Abstract: A model-driven deep learning-based seismic super-resolution inversion method includes the following steps: 1) mapping each iteration of a model-driven alternating direction method of multipliers (ADMM) into each layer of a deep network, and learning proximal operators by using a data-driven method to complete the construction of a deep network ADMM-SRINet; 2) obtaining label data used to train the deep network ADMM-SRINet; 3) training the deep network ADMM-SRINet by using the obtained label data; and 4) inverting test data by using the deep network ADMM-SRINet trained at step 3). The method combines the advantages of a model-driven optimization method and a data-driven deep learning method, and therefore the network has the interpretability; and meanwhile, due to the addition of physical knowledge, the iterative deep learning method lowers requirements for a training set, and therefore an inversion result is more reliable.

Abstract: An x-ray system includes an x-ray source, at least one partially transparent x-ray filter, an x-ray detector, a monitor to display x-ray images detected by the detector and image processing. The display includes a dynamic range. The system is configured to generate at least one x-ray image and modify at least one of the at least one image for display by: using the at least one filter to filter x-ray so as to reduce x-ray intensity in at least one part of the image; maintaining at least one part of the image unfiltered by the at least one filter; determining a range in the dynamic range of the display; and modifying at least one pixel in the at least one filtered part of the image based on the determined range of the dynamic range of the display.

Abstract: A method and an apparatus for processing an image are provided. A method may include: acquiring a to-be-processed foggy image, and performing minimum value filtering on the to-be-processed foggy image to obtain a grayscale image of the to-be-processed foggy image; selecting, in a descending order of pixel values, a target number of pixel points from the grayscale image, and determining an image area corresponding to the selected pixel points from the to-be-processed foggy image; acquiring pixel values of the pixel points included in the image area, and determining an atmospheric light value based on the acquired pixel values; and performing image defogging processing on the to-be-processed foggy image based on the grayscale image and the atmospheric light value to obtain a processed image. The effect of image display and the efficiency of image processing can be improved.

Abstract: An image processing method includes obtaining an original image; partitioning the original image into a first part and a second part such that distortion of at least a part of an image in the first part of the original image is smaller than a predetermined threshold, and distortion of at least a part of an image in the second part of the original image is greater than or equal to the predetermined threshold; correcting the second part of the original image so as to obtain a distortion-corrected image corresponding to the second part; and recognizing the first part of the original image and the distortion-corrected image so as to recognize an object in the original image.

Abstract: An apparatus includes a detection unit configured to detect a contrast from a plurality of images of a focus position, wherein viewing angles of the plurality of images overlaps at least partially, an adjustment unit configured to adjust the contrast, a setting unit configured to set a combining ratio based on the adjusted contrast, and a combining unit configured to combine the plurality of images based on the set combining ratio to generate a combined image, wherein the adjustment unit spreads the contrast of the image in which a subject is brought into focus, among the plurality of images, and the setting unit sets, based on the contrast of a corresponding position in the plurality of images, the combining ratio for the combining to the corresponding position.

Abstract: A blur correction device includes a processor including hardware, and the processor obtains an object image from an imaging section, sets any one of a first region where a blur correction is not applied and a second region where the blur correction is applied, based on the object image, finds a third region representing a result of the blur correction applied to the second region, and combines the third region and the first region.

Abstract: In an example embodiment, a first DCNN is trained to output a value for a first metric by inputting a plurality of sample documents to the first DCNN, with each of the sample documents having been labeled with a value for the first metric. Then a plurality of possible transformations of a first input document are fed to the first DCNN, obtaining a value for the first metric for each of the plurality of possible transformations. A first transformation is selected from the plurality of possible transformations based on the values for the first metric for each of the plurality of possible transformations. Then a second DCNN is trained to output a transformation for a document by inputting the selected first transformation to the second DCNN. The second input document is fed to the second DCNN, obtaining a second transformation of the second input document.

Abstract: Provided are a system and method for correcting an image through estimation of a distortion parameter. The method includes receiving a distorted image including one or more measurement targets, extracting a plurality of feature points from each of the measurement targets, classifying the one or more measurement targets as a distorted target and an undistorted target by comparing distances between the plurality of extracted feature points and a center point of the received distorted image with each other, estimating a distortion parameter on the basis of standard deviations of a plurality of feature points of the classified distorted target and undistorted target, and correcting the received distorted image on the basis of the estimated distortion parameter.

