat shifted in time).
- Figure 2 shooting frames without pauses using a sensor with an electronic shutter.
- Vertical lines 1; 2; 3; 4 on the timeline indicate the time the exposure began. The exposure of the next frame begins immediately after reading the data of the line of the current frame and there is no pause.
- Fig.Z is a sensor circuit, where 301 is the photosensitive region of the sensor; 302 - him the central part used to obtain a visually close image.
- Figure 4 shows the alignment of several frames and their combination into a single image of an enlarged size with a subsequent increase in resolution (block diagram), where:
- Figure 5 shows the use of a nonlinear filter to increase resolution, where:
- 502 is a diagram of a filter including preliminary data preparation (503); neural network (504) and summation (505);
- 506 is the final image with super resolution.
- 6 is a test image used in the preliminary training stage of the neural network.
- the maximum speed of receiving data from the sensor in modern mobile devices is limited by the maximum possible speed of the data transfer interface, and when shooting visually close images there is no need to scan the entire surface of the sensor, when scanning only part of the sensor, it is possible to proportionally increase the frame rate. Expose several frames at a fixed frequency and shutter speed, with a frame rate that eliminates pauses between exposures. In the case of poor illumination of the object, it is possible either to increase the shutter speed of each frame, or, more preferably, to increase the number of exposed frames.
- the maximum frame rate can be increased by 9 times, which means that shooting, for example, nine frames takes the same time
- REPLACE ITS SHEET RULE 26 as shooting a single frame using the standard method. With such a survey, both stationary and moving objects in each frame will be clear, and the noise level will be lower than when shooting by similar known methods, because when combined, the amplitude of the noise component of individual frames increases in proportion to the square root of the total number of frames, and the amplitude of the useful component (the image itself) increases in proportion to the number of frames.
- the next processing step is the identification of the clearest frames, as well as their alignment relative to each other.
- frame clarity There are many ways to assess frame clarity, for example, the method described in [Xin Wang, Baofeng Tian, Chao Liang, Dongcheng Shi “Blind Image Quality Assessment for Measuring Image Blur", Congress on Image and Signal Processing, 2008. CISP ⁇ 8. Volume: 1, ISBN: 978-0-7695-3119-9], where a subset of the sharpest faces is selected to determine the clarity of the frame, the average clarity of the selected faces is determined, and the average value is used as the clarity metric of the entire frame.
- one enlarged frame (404) is used, obtained by aligning and combining the frames of the previous step (Fig. 4).
- the increase in the frame is carried out by interpolation (for example, bicubic).
- the combination of frames is carried out either by simple averaging of the values of the coincident pixels of the enlarged frames, or by more complex summation using weight coefficients. For example, in the case of the presence of moving objects, selective averaging of frames is performed with the selection of data from those frames in which the position (location) of moving objects coincides.
- the combination is performed with more weight being given to those frames in which the noise level is lower in order to lower the overall noise level in the combined frame.
- a super-resolution image is obtained by applying, successively to each pixel, an enlarged frame, a non-linear filter.
- the filter uses the region of pixels located in the immediate vicinity of the pixel for which the resolution is increased (Fig. 5).
- SUBSTITUTE SHEET (RULE 26) They are adapted to a specific lens-sensor system. This provides the maximum increase in resolution for this particular system, as well as optimal noise reduction.
- the filter is constructed using an artificial neural network.
- various types of neural networks can be used.
- a nonlinear multilayer perceptron was used [Medvedev BC, Potemkin VG, "Neural networks. MATLAB 6", M: Dialog-MEPhI, 2002, chapter 4]. It was experimentally established that the most optimal neural network architecture for the task is a perceptron with one hidden layer, sigmoidal or tangential activation functions in all layers, four neurons in the input layer, four neurons in the hidden layer.
- the filter is applied to a monochromatic image or only to the luminance component of the image, one neuron is used in the output layer. If the filter is applied to a multi-color image, the output layer may contain the number of neurons equal to the number of color layers of the image, or a separate filter can be applied to each color layer independently.
