Abstract: Some implementations relate to methods, systems, and computer-readable media to generate text tags for games. In some implementations, a computer-implemented method to generate a 3D mesh for an object using a trained machine learning model includes providing a two-dimensional (2D) image of the object as input to the trained machine learning model, obtaining a template three-dimensional (3D) mesh, and generating, using the trained machine learning model and based on the 2D image and the template 3D mesh, a 3D mesh for the object, wherein the 3D mesh for the object is obtained by deforming the template 3D mesh and is usable to map a texture or to generate a 3D animation of the object.

Abstract: Disclosed is an electronic apparatus. The electronic apparatus includes a processor configured to obtain first upscaling information of an input image using an artificial intelligence (AI) model that is trained to obtain upscaling information of an image. The processor is also configured to downscale the input image based on the obtained first upscaling information, and obtain an output image by upscaling the downscaled image based on an output resolution.

Abstract: The invention provides for a medical imaging system comprising: a memory for storing machine executable instructions; a processor for controlling the medical instrument. Execution of the machine executable instructions causes the processor to: receive MRF magnetic resonance data acquired according to an MRF magnetic resonance imaging protocol of a region of interest; reconstruct an MRF vector for each voxel of a set of voxels descriptive of the region of interest using the MRF magnetic resonance data according to the MRF magnetic resonance imaging protocol; calculate a preprocessed MRF vector (126) for each of the set of voxels by applying a predetermined preprocessing routine to the MRF vector for each voxel, wherein the predetermined preprocessing routine comprises normalizing the preprocessed MRF vector for each voxel; calculate an outlier map for the set of voxels by assigning an outlier score to the preprocessed MRF vector using a machine learning algorithm.

Abstract: A correlation system can be implemented on a user device (e.g., smartphone) to perform image processing tasks e.g., point to point correlations for tasks such as alignment, tracking, and reconstruction of model data. The correlation system can implement normalized cross correlation in a least squares optimization scheme without use of approximations. Normalized cross correlation optimized via least squares can further implement global and local cost functions that are invariant to dynamic lighting conditions.

Abstract: A system and method for determining an object is disclosed. A security appliance, with a processor and memory is provided. The security appliance is configured to receive image of an object captured by a security device. The security device located in a first location. The security device also receives image of another object captured by another security device. The another security device located in a second location. The image of the object is processed to generate a first plurality of attributes for the object. The image of the another object is processed to generate a second plurality of attributes for the another object. The first plurality of attributes for the object and the second plurality of attributes for the another object are compared and based on the comparison, determine that the object and the another object are the same.

Abstract: A versatile template generation unit cuts a first region in which a similarity level to a template image is a first similarity level and a second region in which the similarity level to the template image is a second similarity level different from the first similarity level, from an input image including an alignment mark, to generate a versatile template image.

Abstract: A method of obtaining and geo-registering an aerial image of an object of interest is provided. The method includes obtaining an aerial image and accessing an object detection model trained using a machine learning algorithm. The method includes training set of aerial images of an object of interest, and using the object detection model to detect the object of interest in the aerial image. The object detection includes a prediction of a boundary of the object of interest depicted in the aerial image based on the defined boundary of the object of interest. And the method includes geo-registering the aerial image including the prediction of the boundary of the object of interest with a geographic location of the object of interest.

Abstract: A method of obtaining and geo-registering an aerial image of an object of interest is provided. The method includes obtaining an aerial image by a camera onboard an aircraft. The method includes accessing an object detection model trained using a machine learning algorithm and a training set of aerial images of an object of interest, and using the object detection model to detect the object of interest in the aerial image. The object detection includes a prediction of a boundary of the object of interest depicted in the aerial image based on the defined boundary of the object of interest. The method includes accessing a data store including a geographic location of the object of interest. And the method includes geo-registering the aerial image including the prediction of the boundary of the object of interest with the geographic location of the object of interest.

Abstract: A semiconductor processing method is provided. The method includes providing a first carrier (10). A first adhesive (18) is provided on the first carrier (10) and a plurality of semiconductor chips (20) is placed on the first adhesive (18). A second carrier (28) is provided. The second carrier (28) is provided with a plurality of chip receiving areas (32). The first and second carriers (10) and (28) are then brought together to attach the semiconductor chips (20) to respective ones of the chip receiving areas (32) on the second carrier (28). The first carrier (10) is then separated from the semiconductor chips (20).