Abstract: The present disclosure related to techniques for extracting a respiratory navigation signal for use in a magnetic resonance imaging system. The extracted respiratory navigation signals accurately represent respiratory motions of a patient.

Abstract: An image inspection device includes a mounting unit on which an image projection device that directly projects an image on a retina of a user is to be mounted; a condensing lens that condenses a light beam emitted from the image projection device mounted on the mounting unit; a detector on which an inspection image is projected and detected; and a controller that inspects the inspection image detected by the detector. The image inspection device inspects the image projected by the image projection device, which directly projects the image on the retina of the user.

Abstract: An apparatus for video coding using dual deblocking filter thresholds may include a processor generating a reconstructed frame by decoding an encoded bitstream and outputting the reconstructed frame. Decoding may include generating a decoded block by decoding a portion of the encoded bitstream, identifying a first deblocking threshold index from the encoded bitstream, identifying a second deblocking threshold index from the encoded bitstream, generating a reconstructed block based on the decoded block, and including the reconstructed block in the reconstructed frame. Generating the reconstructed block may include deblocking based on the first deblocking threshold index and the second deblocking threshold index.

Abstract: Electronic devices, methods, and program storage devices for leveraging machine learning to perform improved image fusion and/or noise reduction are disclosed. An incoming image stream may be obtained from an image capture device, wherein the incoming image stream comprises a variety of differently-exposed captures, e.g., EV0 images, EV? images, EV+ images, long exposure images, EV0/EV? image pairs, etc., which are received according to a particular pattern. When a capture request is received, two or more intermediate assets may be generated based on determined combinations of images from the incoming image stream, and the intermediate assets may then be fed into a neural network that has been trained to determine one or more sets of parameters to optimally fuse and/or noise reduce the intermediate assets. In some embodiments, the network may be trained to operate on levels of pyramidal decompositions of the intermediate assets independently, for increased efficiency and memory utilization.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving mission specific data. Converting the mission specific data into attribute classes recognizable by image recognition classifiers, where the image recognition classifiers are pre-trained to detect objects corresponding to the attributed classes within digital images. Obtaining, from a wearable camera system, low resolution images of a scene and high resolution images of the scene. Detecting, within the low resolution images using a low resolution image classifier, an object that corresponds to one of the attribute classes. In response to detecting the object that corresponds to one of the attribute classes, providing, for presentation to a user of the wearable camera system by a presentation system, an first alert indicating a potential detection of the suspect, and providing the high resolution images to a high resolution image classifier.

Abstract: The present invention relates to a method for decoding a bitstream for a video signal and a device therefor, the method comprising the steps of: acquiring a value indicating the similarity between reference samples for intra prediction of a current block; determining an intra prediction mode of the current block on the basis of the acquired value and a predetermined threshold value; and reconstructing the current block on the basis of the determined intra prediction mode, wherein the step of determining the intra prediction mode of the current block comprises: when the value indicating the similarity is smaller than the predetermined threshold value, acquiring index information from the bitstream; preparing a list of candidate modes on the basis of intra prediction modes of blocks neighboring the current block, without acquiring flag information indicating whether the intra prediction mode of the current block is derived from the blocks neighboring the current block; and determining, as the intra prediction m

Abstract: The disclosure analyzes moon-earth relationship and influencing factors for geometric distortion of a moon-based earth image, focuses on influences of position change of a sublunar point, a curvature of the earth, a terrain fluctuation on a large-scale hemisphere image, and the problem of small number of ground control points and uneven distribution thereof, and proposes a projection polar coordinate geometric expression method for the moon-based earth observation image, which takes into account of movement of the sublunar point, while considering multi-platform earth observation data, to realize accurate automatic geometric correction of the moon-based platform earth observation data.

Abstract: Disclosed is a computer-readable medium including a program code that, when executed by processing circuitry, causes the processing circuitry to generate a feature map from an input image, to extract a region of interest from the feature map, and to generate a predicted mask based on the region of interest. The processing circuitry may use a predicted mask and a real mask to learn a convolutional neural network system. The real mask includes first pixels corresponding to the real boundary and second pixels corresponding to a fake boundary adjacent to the real boundary.