- Non-linear filter includes:
- the preliminary modification of digitized data consists of the following steps:
- Low-frequency separation Produced using a conventional linear filter (for example, calculating the average value of all pixels within a radius of 8 pixels from a given one). Low-frequency separation reduces the dynamic range of input data.
- the neural network operates in the most favorable mode, if the dynamic range of the input values is small, rationing allows you to further reduce the dynamic range. All arrays of pixels are normalized so that the pixel values fall in a certain range (for example, [0..1]).
- Modified in this way data is transmitted to the input of the neural network. They are used both when training a neural network, and when using a neural network as part of a non-linear filter.
- a test image is used (Fig. 6), specially prepared for this purpose, captured by the lens-sensor system, for which the filter will be used.
- neural networks have the ability to 'generalize' (i.e., to derive some general rules and dependencies based on a limited set of data), then there is no need to use all possible image options at the stage of training the neural network.
- the image used during training should meet the minimum requirements in order to get a network that works well enough for all images as a result of training.
- the image must include:
- the shots of the test image as well as their alignment and association are made by the claimed method using reference points (crosshairs) to simplify the alignment of frames of the captured image.
- reference points crosshairs
- the normalized pixels of the test image are used according to the coordinates, from which the high and low frequencies are pre-filtered.
- the cutoff frequency of the high frequencies is chosen experimentally, based on the requirements for the clarity of the final image and the acceptable level of noise / distortion in it.
- the cutoff frequency of the low frequencies is chosen equal to the frequency of the low-pass filter used to modify the input digitized data.
- the Levenberg – Marquardt algorithm [7] is used, which gives the best results for medium and small sized neural networks.
- step 5 of the preliminary modification After receiving the data at the output of the neural network, they are reverse normalized. For example, if in step 5 of the preliminary modification, rationing was carried out by simple multiplication by a constant, then inverse rationing is performed by dividing the data obtained from the output of the neural network by the same constant.
- the non-linear filter data processing procedure consists of:
- the proposed method of superresolution of images and a non-linear digital filter for its implementation makes it possible to obtain high-quality images with high resolution and is applicable when using various kinds of mobile devices currently manufactured by the industry.
- the post-processing of data received from the sensor necessary for obtaining high resolution, imposes low requirements on the computing resources of the device and can be
- SUBSTITUTE SHEET (RULE 26) 13. Masaaki Hayashi, "Neurofilter, and method of training same to operate on image data such as to discriminate between text and picture regions jf an image which is expressed by image data" United States Patent 6,301, 381
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Studio Devices (AREA)
- Image Analysis (AREA)
Abstract