Abstract: A representative waveform providing apparatus and method, a blood pressure estimation apparatus, and a wearable device are provided. The representative waveform providing apparatus may include a signal obtainer configured to obtain a periodic signal and a processor configured to determine a representative waveform of the periodic signal by using a dynamic time warping algorithm.

Abstract: There is provided herein an apparatus and method for roadside asset tracking and maintenance monitoring having a mobile unit with data capture devices for capturing roadside asset imagery, global positioning system (GPS) receivers and data interfaces for communicating with an asset management server. As such, the apparatus may take roadside imagery for automated asset identification which may include utilising an asset type image recognition technique for automating the identification of the roadside assets.

Abstract: Provided is an image processing device that can suppress deterioration in the accuracy of identifying a subject, even in cases where the shape of the subject in an image is deformed. A feature amount calculation means 81 detects feature points from an image, and calculates, for each feature point, a local feature amount for the feature point on the basis of a peripheral region of said feature point, including said feature point. A correlation identifying means 82 specifies the correlation between feature points in a first image and feature points in a second image on the basis of the local feature amount of each feature point in the first image and the local feature amount of each feature point in the second image. A partial region detection means 83 detects, from one of the first image or the second image, partial regions each including a feature point in said image.

Abstract: A sensor acquires an input image X of a scene. The image includes noise with a variance ?2. A deep Gaussian conditional random field (GCRF) network is applied to the input image to produce an output image Y, where the output image is denoised, and wherein the deep GCRF includes a prior generation (PgNet) network followed by an inference network (InfNet), wherein the PgNet produces patch covariance priors ?ij for patches centered on every pixel (i,j) in the input image, and wherein the InfNet is applied to the patch covariance priors and the input image to solve the GCRF.

Abstract: The application provides a method and system for determinism in non-linear systems for speech processing, particularly automatic speech segmentation for building speech recognition systems. More particularly, the application enables a method and system for detecting boundary of coarticulated units from isolated speech using recurrence plot.

Abstract: A method is provided for content alignment. The method includes obtaining a first content sequence and a second content sequence different from the first content sequence. The method also includes representing each of the first content sequence and the second content sequence in a hierarchical structure containing an ordered root element sequence and a sub-tree structure. The ordered root element sequence includes a plurality of root elements and each root element is associated with a sub-tree of elements. The method also includes determining a desired alignment between the first content sequence and the second content sequence using dynamic programming, and outputting results of the desired alignment between the first content sequence and the second content sequence.

Abstract: A temporal information integration dis-occlusion system and method for using historical data to reconstruct a virtual view containing an occluded area. Embodiments of the system and method use temporal information of the scene captured previously to obtain a total history. This total history is warped onto information captured by a camera at a current time in order to help reconstruct the dis-occluded areas. The historical data (or frames) from the total history match only a portion of the frames contained in the captured information. This warping yields warped history information. Warping is performed by using one of two embodiments to match points in an estimation of the current information to points in the captured information. Next, regions of current information are split using a classifier. The warped history information and the captured information then are merged to obtain an estimate for the current information and the reconstructed virtual view.

Abstract: A visual quality assessment method and system are based on deriving a quality metric by comparing sub-band approximations of a distorted image and an undistorted version of the same image, providing a good compromise between computational complexity and accuracy. The sub-band approximations are derived from Discrete Wavelet (Haar) transforms of small image blocks of each image. Due to inherent symmetries, the wavelet transform is “blind” to certain types of distortions. But the accuracy of the method is enhanced, and the blindness of the transform is overcome, by computing quality metrics for the distorted image as well as computing quality metrics for a shifted version of the distorted image and combining the results.

Abstract: Candidate points belonging to a predetermined structure are extracted from image data DV. A shape model which represents a known shape of the predetermined structure and is formed by model labels having a predetermined connection relationship is obtained. Corresponding points corresponding to the model labels are selected from the extracted candidate points under the following constraints: (a) each model label is mapped with only one of the candidate points or none of the candidate points; (b) each candidate point is mapped with only one of the model labels or none of the model labels; and (c) when a path between two candidate points which are mapped with each pair of the model labels connected with each other is determined, each candidate point which is mapped with none of the model labels is included in only one of the determined paths or none of the determined paths.