Abstract: An image recognition apparatus is provided with: a divider configured to divide a frame image into a plurality of areas, wherein the frame image is obtained by imaging or capturing a scene outside through a window glass on which a wiper is mounted, from an interior of a moving body; a determinator configured to determine whether or not at least one of the wiper and an attached substance to the window glass appears, for each of the plurality of areas; an object detector configured to detect an object included in such an area of the plurality of areas that it is determined that the wiper and the attached substance do not appear; and a tracking device configured to track a tracking target, which is an object included in two or more frame images, on the basis of a detection result of the object detector.

Abstract: A display device includes a display panel including a first pixel that emits light with a first luminance that is lower than target luminance and a second pixel that emits light with a second luminance that is higher than the target luminance, a sensor configured to measure a first characteristic of a first light emitting element in the first pixel and a second characteristic of a second light emitting element in the second pixel, and a data compensator configured to calculate a degradation amount of the second pixel based on the first characteristic and the second characteristic.

Abstract: Methods for recovering saturated pixel values in raw pixel data are described. Given an image with raw pixel values captured using sensors with a color filter array (CFA), image regions are classified according to how many color channels of the CFA are saturated. Values of saturated pixels are estimated based on weighted color ratios of unsaturated pixels and recovered pixels. Example methods for a Bayer CFA are provided.

Abstract: A video processing system includes: an acquirer that acquires video data including a main video; a format acquirer that acquires a display format that indicates a luminance dynamic range format displayable by a video display apparatus; a generator that generates first characteristics information that indicates first dynamic luminance characteristics that correspond to the display format by using the video data when a luminance dynamic range format of the main video indicated by the video data is different from the display format, the first dynamic luminance characteristics being dynamic luminance characteristics indicating a time-dependent change in luminance characteristics of the main video; and a video transmitter that outputs the first characteristics information generated by the generator.

Abstract: An image processing apparatus includes a receiving unit that receives a first image that is captured with an object irradiated with light from a light irradiating unit and a second image that is captured with the object not irradiated with light from the light irradiating unit, and a generating unit that generates a composite image by adjusting brightness of a whole area of the second image such that brightness at a specific location in the first image is approximately equal to brightness at a corresponding specific location in the second image, and by replacing in the first image a pixel value of a pixel that is affected by light irradiation by the light irradiating unit with a pixel value of a pixel in the second image that has been adjusted in brightness.

Abstract: An image processing device image processing device (100), in which correction is less likely to be impossible or correction is less likely to be inappropriate as compared with that in the related art, performs, on an image including a face (500), correction processing to reduce/expand a first correction region (501) of a circle that is located at a position corresponding to a center of the face (500) and that has a size depending on a size of the face (500) and to expand/reduce a second correction region (502) around the first correction region.

Abstract: Supervised machine learning using neural networks is applied to denoising images rendered by MC path tracing. Specialization of neural networks may be achieved by using a modular design that allows reusing trained components in different networks and facilitates easy debugging and incremental building of complex structures. Specialization may also be achieved by using progressive neural networks. In some embodiments, training of a neural-network based denoiser may use importance sampling, where more challenging patches or patches including areas of particular interests within a training dataset are selected with higher probabilities than others. In some other embodiments, generative adversarial networks (GANs) may be used for training a machine-learning based denoiser as an alternative to using pre-defined loss functions.

Abstract: An omnidirectional barcode may include a graphical representation of a one-dimensional barcode wherein the graphical representation of the one-dimensional barcode is extended circumferentially around a three-dimensional (3D) object. A method of forming an omnidirectional barcode may include creating a graphical representation of a one-dimensional barcode and extending the one-dimensional representation in at least one dimension.

Abstract: An image sensor includes a pixel array that includes a plurality of pixels, a first interface directly connected to an external gyro sensor and that receives gyro data output by the gyro sensor in response to motion, and a control logic that generates image data by exposing the plurality of pixels for a predetermined exposure period, generates valid data that correspond to the exposure period using the gyro data, and generates, based on the valid data, compensation information that represents a movement path of the motion.