Description
Claims
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE112011102647T DE112011102647T5 (de) | 2010-08-06 | 2011-05-16 | Verfahren zur Herstellung von super-auflösenden Bildern und nicht-linearer Filter zu dessen Implementierung |
KR1020137004968A KR101612165B1 (ko) | 2010-08-06 | 2011-05-16 | 초고해상도 이미지 생성 방법 및 이를 구현하기 위한 비선형 디지털 필터 |
JP2013523122A JP5784723B2 (ja) | 2010-08-06 | 2011-05-16 | 超解像度画像を生成する方法及びこれを実施するための非線形デジタルフィルター |
US13/814,664 US9020302B2 (en) | 2010-08-06 | 2011-05-16 | Method for producing super-resolution images and nonlinear digital filter for implementing same |
CN201180037162.7A CN103098089B (zh) | 2010-08-06 | 2011-05-16 | 图像超分辨率的方法 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
RU2010133219/09A RU2431889C1 (ru) | 2010-08-06 | 2010-08-06 | Способ суперразрешения изображений и нелинейный цифровой фильтр для его осуществления |
RU2010133219 | 2010-08-06 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012018282A1 true WO2012018282A1 (ru) | 2012-02-09 |
Family
ID=44999290
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/RU2011/000329 WO2012018282A1 (ru) | 2010-08-06 | 2011-05-16 | Способ суперразрешения изображений и нелинейный цифровой фильтр для его осуществления |
Country Status (7)
Country | Link |
---|---|
US (1) | US9020302B2 (ru) |
JP (1) | JP5784723B2 (ru) |
KR (1) | KR101612165B1 (ru) |
CN (1) | CN103098089B (ru) |
DE (1) | DE112011102647T5 (ru) |
RU (1) | RU2431889C1 (ru) |
WO (1) | WO2012018282A1 (ru) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019207350A1 (ru) | 2018-04-28 | 2019-10-31 | Алмаленс Инк. | Оптическая система гибридной реальности с цифровой коррекцией аберраций |
US10805634B2 (en) | 2017-05-17 | 2020-10-13 | Samsung Electronics Co., Ltd | Super-resolution processing method for moving image and image processing apparatus therefor |
CN112345251A (zh) * | 2020-11-04 | 2021-02-09 | 山东科技大学 | 基于信号分辨率增强的机械智能故障诊断方法 |
Families Citing this family (32)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2538941C1 (ru) * | 2013-06-14 | 2015-01-10 | Общество с ограниченной ответственностью "Аби Девелопмент" | Улучшения качества распознавания за счет повышения разрешения изображений |
RU2549353C1 (ru) * | 2014-02-03 | 2015-04-27 | Федеральное государственное унитарное предприятие "Государственный научно-исследовательский институт авиационных систем" | Способ повышения разрешающей способности видеосистем |
KR101485512B1 (ko) * | 2014-03-28 | 2015-01-23 | 주식회사 성우음향정보통신 | 객체 행동패턴의 해마 신경망 학습을 이용한 미래범죄 발생확률 우선순위 이벤트 cctv 영상처리 방법 |
RU2583725C1 (ru) * | 2014-10-17 | 2016-05-10 | Самсунг Электроникс Ко., Лтд. | Способ и система для обработки изображения |
US10007970B2 (en) | 2015-05-15 | 2018-06-26 | Samsung Electronics Co., Ltd. | Image up-sampling with relative edge growth rate priors |
US9741107B2 (en) * | 2015-06-05 | 2017-08-22 | Sony Corporation | Full reference image quality assessment based on convolutional neural network |
US9582762B1 (en) | 2016-02-05 | 2017-02-28 | Jasmin Cosic | Devices, systems, and methods for learning and using artificially intelligent interactive memories |
RU2664540C2 (ru) * | 2016-05-04 | 2018-08-20 | Федеральное государственное казенное военное образовательное учреждение высшего образования "Военный учебно-научный центр Военно-воздушных сил "Военно-воздушная академия имени профессора Н.Е. Жуковского и Ю.А. Гагарина" (г. Воронеж) Министерства обороны Российской Федерации | Способ повышения разрешения изображения |