Abstract: A video frame compression system includes a rendering engine that provides a current video frame and current additional rendering information. Additionally, the video frame compression system includes a warping engine that generates a warped video frame, wherein the warped video frame is a transformation of a previous video frame that is based on the current additional rendering information. Further, the video frame compression system includes a video encoder that compresses the current video frame by using the warped video frame as a reference frame and separately compresses the current additional rendering information. Still further, the video frame compression system includes a packetizer that provides main and auxiliary data streams corresponding to the compressed current video frame and the compressed current additional rendering information, respectively. A video frame decompression system and methods of video frame compression and decompression are also provided.

Abstract: A novel technique for unsupervised feature selection is disclosed. The disclosed methods include automatically selecting a subset of a feature of an image. Additionally, the selection of the subset of features may be incorporated with a congealing algorithm, such as a least-square-based congealing algorithm. By selecting a subset of the feature representation of an image, redundant and/or irrelevant features may be reduced or removed, and the efficiency and accuracy of least-square-based congealing may be improved.

Abstract: A computer-implemented method for matching objects is disclosed. At least two images where one of the at least two images has a first target object and a second of the at least two images has a second target object are received. At least one first patch from the first target object and at least one second patch from the second target object are extracted. A distance-based part encoding between each of the at least one first patch and the at least one second patch based upon a corresponding codebook of image parts including at least one of part type and pose is constructed. A viewpoint of one of the at least one first patch is warped to a viewpoint of the at least one second patch. A parts level similarity measure based on the view-invarient distance measure for each of the at least one first patch and the at least one second patch is applied to determine whether the first target object and the second target object are the same or different objects.

Abstract: In one embodiment, an image editing method, comprising: defining a region as a foreground object and a remaining region as background; applying an image warping process to a predefined object region, wherein a distortion to at least the remaining region occurs based on the image warping process applied to the predefined object region; responsive to the image warping process, determining a region to be repaired, the region to be repaired comprising a missing background; repairing the region to be repaired based on an inpainting process; and generating a repaired image by combining the predefined object region to which the image warping process is applied, one or more non-distorted regions, and the repaired region.

Abstract: A method for aligning and unwarping distorted images in which lens profiles for a variety of lens and camera combinations are precomputed. Metadata stored with images is used to automatically determine if a set of component images include an excessive amount of distortion, and if so the metadata is used to determine an appropriate lens profile and initial unwarping function. The initial unwarping function is applied to the coordinates of feature points of the component images to generate substantially rectilinear feature points, which are used to estimate focal lengths, centers, and relative rotations for pairs of the images. A global nonlinear optimization is applied to the initial unwarping function(s) and the relative rotations to generate optimized unwarping functions and rotations for the component images. The optimized unwarping functions and rotations may be used to render a panoramic image.

Abstract: Three dimensional models corresponding to a target image and a reference image are selected based on a set of feature points defining facial features in the target image and the reference image. The set of feature points defining the facial features in the target image and the reference image are associated with corresponding 3-dimensional models. A 3D motion flow between the 3-dimensional models is computed. The 3D motion flow is projected onto a 2D image plane to create a 2D optical field flow. The target image and the reference image are warped using the 2D optical field flow. A selected feature from the reference image is copied to the target image.

Abstract: A system and a method for indicating at least one of the internal structures of an organ on an X-ray image are proposed. The system includes an interface adapted to receive the X-ray image and a non-X-ray image pertaining to the organ. The system also includes a database having a geometric model of the internal structures of the organ, a first module for determining at least a dimension of one of the internal structures of the organ from the non-X-ray image, and a second module for indicating the at least one of the internal structures of the organ in the X-ray image based on the geometric model adjusted by the at least one dimension.

Abstract: An apparatus and method for identifying image capture opportunities. Expert photo agents (526) that process metadata extracted from a captured image are obtained (706) at a user-device (102). Metadata (214) characterizing visual content of a captured image is produced (310) by processing the visual content of the captured image. Image capture opportunities are selected (410) based upon processing of the metadata with the at least one expert photo agent. The image capture opportunity selection specifies content to capture in a subsequent image.