Abstract: A motion vector (MV) projection method includes generating motion field motion vectors (MFMVs) for a first portion of a current frame by applying MV projection to MVs of a portion of each of reference frames and storing the MFMVs of the first portion of the current frame into an MFMV buffer, and generating MFMVs for a second portion of the current frame by applying MV projection to MVs of a portion of each of the reference frames and storing the MFMVs of the second portion of the current frame into the MFMV buffer. The second portion does not overlap the first portion. Before generating the MFMVs for the second portion of the current frame is done, at least one of the MFMVs of the first portion is read from the MFMV buffer and involved in motion vector determination of at least one coding block included in the first portion.

Abstract: A system includes an image sensor, an imaging pipeline, and a display device. The image sensor is configured to capture a first frame of pixel data. The imaging pipeline is coupled to the image sensor to receive the first frame of pixel data. The imaging pipeline includes an adaptive noise filter. The adaptive noise filter is configured to filter a pixel based on noise in the pixel. The imaging pipeline is configured to output a second frame of pixel data. The second frame of pixel data includes pixels filtered by the adaptive noise filter. The display device is coupled to the imaging pipeline to receive the second frame of pixel data. The display device being configured to display the second frame of pixel data.

Abstract: Reader observation and monitoring of the imaging of biological development is shown and described. In one embodiment, a method for monitoring biological growth includes identifying an alert from comparing a background value identified on a plate to a non-background value identified on a plate. In particular examples, a pre-filter, threshold, and histogram analysis identifies an alert count.

Abstract: Various embodiments of the present disclosure relate generally to systems and methods for drone-based systems and methods for capturing a multi-media representation of an entity. In some embodiments, the multi-media representation is digital, or multi-view, or interactive, and/or the combinations thereof. According to particular embodiments, a drone having a camera is controlled or operated to obtain a plurality of images having location information. The plurality of images, including at least a portion of overlapping subject matter, are fused to form multi-view interactive digital media representations.

Abstract: A coded aperture mask is provided. The coded aperture mask may include a 2-D planar substrate having a plurality of holes constructed based on a Cartesian product of a first 1-D aperture set and a second 1-D aperture set. The first 1-D aperture set may have a first balanced decoder. The second 1-D aperture set may have a second balanced decoder. The Cartesian product may involve the first 1-D aperture set and the second 1-D aperture set arranged in a non-zero angle (e.g., 90 degrees) to each other. The first 1-D aperture set may define a first axis of the 2-D planar substrate. The second 1-D aperture set may define a second axis of the 2-D planar substrate. The plurality of holes on the 2-D planar substrate may correspond to holes in both of the first 1-D aperture set and the second 1-D aperture set.

Abstract: An inspection apparatus that inspects printed matter generated based on print data by using a read image of a printed surface of the printed matter includes processing circuitry. The processing circuitry performs acquisition, calculation, correction, and comparison. The processing circuitry acquires the read image as an inspection target image, and acquires a reference image based on the print data. The processing circuitry calculates a flare inverse correction value based on flare that occurs at time of creation of the inspection target image. The processing circuitry performs flare inverse correction on the reference image based on the flare inverse correction value to acquire a corrected reference image. The processing circuitry compares the corrected reference image with the inspection target image to perform inspection.

Abstract: Codecs that use larger blocks may have larger boundary regions that may benefit from filtering. In some embodiments, the deblocking filter determines filters and/or a number of samples of the block to be filtered based on block dimensions. For example, In one embodiment, deblocking filter parameters for the video block are determined based on at least one dimension of the size of the video block. The filter parameters include a filter to be applied or a number of pixels along a boundary with a neighboring block to which the filter is to be applied determined based on the at least one dimension.

Abstract: Devices and methods for determining the thickness of hair are provided. An exemplary device includes a computing unit with a processor and a local memory. The processor is connected to the local memory and configured to read data from and/or to write data into the local memory. The device further includes a user input/output unit configured for interaction with a user. Also, the computing unit includes an image data input interface for receiving image data of at least one individual hair. The image data input interface is connected to the processor. Further, the computing unit includes a user interface connected to the user input/output unit and to the processor. The computing unit is configured to process the image data of the at least one individual hair to create hair thickness values of the at least one individual hair and to output the hair thickness values to the user interface.