RU2635883C1 (ru) * | 2016-06-02 | 2017-11-16 | Самсунг Электроникс Ко., Лтд. | Способ и система обработки изображений для формирования изображений сверхвысокого разрешения |
US10366480B2 (en) | 2016-07-01 | 2019-07-30 | Analytical Mechanics Associates, Inc. | Super-resolution systems and methods |
US10554877B2 (en) | 2016-07-29 | 2020-02-04 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Image synthesis method and apparatus for mobile terminal, and mobile terminal |
CN106101556B (zh) | 2016-07-29 | 2017-10-20 | 广东欧珀移动通信有限公司 | 移动终端的图像合成方法、装置及移动终端 |
US10380767B2 (en) * | 2016-08-01 | 2019-08-13 | Cognex Corporation | System and method for automatic selection of 3D alignment algorithms in a vision system |
US9864933B1 (en) | 2016-08-23 | 2018-01-09 | Jasmin Cosic | Artificially intelligent systems, devices, and methods for learning and/or using visual surrounding for autonomous object operation |
RU2661537C2 (ru) * | 2016-08-30 | 2018-07-17 | Самсунг Электроникс Ко., Лтд. | Способ и система сверхразрешения с помощью комбинированной разреженной аппроксимации |
US10452974B1 (en) | 2016-11-02 | 2019-10-22 | Jasmin Cosic | Artificially intelligent systems, devices, and methods for learning and/or using a device's circumstances for autonomous device operation |
JP6929047B2 (ja) * | 2016-11-24 | 2021-09-01 | キヤノン株式会社 | 画像処理装置、情報処理方法及びプログラム |
US11430199B2 (en) | 2016-12-09 | 2022-08-30 | Google Llc | Feature recognition assisted super-resolution method |
US10607134B1 (en) | 2016-12-19 | 2020-03-31 | Jasmin Cosic | Artificially intelligent systems, devices, and methods for learning and/or using an avatar's circumstances for autonomous avatar operation |
JP6236731B1 (ja) * | 2017-03-10 | 2017-11-29 | 株式会社ツバサファクトリー | 超解像度処理装置、超解像度処理方法およびコンピュータプログラム |
TWI754764B (zh) * | 2017-06-30 | 2022-02-11 | 美商克萊譚克公司 | 由低解析度影像產生高解析度影像以用於半導體應用 |
US10102449B1 (en) | 2017-11-21 | 2018-10-16 | Jasmin Cosic | Devices, systems, and methods for use in automation |
US10474934B1 (en) | 2017-11-26 | 2019-11-12 | Jasmin Cosic | Machine learning for computing enabled systems and/or devices |
US10402731B1 (en) | 2017-12-15 | 2019-09-03 | Jasmin Cosic | Machine learning for computer generated objects and/or applications |
KR20200003444A (ko) | 2018-07-02 | 2020-01-10 | 삼성전자주식회사 | 영상 모델 구축 장치 및 방법 |
KR102614908B1 (ko) * | 2018-09-13 | 2023-12-18 | 삼성전자주식회사 | 이미지 복원 방법 및 장치 |
US11030724B2 (en) | 2018-09-13 | 2021-06-08 | Samsung Electronics Co., Ltd. | Method and apparatus for restoring image |
KR102604016B1 (ko) | 2018-10-24 | 2023-11-22 | 삼성전자주식회사 | 전자 장치 및 이의 제어방법 |
KR102545667B1 (ko) * | 2018-10-30 | 2023-06-21 | 삼성전자주식회사 | 인공 신경망을 이용한 이미지 처리 방법 및 이를 지원하는 전자 장치 |
KR20200142883A (ko) | 2019-06-13 | 2020-12-23 | 엘지이노텍 주식회사 | 카메라 장치 및 카메라 장치의 이미지 생성 방법 |
CN114708144B (zh) * | 2022-03-16 | 2023-05-26 | 荣耀终端有限公司 | 图像数据处理方法及装置 |
CN115908142B (zh) * | 2023-01-06 | 2023-05-09 | 诺比侃人工智能科技(成都)股份有限公司 | 一种基于视觉识别的接触网微小部件验损方法 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5262871A (en) * | 1989-11-13 | 1993-11-16 | Rutgers, The State University | Multiple resolution image sensor |
JPH08263649A (ja) * | 1995-02-21 | 1996-10-11 | Xerox Corp | ニューラルベースの空間画像強化システム |
WO2002005213A2 (en) * | 2000-07-12 | 2002-01-17 | Koninklijke Philips Electronics N.V. | Method and apparatus for digital image defect correction and noise filtering |
US20090129704A1 (en) * | 2006-05-31 | 2009-05-21 | Nec Corporation | Method, apparatus and program for enhancement of image resolution |