Abstract: The invention is related to a method for registering at least one part of a first image and of a second image, including the steps of providing a first image of the real environment and a coordinate system associated to the first image, a second image of the real environment, and measurements of orientation and/or distance of a capturing device with respect to the real environment while capturing the second image. A collineation warping function is then determined from the measurements associated to at least one part of the second image.

Abstract: It is provided a computer system for matching between images/graphs comprising a processing unit, a storage unit, an input unit and a data recording unit, wherein the computer system includes a feature pattern generating unit for generating a set of feature patterns of the image and the graph from the received image and the received graph, segmenting each of the image and the graph into a plurality of regions centered on a sample point, and determining characteristics of the plurality of regions; a correspondence computing unit for computing correspondence between the image and the graph to be matched by comparing the generated feature patterns; a transform unit for transforming a coordinate system of at least one of the image and the graph to be matched; and a combination unit for combining the image and the graph that have been subjected to the transform of the coordinate system.

Abstract: Input image data may be subjected to a spatial convolution to produce a plurality of different filter matrices or kernels. Thus, a single system may be capable of producing a plurality of different filter sizes which may be selected based on particular circumstances, uses and described precision. Thus, a single system may provide variable spatial filtering. By reducing the input data matrix size and by reusing components, in some cases, the computational complexity of producing a plurality of symmetrical filters is not significantly greater than that involved in producing only a single filter.

Abstract: An apparatus is provided that includes a plurality of modules, a plurality of memory banks, and a multiplexor. Each module includes at least one agent that interfaces between a module and a memory bank. Each memory bank includes an arbiter that interfaces between the at least one agent of each module and the memory bank. The multiplexor is configured to assign data paths between the at least one agent of each module and a corresponding arbiter of each memory bank based on the assigned data path. The at least one agent of each module is configured to read data from the corresponding arbiter of the memory bank or write modified data to the corresponding arbiter of the memory bank.

Abstract: A computer program product tangibly embodied in a computer-readable storage medium includes instructions that when executed by a processor perform a method. The method includes identifying a frame of a video sequence, transforming a model into an initial guess for how the region appears in the frame, performing an exhaustive search of the frame, performing a plurality of optimization procedures, wherein at least one additional model parameter is taken into account as each subsequent optimization procedure is initiated. A system includes a computer readable storage medium, a graphical user interface, an input device, a model for texture and shape of the region, the model generated using the video sequence and stored in the computer readable storage medium, and a solver component.

Abstract: A method for text recognition includes generating a number of text hypotheses for an image, for example, using an HMM based approach using fixed-width analysis features. For each text hypothesis, one or more segmentations are generated and scored at the segmental level, for example, according to character or character group segments of the text hypothesis. In some embodiments, multiple alternative segmentations are considered for each text hypothesis. In some examples, scores determined in generating the text hypothesis and the segmental score are combined to select an overall text recognition of the image.

Abstract: Described are a method and an apparatus for generating a forecast weather image such as a forecast weather radar image. The method uses filters to approximate viscous alignment and to thereby determine displacement fields having meaningful structure. In various embodiments a power-law energy spectrum is utilized for deformations in the displacement field through the application of a set of Gabor filters. The Gabor filters are applied in a sequential manner to gradient error images and values of modes of motions corresponding to the Gabor filters are determined. The values are used to generate the displacement field which can then be applied to an existing weather image to generate a forecast weather image.

Abstract: The present technology relates to an image processing apparatus, an image processing method, and a program capable of improving the image quality of decoded images of images from multiple viewpoints. A warping unit 461 warps a picture of an image of a viewpoint #1 between the image of the viewpoint #1 and an image of a viewpoint #2 different of the viewpoint #1 to generate a picture of a warped image obtained by converting the picture of the image of the viewpoint #1 into an image obtainable in the viewpoint #2. A disparity prediction unit 463 acquires, from header information, a reference index indicating a reference picture that is referred to generate a predicted image of an object block to be decoded of a picture of the image of the viewpoint #2, and selects the reference picture from among reference picture candidates including at least the picture of the warped image based on the reference index.