Abstract: Systems and methods for improved image denoising using a deep learning network model are disclosed. An example system includes an input data processor to process a first patient image of a first patient to add a first noise to the first patient image to form a noisy image input. The example system includes an image data denoiser to process the noisy image input using a first deep learning network to identify the first noise. The image data denoiser is to train the first deep learning network using the noisy image input. When the first deep learning network is trained to identify the first noise, the image data denoiser is to deploy the first deep learning network as a second deep learning network model to be applied to a second patient image of the first patient to identify a second noise in the second patient image.

Abstract: An information processing apparatus includes an acquiring unit configured to acquire first data and second data, a deriving unit configured to derive gradient data describing a gradient of correspondence data on the basis of the correspondence data describing a correspondence between partial data of the first data and partial data of the second data, and a propagating unit configured to propagate the correspondence data on the basis of the gradient data.

Abstract: An information processing apparatus provided with a display portion including a plurality of sub-pixels every pixel, connected to an output apparatus for presenting an image displayed on the display portion to a user through a lens, and producing the image to be displayed on the display portion receives source image data becoming an object of display, obtains a corresponding pixel on the source image data by information on a position of the sub-pixel and a predetermined conversion expression every sub-pixel included in a pixel with respect to the respective pixels constituting corrected image data, and determines luminances of the respective sub-pixels based on a pixel value of the pixel on the source image obtained every sub-pixel, thereby producing the corrected image data.

Abstract: Embodiments are disclosed that apply adaptive sub-tiles to captured images for distortion correction in vision-based assistance systems and methods. A captured image is processed to generate corrected tiles, and selected numbers of sub-tiles are used to generate each of the corrected tiles depending upon the pixel densities for regions of the captured image. The corrected sub-tiles are combined to form corrected tiles, and corrected tiles are combined for form a corrected image. The corrected image can be used to output control signals to cause actions to be issued to a user of the system such as a driver of a vehicle. For one embodiment, the corrected tiles are generated one at a time, and corrected sub-tiles for each corrected tile are also generated one at a time based upon individual source data blocks determined by a pre-determined sub-tile configuration. Efficient memory use and data transfers are provided.

Abstract: A system for filtering an input image dataset is disclosed. A plurality of sequences of input signal values, wherein each sequence corresponds to a different attribute, wherein an input signal value is associated with an attribute and a sampling point i. At least one processor 4 is configured to control computing an output signal value corresponding to a sampling point and an attribute. For a particular sampling point i and for each of a plurality of different attributes of the set of attributes, associating a weight to an input signal value based on a similarity between the signal value and for a plurality of the sampling points j excluding the sampling point i. Also the system computes a weighted sum based on the input signal values and weights. The attribute is a location or a frequency.

Abstract: Embodiments and processes of computer tomography perform tasks associated with denoising a reconstructed image using an anisotropic diffusion filter and adaptively weighting an iterative instance of the diffused image based upon the product of a weight value and a difference between the iterative instance of the diffused image and the original image. In general, the adaptive weighting is a negative feedback in the iterative steps.

Abstract: Certain aspects involve video inpainting in which content is propagated from a user-provided reference frame to other video frames depicting a scene. For example, a computing system accesses a set of video frames with annotations identifying a target region to be modified. The computing system determines a motion of the target region's boundary across the set of video frames, and also interpolates pixel motion within the target region across the set of video frames. The computing system also inserts, responsive to user input, a reference frame into the set of video frames. The reference frame can include reference color data from a user-specified modification to the target region. The computing system can use the reference color data and the interpolated motion to update color data in the target region across set of video frames.

Abstract: Methods and apparatus for generating improved image data from received input image data comprising first input image data associated with a first exposure level and second input image data associated with a second, different, exposure level. Motion detection data is generated from the received input image data by applying a noise model and improved image data is generated by combining data from the first and second input data in dependence on the motion detection data.

Abstract: A system and a method are provided that prevent a wrong panel of a structure such as an aircraft from being opened for inspection of an interior of the structure. A system includes a computer, a portable terminal and a camera configured to capture an image of a structure with panels attached to generate a first image and capture another image of the structure with a predetermined panel among the panels removed to generate a second image and configured to send the first image and the second image to the computer. The computer combines the first image and the second image to generate a third image and sends the third image to the portable terminal. The portable terminal displays the third image. In the third image, the predetermined panel is translucently shown with at least one component at an interior of the predetermined panel being seen through.