JP2009194896A (ja) * | 2008-01-18 | 2009-08-27 | Sanyo Electric Co Ltd | 画像処理装置及び方法並びに撮像装置 |
Family Cites Families (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH06348840A (ja) * | 1993-06-03 | 1994-12-22 | Konica Corp | 画像復元方法 |
JPH08305846A (ja) | 1995-03-07 | 1996-11-22 | Matsushita Electric Ind Co Ltd | ニューロフィルターと像域分割方法とフィルター装置 |
US7085436B2 (en) * | 2001-08-28 | 2006-08-01 | Visioprime | Image enhancement and data loss recovery using wavelet transforms |
US7187811B2 (en) | 2003-03-18 | 2007-03-06 | Advanced & Wise Technology Corp. | Method for image resolution enhancement |
FR2864631B1 (fr) * | 2003-12-31 | 2006-04-14 | Mauna Kea Technologies | Procede et systeme de super-resolution d'images confocales acquises a travers un guide d'image, et dispositif utilise pour la mise en oeuvre d'un tel procede |
KR100555866B1 (ko) * | 2004-06-08 | 2006-03-03 | 삼성전자주식회사 | 패턴 적응형 필터링에 의한 영상신호의 평활화 장치 및 그평활화 방법 |
JP4184319B2 (ja) * | 2004-06-10 | 2008-11-19 | オリンパス株式会社 | 撮像装置 |
US20060291750A1 (en) * | 2004-12-16 | 2006-12-28 | Peyman Milanfar | Dynamic reconstruction of high resolution video from low-resolution color-filtered video (video-to-video super-resolution) |
US7412107B2 (en) * | 2004-12-17 | 2008-08-12 | The Regents Of The University Of California, Santa Cruz | System and method for robust multi-frame demosaicing and color super-resolution |
WO2008133951A2 (en) * | 2007-04-24 | 2008-11-06 | Massachusetts Institute Of Technology | Method and apparatus for image processing |
JP4886583B2 (ja) * | 2007-04-26 | 2012-02-29 | 株式会社東芝 | 画像拡大装置および方法 |
JP5040449B2 (ja) * | 2007-05-31 | 2012-10-03 | 富士通セミコンダクター株式会社 | 固体撮像素子および固体撮像素子を用いた信号処理方法 |
US8743963B2 (en) * | 2007-08-13 | 2014-06-03 | Ntt Docomo, Inc. | Image/video quality enhancement and super-resolution using sparse transformations |
JP2009273072A (ja) * | 2008-05-12 | 2009-11-19 | Hitachi Ltd | 高精細画像処理装置、高精細画像処理方法、およびプログラム |
JP5156982B2 (ja) * | 2008-06-05 | 2013-03-06 | 富士フイルム株式会社 | 画像処理システム、画像処理方法、およびプログラム |
US8374444B2 (en) * | 2008-08-06 | 2013-02-12 | Sony Corporation | Method and apparatus for providing higher resolution images in an embedded device |
US8233747B2 (en) * | 2009-04-03 | 2012-07-31 | Sony Corporation | Method and apparatus for forming super resolution images from raw data representative of color filter array images |
-
2010
- 2010-08-06 RU RU2010133219/09A patent/RU2431889C1/ru active
-
2011
- 2011-05-16 CN CN201180037162.7A patent/CN103098089B/zh active Active
- 2011-05-16 US US13/814,664 patent/US9020302B2/en active Active
- 2011-05-16 DE DE112011102647T patent/DE112011102647T5/de active Pending
- 2011-05-16 KR KR1020137004968A patent/KR101612165B1/ko active IP Right Grant
- 2011-05-16 JP JP2013523122A patent/JP5784723B2/ja active Active
- 2011-05-16 WO PCT/RU2011/000329 patent/WO2012018282A1/ru active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5262871A (en) * | 1989-11-13 | 1993-11-16 | Rutgers, The State University | Multiple resolution image sensor |
JPH08263649A (ja) * | 1995-02-21 | 1996-10-11 | Xerox Corp | ニューラルベースの空間画像強化システム |
WO2002005213A2 (en) * | 2000-07-12 | 2002-01-17 | Koninklijke Philips Electronics N.V. | Method and apparatus for digital image defect correction and noise filtering |
US20090129704A1 (en) * | 2006-05-31 | 2009-05-21 | Nec Corporation | Method, apparatus and program for enhancement of image resolution |
