Abstract: A computing device predicts an event or classifies an observation. A trained labeling model is executed with unlabeled observations to define a label distribution probability matrix used to select a label for each observation. Unique combinations of observations selected from the unlabeled observations are defined. A marginal distribution value is computed from the label distribution probability matrix. A joint distribution value is computed between observations included in each combination. A mutual information value is computed for each combination as a combination of the marginal distribution value and the joint distribution value computed for the respective combination. A predefined number of observation vector combinations is selected from the combinations that have highest values for the computed mutual information value. Labeled observation vectors are updated to include each observation vector included in the selected observation vector combinations with a respective obtained label.

Abstract: An image processing device includes: a plurality of label data generation units which generate label data in which a predetermined label is assigned to each of a plurality of pixels in each of a plurality of divided images; a plurality of label integration information generation units which correspond to the respective label data generation units and generate label integration information representing association of labels included in the label data in order to integrate label data generated by a corresponding label data generation unit and label data generated by another label data generation unit; a plurality of label integration units which generate integrated label data in which respective pieces of label data corresponding to neighboring divided images are integrated; and a label integration processing control unit which distributes the label data to the respective label integration units such that computational loads to integrate the label data are equalized.

Abstract: An object detection system for an autonomous vehicle processes sensor data, including one or more images, obtained for a road segment on which the autonomous vehicle is being driven. The object detection system compares the images to three-dimensional (3D) environment data for the road segment to determine pixels in the images that correspond to objects not previously identified in the 3D environment data. The object detection system then analyzes the pixels to classify the objects not previously identified in the 3D environment data.

Abstract: Implementations relate to generating and displaying blur in images. In some implementations, a method includes generating a plurality of mipmap images based on an input image, including applying a blur to a respective plurality of pixels derived from the input image for each mipmap image. In some examples, the blur is at least partially based on depth data for the image. Parameter data is obtained that indicates an output focal plane depth for an output focal plane of an output image and an output focal range in front of the output focal plane. Output pixel values of the output image are generated, including determining blurred pixel values based on one or more of the mipmap images selected based on the output focal plane depth and the output focal range. The blurred pixel values are based on particular pixels associated with a depth outside the output focal range.

Abstract: In an example, a method is described that includes dividing an input image into a plurality of strips, where each strip is smaller than a whole of the input image. A plurality of binary images is then generated, where each of the binary images corresponds to one of the strips. Connected component labeling is performed on the binary images, one binary image at a time. An object map for the input image is then generated based on the results of the connected component labeling.

Abstract: An image processing device includes a label memory to store a first label assigned to a pixel in a previous line of a processing target image subjected to an assignment process prior to the assignment process on a processing target line of the processing target image, a label area memory to store identification information identifying a label area including pixels assigned with an identical label, and circuitry to, if at least one target pixel in the processing target line is an effective pixel, perform the assignment process on the target pixel based on at least one label that includes the first label and is assigned to at least one peripheral pixel peripheral to the target pixel, and to perform an updating process of updating the identification information based on a position of the target pixel and the label assigned to the target pixel.

Abstract: Embodiments of the present disclosure relate to a document layout system that automatically optimizes dynamic documents (i.e., documents having dynamic content). In particular, the document layout system intelligently populates and arranges content items, including dynamic content items, into a dynamic document. Further, upon receiving dynamic content updates that change the dimensions (e.g., expand or shrink) of one or more dynamic content items within a dynamic document, the document layout system automatically modifies the arrangement of the content items to resolve any position conflicts detected between content items within the dynamic document. Further, the document layout system provides a number of features that assist users in organizing and arranging dynamic documents.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for facial recognition using fractal features are disclosed. In one aspect, a method includes the actions of accessing data encoding a facial image, the facial image including a face. The actions further include generating a hierarchical graphical model of the face in the facial image, the hierarchical graphical model including more than nodes from at least one level, each level approximating the face within a contour. The actions further include applying a bank of filters to the face at a particular node, each filter spanning more than one scale and at least one direction. The actions further include analyzing filter responses from the bank of filters applied at each direction to obtain a similarity measure that consolidates filter responses from filters applied at more than one scale. The actions further include generating a vector representation.

Abstract: A control method for an electronic device comprising a sensor unit having a first detection range and a second detection range measured from the electronic device, and the second detection range being smaller than the first detection range. The method includes: detecting a 3D movement track of an operation body when the relative distance between the operation body and the electronic device is smaller than the first detection range, the 3D movement track being a continuous movement track having one end between the first and second detection range and the other end equal to or smaller than the second detection range; generating a status change signal when the operation body reaches close to or moves away from the second detection range of the sensor unit; executing a corresponding control command according to the 3D movement track and the status change signal.