Abstract: There is provided an image processing apparatus in which a part used in predetermined processing is specified with use of a relatively small binarization threshold, from parts that have been converted into black pixels through binarization processing using a relatively large binarization threshold.

Abstract: The present disclosure relates to using a neural network to efficiently denoise images that were generated by a ray tracer. The neural network can be trained using noisy images generated with noisy samples and corresponding denoised or high-sampled images (e.g., many random samples). An input feature to the neural network can include color from pixels of an image. Other input features to the neural network, which would not be known in normal image processing, can include shading normal, depth, albedo, and other characteristics available from a computer-generated scene. After the neural network is trained, a noisy image that the neural network has not seen before can have noise removed without needing manual intervention.

Abstract: The present disclosure relates to an information processing device, an information processing method, and a program that enable generation of stable PUFs. The information processing device includes: a reading unit that reads output data from a predetermined element a plurality of times; an average value calculation unit that calculates average values of the output data read by the reading unit; a median calculation unit that calculates the median of the average values calculated by the average value calculation unit; and a PUF generation unit that generates a physical unclonable function (PUF) by comparing the median with the average values. The predetermined element is an image sensor, and the reading unit reads output data from the image sensor when the image sensor is shielded from light. The present technology can be applied to imaging devices, for example.

Abstract: A method for processing at least one X-ray image is provided. A variance of noise is signal dependent. The method includes applying a variance-stabilizing transformation to image data of the at least X-ray image to generate variance-stabilized data. At least one transform parameter of the variance-stabilizing transformation is dependent on a property of the at least one X-ray image that depends on an X-ray imaging device and/or a measurement parameter used to record the at least one X-ray image. A noise reduction algorithm is applied to the variance-stabilized data to generate noise-reduced data, and an inverse transformation of the variance stabilizing transformation is applied to the noise-reduced data to generate a denoised X-ray image.

Abstract: From a first image using a model, a first uncertainty map is generated. An uncertainty level of a location in the first uncertainty map corresponds to a detection of a known structure in a portion of the first image. A first weighted image corresponding to the first uncertainty map is generated, the generating including assigning a first weight to a pixel of the first image, the first weight corresponding to the uncertainty level of a location in the first uncertainty map corresponding to the pixel. From a second image using a model, a second uncertainty map is generated. A second weighted image corresponding to the second uncertainty map is generated. The first image and the second image are combined to form a composite image, each image participating in the composite image according to the corresponding weighted image.

Abstract: A method of generating a high dynamic range (HDR) image is provided that includes capturing a long exposure image and a short exposure image of a scene, computing a merging weight for each pixel location of the long exposure image based on a pixel value of the pixel location and a saturation threshold, and computing a pixel value for each pixel location of the HDR image as a weighted sum of corresponding pixel values in the long exposure image and the short exposure image, wherein a weight applied to a pixel value of the pixel location of the short exposure image and a weight applied to a pixel value of the pixel location in the pixel long exposure image are determined based on the merging weight computed for the pixel location and responsive to motion in a scene of the long exposure image and the short exposure image.

Abstract: Provided are methods and apparatus related to Artificial Intelligence (AI) downscaling and upscaling and techniques related to reducing artifact problems. Some embodiments include down-scaling an original image through a Deep Neural Network (DNN); generating, from the original image and based on frequency transform coefficients, artifact information representing a region in the first image including an artifact in the first image. Post-processing may be performed based on the artifact information to change pixels in the first image, thus reducing the effect of artifacts.

Abstract: Systems and methods for hybrid depth regularization in accordance with various embodiments of the invention are disclosed. In one embodiment of the invention, a depth sensing system comprises a plurality of cameras; a processor; and a memory containing an image processing application. The image processing application may direct the processor to obtain image data for a plurality of images from multiple viewpoints, the image data comprising a reference image and at least one alternate view image; generate a raw depth map using a first depth estimation process, and a confidence map; and generate a regularized depth map. The regularized depth map may be generated by computing a secondary depth map using a second different depth estimation process; and computing a composite depth map by selecting depth estimates from the raw depth map and the secondary depth map based on the confidence map.