Abstract: A subject tracking device includes: an input unit that sequentially inputs input images; an arithmetic operation unit that calculates a first similarity level between an initial template image and a target image and a second similarity level between an update template image and the target image; a position determining unit that determines a subject position based upon at least one of the first and the second similarity level; a decision-making unit that decides whether or not to update the update template image based upon the first and the second similarity level; and an update unit that generates a new update template image based upon the initial template image multiplied by a first weighting coefficient and the target image multiplied by a second weighting coefficient, and updates the update template image with the newly generated update template image, if the update template image is decided to be updated.

Abstract: A paper sheet identification apparatus capable of identifying an authenticity of a watermark area formed on a paper sheet is provided without increasing the cost. The paper sheet identification apparatus includes: a light receiving part receiving reflected light from a watermarked image formed on a paper sheet to be conveyed, a converter converting the reflected light from the watermarked image received by the light receiving part for each pixel as a unit of a predetermined size including color information having brightness; and an identification processing part identifying the authenticity of the watermarked image based on a correlation coefficient, which is calculated from a density value for each pixel converted by the converter and a density value for each pixel by the transmitted light from the watermarked image of the bill serving as a reference.

Abstract: Per one embodiment, a method is provided for refining a pose estimate of a model. The model is coarsely aligned with a run-time image, and it represents a 2D pattern. The pose estimate includes at least one pose estimate parameter. The model has a plurality of model features, and the run-time image has a plurality of run-time features. A given distance value is determined representing a given distance between a given one of the plurality of the model features mapped by a given pose estimate and a corresponding given run-time feature. A two-dimensional model description of the two-dimensional model pattern is provided. The two-dimensional model pattern is mapped using the given pose estimate to create a transformed version of the two-dimensional model pattern. The transformed version represents a non-linear movement of at least portions of the two-dimensional model pattern in a direction orthogonal to a plane of the two-dimensional model description.

Abstract: To provide an inter-pattern feature corresponding device (1) for determining a feature correspondence relationship with high accuracy even if a pattern to be collated has distortion.

Abstract: The present invention provides a band-pass filter, being capable of fitting a frequency spectrum area having phase information in a frequency spectrum image, to obtain a spectrum information corresponding to the phase information during the process of obtaining the phase information from the frequency spectrum image with respect to an object's surface profile. In another embodiment, the present invention further provides a method to optimize the spectrum range of the band-pass filter so as to enhance measuring accuracy and efficiency while restoring the surface of the object. In addition, by employing the foregoing method, the present invention further provides a measurement system for measuring three-dimensional surface shapes in which a deformed fringe pattern with respect to the measured object's surface is acquired and the phase information is obtained from the fringe pattern according to the foregoing method so as to restore the surface profile of the measured object.

Abstract: A method for aligning and unwarping distorted images in which lens profiles for a variety of lens and camera combinations are precomputed. Metadata stored with images is used to automatically determine if a set of component images include an excessive amount of distortion, and if so the metadata is used to determine an appropriate lens profile and initial unwarping function. The initial unwarping function is applied to the coordinates of feature points of the component images to generate substantially rectilinear feature points, which are used to estimate focal lengths, centers, and relative rotations for pairs of the images. A global nonlinear optimization is applied to the initial unwarping function(s) and the relative rotations to generate optimized unwarping functions and rotations for the component images. The optimized unwarping functions and rotations may be used to render a panoramic image.

Abstract: Performing word recognition operations to determine what an image of a word represents. The method includes accessing a first image. The first image represents an image version of a word. The method further includes accessing a second image. The second image also represents an image version of a word. Using a warp mesh, the method includes warping the second image to cause the second image to approximately match the first image by applying a mesh to the second image and moving vertices of the mesh to warp the second image. The difference between the warped second image and the unwarped first image are determined.

Abstract: A method and system for increasing the certainty of a silhouette matching process, where the process is being used for attitude determination of an object of interest, for example an aircraft. The method involves using one or more mask images that include structure or features that may or may not always be associated with the object of interest, and overlaying the mask image(s) onto a library image of the aircraft. Each pixel of the library image is compared against corresponding pixels of the mask image(s) to determine which pixels represent ambiguous areas of the library image. Those pixels are eliminated from consideration in determining a Fit score, where the Fit score represents a percentage value indicative of a certainty of the matching process in identifying the attitude of the aircraft. The method and system is applicable to a wide ranging variety of object detection applications.