Abstract: Techniques for providing an accurate auto-crop feature for images captured by an image capture device may be described herein. For example, one or more image masks for a color image captured by an image capture device may be received by a computer system. Metadata about the color image that identifies portions of the image as foreground and the color image itself may also be received by the computer system. Further, a representation of a user and a floor region associated with a user may be extracted from the color image using the one or more image masks and the metadata. A first area of the color image may be cropped with respect to the extracted representation of the user and the floor region associated with the user to generate a second area of the color image. In embodiments, a third area of the color image may be obscured based on the received metadata.

Abstract: A system, method, and media are provided for generating a structured labeling graphical user interface. The user interface receives user input that associates multimedia content with categories. The user input may include user-supplied tags that further define the category for the multimedia content. The user-supplied tags are rendered proximate to the categories. In turn, a database logs user events to store, among other things, the categories, the user-supplied tags, time associated with completing the user-supplied tags, and time for associating multimedia content with the categories or tags.

Abstract: The image data analytics for computation configuration method and computer system with computation accessibility is provided. The computation accessibility and configuration on image data processing are implemented by a plurality of computers having a plurality of processors and a plurality of data storages, the image data analytics for computation accessibility and configuration includes the following steps: inputting an original image by an input device and initializing the original image; defining a plurality of tiles equally dividing the original image into a same size of a regular shape; transferring the plurality of image regions to form a graph having vertices and edges; cutting the graph into a plurality of sub-graphs; arranging the plurality of sub-graphs to the plurality of processors or cores to conduct parallel processing simultaneously for analyzing the image data; and storing a plurality of processing results respectively in the plurality of data storages.

Abstract: In one embodiment, a method includes detecting one or more objects in an image, generating at least one mask for each of the detected objects, wherein each of the masks is defined by a perimeter, classifying the detected objects, receiving gesture input in relation to the image, determining whether one or more locations associated with the gesture input correlate with any of the masks, and providing feedback regarding the image in response to the gesture input. Each of the masks may include data identifying the corresponding detected object, and the perimeter of each mask may correspond to a perimeter of the corresponding detected object. The perimeter of the corresponding detected object may separate the detected object from one or more portions of the image that are distinct from the detected object.

Abstract: A recognition training apparatus according to the present invention includes a generation unit configured to generate relevance between a specific domain and a candidate of a recognition target based on ontology information expressing a concept structure of the specific domain, a selection unit configured to select the recognition target from the candidate of the recognition target based on the relevance generated by the generation unit, and a training unit configured to train a recognizer using training data regarding the recognition target selected by the selection unit.

Abstract: Virtual reality is provided by a mobile VR device. VR scenes are displayed by serving an application's image requests from a local cache of pre-rendered image frames. As the device user moves, image frames that match a VR scene viewed from a current pose (3D position and 3D orientation of the device translated to the virtual environment) are retrieved from the cache and displayed to the user as a VR scene. The cache may include a static cache including static image frames and a dynamic object cache including dynamic object image frames. The static cache may be indexed by the device current pose. The dynamic object cache may be indexed by the dynamic object's animation stage, orientation, and relative distance from the device user's current pose for a given time or movement-based trigger. Static image frames and dynamic object image frames may be composited, and displayed as a scene.

Abstract: An approach to facilitating proximate-tile-type-based map generation is provided. Tile type information that indicates tile types for map tiles of a map to be generated may be received. One or more proximate tile types may be determined for individual ones of the map tiles based on the tile types. The proximate tile types may include one or more of the tile types of other ones of the map tiles that are proximate the individual map tiles on the map. Level information that indicates levels for the individual map tiles may be determined based on the proximate tile types of the individual map tiles. The map may be generated based on the tile type information and the level information. In some implementations, the tile type information may be received as a graphic representation having colors that indicate the tile types.

Abstract: According to some embodiments of the present invention there is provided a computerized method for labeling a cluster of text documents. The method comprises receiving a document cluster and producing automatically multiple document sub-clusters determined by randomly changing some documents. The method applies multiple cluster labeling algorithms on the cluster and on each sub-cluster, to generate ordered lists. The method comprises generating a ranked label list for each cluster labeling algorithm by computing automatically label values, one for each cluster label in the lists of the selected algorithm, and re-ranking the ordered list. The method combines the re-ranked label lists using a label fusing algorithm to produce a fused label list.