JP2009194896A (ja) * | 2008-01-18 | 2009-08-27 | Sanyo Electric Co Ltd | 画像処理装置及び方法並びに撮像装置 |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10805634B2 (en) | 2017-05-17 | 2020-10-13 | Samsung Electronics Co., Ltd | Super-resolution processing method for moving image and image processing apparatus therefor |
WO2019207350A1 (ru) | 2018-04-28 | 2019-10-31 | Алмаленс Инк. | Оптическая система гибридной реальности с цифровой коррекцией аберраций |
US11372243B2 (en) | 2018-04-28 | 2022-06-28 | Almalence, Inc. | Optical hybrid reality system having digital correction of aberrations |
CN112345251A (zh) * | 2020-11-04 | 2021-02-09 | 山东科技大学 | 基于信号分辨率增强的机械智能故障诊断方法 |
CN112345251B (zh) * | 2020-11-04 | 2022-03-04 | 山东科技大学 | 基于信号分辨率增强的机械智能故障诊断方法 |
Also Published As
Publication number | Publication date |
---|---|
JP5784723B2 (ja) | 2015-09-24 |
US9020302B2 (en) | 2015-04-28 |
KR101612165B1 (ko) | 2016-04-12 |
DE112011102647T5 (de) | 2013-07-11 |
CN103098089B (zh) | 2017-03-29 |
KR20130102550A (ko) | 2013-09-17 |
CN103098089A (zh) | 2013-05-08 |
RU2431889C1 (ru) | 2011-10-20 |
US20130156345A1 (en) | 2013-06-20 |
JP2013532878A (ja) | 2013-08-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
RU2431889C1 (ru) | Способ суперразрешения изображений и нелинейный цифровой фильтр для его осуществления | |
Wronski et al. | Handheld multi-frame super-resolution | |
JP6957197B2 (ja) | 画像処理装置および画像処理方法 | |
JP5909540B2 (ja) | 画像処理表示装置 | |
Delbracio et al. | Burst deblurring: Removing camera shake through fourier burst accumulation | |
CN1870715B (zh) | 手抖动校正装置 | |
JP5220777B2 (ja) | 画像処理装置、撮像装置、プログラム及び画像処理方法 | |
US10282822B2 (en) | Digital correction of optical system aberrations | |
JP5341010B2 (ja) | 画像処理装置、撮像装置、プログラム及び画像処理方法 | |
Delbracio et al. | Removing camera shake via weighted fourier burst accumulation | |
JP2003078812A (ja) | デジタル画像における動きによるぼけを低減する方法 | |
KR101813292B1 (ko) | 이미지 개선 방법 | |
Chang et al. | Beyond camera motion blur removing: How to handle outliers in deblurring | |
Wang et al. | High-quality image deblurring with panchromatic pixels. | |
US20080002909A1 (en) | Reconstructing Blurred High Resolution Images | |
JP4423537B2 (ja) | 信号処理装置および信号処理方法、並びにプログラムおよび記録媒体 | |
Bhagat et al. | Novel Approach to Estimate Motion Blur Kernel Parameters and Comparative Study of Restoration Techniques | |
Soulez et al. | Joint deconvolution and demosaicing | |
JP6611509B2 (ja) | 画像処理装置、撮像装置および画像処理プログラム | |
JP6818461B2 (ja) | 撮像装置、画像処理装置、画像処理方法および画像処理プログラム | |
Lim et al. | Image resolution and performance analysis of webcams for ground-based astronomy | |
US20230041139A1 (en) | Image processing for oversampled infrared imaging | |
Eboli | Hybrid Non-blind Image Deblurring for Real Scenarios | |
KRAININ et al. | Handheld Multi-Frame Super-Resolution | |
Park | Super resolution image reconstruction using averaged image and regularized deconvolution |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201180037162.7 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11814859 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2013523122 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13814664 Country of ref document: US Ref document number: 1120111026471 Country of ref document: DE Ref document number: 112011102647 Country of ref document: DE |
|
ENP | Entry into the national phase |
Ref document number: 20137004968 Country of ref document: KR Kind code of ref document: A |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 11814859 Country of ref document: EP Kind code of ref document: A1 |