Abstract: Disclosed is a method for determining the absence or presence of one or more instances of a predetermined pattern in an image, and for determining the location of each found instance within a multidimensional space. A model represents the pattern to be found, the model including a plurality of probes. Each probe represents a relative position at which a test is performed in an image at a given pose, each such test contributing evidence that the pattern exists at the pose. The method further includes a comparison of the model with a run-time image at each of a plurality of poses. A match score is computed at each pose to provide a match score surface. Then, the match score is compared with an accept threshold, and used to provide the location any instances of the pattern in the image.

Abstract: A 2D model fitting means 1201 estimates values of parameters optimum to generate a probe model image similar to a probe image and a gallery model image similar to a gallery image with an image variation model 1204. At that time, among a plurality of parameters of the image variation model 1204, the value of a parameter of which sameness between the probe image and the gallery image is to be judged as a target parameter is set to be the same for both images. A model goodness-of-fit evaluation means 1202 computes a model goodness of fit for the probe model image and the gallery model image to the probe image and the gallery image under the estimated parameters. A collation means 1203 compares the model goodness of fit with a threshold value to judge the sameness between the probe image and the gallery image.

Abstract: An image search method that is robust and fast (with computational complexity of logarithmic order relative to the number of models). The image search method including: extracting a plurality of specific regions possessing such a property that a shape can be normalized regardless of an affine transformation thereof, as affine-invariant regions from one or more learning images; calculating, with respect to a reference affine-invariant region, other neighboring affine-invariant regions as a set; deforming the neighboring affine-invariant regions by a transformation to normalize the shape of the reference affine-invariant region; and outputting the deformed shapes of the neighboring affine-invariant regions, together with combination of the reference affine-invariant region and the neighboring affine-invariant regions.

Abstract: A method for creating a coordinated image display by designing a digital template for the image display. The template includes a number of openings for displaying digital images. Each of the openings has at least one stored required image attribute that must be satisfied to a programmed degree according to the artistic design of the template. A template recipe includes an image attribute for a dependent opening, which means that it's defined to be dependent upon at least one image attribute of an image in another opening of the template. The computer system automatically searches an image database for an image satisfying the required image attribute for the dependent opening. If a user of the computer system manually replaces or modifies an image upon which another image depends, the computer system will automatically search the image database for a new image that satisfies the required dependency upon the replaced or modified image, if necessary.

Abstract: A method for creating an artistically coordinated image display. A digital template is provided for said image display and it includes of openings for placing images each having at least one required attribute for an image. A programmed computer system automatically searches a database of images for images to be placed in the openings and each of the images satisfy the openings required image attributes. One or more vertical and horizontal lines are demarcated in the so that subjects in the image can be placed on the lines or their intersecting points by modifying and shifting the image appropriately. A subject of the image can also be measured and its size can be set as a reference measurement unit to assist in aesthetically placing subjects proportionally within the image.

Abstract: The present invention provides a system and method for detecting deformable objects in images even in the presence of partial occlusion, clutter and nonlinear illumination changes. A holistic approach for deformable object detection is disclosed that combines the advantages of a match metric that is based on the normalized gradient direction of the model points, the decomposition of the model into parts and a search method that takes all search results for all parts at the same time into account. Despite the fact that the model is decomposed into sub-parts, the relevant size of the model that is used for the search at the highest pyramid level is not reduced. Hence, the present invention does not suffer the speed limitations of a reduced number of pyramid levels that prior art methods have.

Abstract: Disclosed is a method for determining the absence or presence of one or more instances of a predetermined pattern in an image, and for determining the location of each found instance within a multidimensional space. A model represents the pattern to be found, the model including a plurality of probes. Each probe represents a relative position at which a test is performed in an image at a given pose, each such test contributing evidence that the pattern exists at the pose. The method further includes a comparison of the model with a run-time image at each of a plurali of poses. A match score is computed at each pose to provide a match score surface. Then, the match score is compared with an accept threshold, and used to provide the location any instances of the pattern in the image.

Abstract: The method is for the identification and characterization of structures in electron micrographs. Structures in a first image are selected. The structures have a first shape type deformed in a first direction. The selected structures are transformed to a second shape type different from the first shape type. The transformed structures of the second shape type are used to form a plurality of templates. A new structure in a second image is identified. The new structure has the first shape type. The second shape type structure of each template is deformed in the first direction. It is determined which template is a preferred template that best matches the new structure.