Abstract: Design analysis for framework assessment is described. A file including a designed image may be obtained from a storage device, and a guide structure may be extracted from the file. The guide structure may include multiple guide lines, and the guide lines may be analyzed to determine column-related data, such as at least relative widths or positional orders, a total number of columns, groups of column widths, a combination thereof, and so forth. From multiple frameworks, a framework may be ascertained that substantially matches the guide structure based at least partially on a comparison of column-related data to each framework of the multiple frameworks. Ascertainment of a framework may be indicated to an end-user. The ascertained framework may be applied to facilitate development of a coded functional version of at least a visual design of the image.

Abstract: A panoramic video camera includes a lens having a horizontal FOV portion of 360° about an optical axis. The panoramic video camera includes an image sensor, a processor, and a storage medium. The processor executes programmatic instructions to analyze digital image files to locate an object and positions thereof to produce object position data. Executed programmatic instructions also generate a plurality of zoom levels for the respective digital image files based on the object position data. The zoom levels enable a playback device to visually show the object at varying magnifications during playback of a video segment derived from the digital image files. Zoom levels may alter image capture settings of the panoramic video camera, may alter playback device playback settings, or both.

Abstract: A first panoramic video camera of the multi-camera imaging system captures images to produce captured panoramic images. At least some panoramic images include images of an object being tracked by the multi-camera imaging system. The first panoramic video camera generates a set of digital image files based on the captured panoramic images. Each digital image file of the set of digital image files is time sequenced to reflect a time at which a panoramic image was captured. The first panoramic video camera receives over a, first wireless communication network, a control message relating to tracking the object. Responsive to the control message, the first panoramic video camera transmits at least a portion of the set of digital image files over a second wireless communication network, which supports higher data rate transmission than does the first wireless communication network.

Abstract: Computer program products for discriminating hand and machine print from each other, and from signatures, are disclosed and include program code readable and/or executable by a processor to: receive an image, determine a color depth of the image; reducing the color depth of non-bi-tonal images to generate a bi-tonal representation of the image; identify a set of one or more graphical line candidates in either the bi-tonal image or the bi-tonal representation, the graphical line candidates including true graphical lines and/or false positives; discriminate any of the true graphical lines from any of the false positives; remove the true graphical lines from the bi-tonal image or the bi-tonal representation without removing the false positives to generate a component map comprising connected components and excluding graphical lines; identify one or more of the connected components in the component map; and output and/or display and indicator of each of the connected components.

Abstract: A data processing method comprising receiving a digital image file comprising a plurality of digital channel datasets corresponding to a plurality of image channels, wherein the digital image file includes an image region and a non-image region; determining a boundary for the image region; for each channel in the plurality of channels, generating a buffer region by using the digital channel datasets, identifying a plurality of pixels along the boundary and for each pixel of the plurality of pixels, copying the pixel one or more times outward into the non-image region in digital memory; and storing a modified digital image file that includes the buffer region.

Abstract: The invention provides an apparatus, system and method for the comparison between songs to analyze how similar each song is to the next using a new system of classification. A new and improved method and apparatus for the weighting of these comparable variables between songs in order to find the best objective criteria for finding a likeness between songs is disclosed. The time complexity associated with the computation of the optimal path for smoothing a playlist of songs is substantially reduced using the system of classification.

Abstract: According to an embodiment, an image processing apparatus includes an interface that receives a first image and a second image. A processor extracts a first character string appearing on the first image, and a second character string appearing on the second image. The processor determines that both the first character string and the second character string correspond to a photographed subject in the first image and second image. The processor cuts out each of the characters constituting the first character string, in units of a character. The processor cuts out each of the characters constituting the second character string, in units of a character. The processor calculates a first similarity degree group and a second similarity degree group. The processor recognizes a character string of the subject, based on the first similarity degree group and the second similarity degree group.

Abstract: A method includes receiving an image comprising a human head from an imaging device. Further, the method includes grouping one or more pixels of the image into one or more super pixels. Furthermore, the method includes generating a polar coordinate grid for the image, wherein the polar coordinate grid comprises one or more grid points. Additionally, the method includes labeling the one or more grid points as one of a face region, a hair region, and a background region based on tiered structure constraint. The method also includes labeling the one or more super pixels as one of the face region, the hair region, and the background region based on the one or more labels of the one or more grid points. Further, the method includes transmitting a notification comprising the one or more labels of the one or more super pixels.

Abstract: According to an embodiment, an image processing apparatus includes an interface that receives a first image and a second image. A processor extracts a first character string appearing on the first image, and a second character string appearing on the second image. The processor determines that both the first character string and the second character string correspond to a photographed subject in the first image and second image. The processor cuts out each of the characters constituting the first character string, in units of a character. The processor cuts out each of the characters constituting the second character string, in units of a character. The processor calculates a first similarity degree group and a second similarity degree group. The processor recognizes a character string of the subject, based on the first similarity degree group and the second similarity degree group.

Abstract: A method for reducing visual artifacts in a frame of a digital video signal, which is coded by blocks and then decoded, includes defining a block type is according to the coding method for a block selected according to a predetermined set of coding types. In the method filtering is performed to reduce visual artifacts due to a block boundary The filtering performed depends on block types of the frame in the environment of the block boundary.

Abstract: A method for segmenting a target image includes receiving the target image of an anatomical structure, registering a plurality of atlases to the target image, each of the atlases including an image and a plurality of labels corresponding to portions of the image, selecting a plurality of registered atlases, transferring the labels of selected registered atlases to the target image, combining the labels that are transferred to the target image using a fusion of a discriminative model and a generative model, and outputting a segmentation of the target image isolating the anatomical structure, wherein a segmentation of the target image is displayed.

Abstract: Provided is a method and system of tracking partially occluded objects using an elastic deformation model. According to an exemplary method and system, partially occluded vehicles are detected and tracked in a scene including side-by-side drive-thru lanes.

Abstract: Disclosed are systems, computer-readable mediums, and methods for creating a flexible structure description. To create the flexible structure description an image of a document of a particular document type that contains a table is received. An entry describing an item in the table is received. Title elements within the document are searched for based upon the entry. Data fields and anchor elements are detected for the entry. A flexible structure description for the particular document type is generated that includes a set of search elements for each data field in the image of the document and the title elements. The flexible structure description is matched against the image. Data from the image is extracted based upon the matching of the flexible structure description against the image.

Abstract: An image processing system automatically segments and labels an image using a hierarchical classification model. A global classification model determines initial labels for an image based on features of the image. A label-based descriptor is generated based on the initial labels. A local classification model is then selected from a plurality of learned local classification model based on the label-based descriptor. The local classification model is applied to the features of the input image to determined refined labels. The refined labels are stored in association with the input image.

Abstract: A method, apparatus and computer program product are provided for generating map data based on construction designs. In the context of a method, the method includes receiving a construction design, performing optical character recognition to detect characters within the construction design, extracting design features within the construction design, and generating map data based on the detected characters, and the extracted design features.

Abstract: An image is built. The building of an image includes selecting an image to be built, performing a hashing function on the image to produce a checksum, querying a registry for an existing image including the same checksum, and if no existing image including the same checksum is found, acquiring a parent image for the image to be built, building the image to produce a built image, adding the checksum for the image to the built image, and storing the built image in the registry.

Abstract: This disclosure provides an exemplary method and system for extracting structured label and value pairwise textual data from a textual document. According to an exemplary method, initially a layout analysis is performed resulting in one or more alternatives for grouping and ordering the textual elements of interest. Next, textual elements are tagged as including a label term, a value term or a label and value term. Finally, a sequence-based method is applied to the tagged elements to generate one or more sequence listings representative of the label and value pairwise data structure(s) and label:value pairwise data is extracted.

Abstract: In an image classifying apparatus, an image is designated as a designated image from among plural candidate images, the designated image being one of images to be contained in at least one group, attribute information is obtained from the designated image, selecting conditions of images are decided based on the attribute information obtained from the designated image, at least one image is selected as a selected image to be contained in the group, from among the plural candidate images based on the selecting conditions, and both the designated image and the selected images are classified into the group.

Abstract: A method registers a source image with a target image, wherein the images are deformable, by first measuring dissimilarity between the source image and the target image. The dissimilarity minimized using a discrete energy function. At multiple scales, multi-scale Markov random field registration is applied to the source and target images to determine a deformation vector field. Then, the target image is warped according to the deformation field vector to obtain a warped target mage registered to the source image.

Abstract: A processing system for an input device includes a sensor module and a determination module. The sensor module includes sensor circuitry coupled to transmitter electrodes and receiver electrodes. The sensor module is configured to transmit transmitter signals with the transmitter electrode and receive resulting signals with the receiver electrodes. The determination module is configured to obtain a set of measurements for an input object detected in a sensing region of the input device from a capacitive image, and perform, based on the set of measurements, a multistage classification to obtain a classification result identifying a type of the input object. wherein the capacitive image is generated based on the resulting signals.

Abstract: Disclosed are various embodiments labeling objects using multi-scale partitioning, rare class expansion, and/or spatial context techniques. An input image may be partitioned using different scale values to produce a different set of superpixels for each of the different scale values. Potential object labels for superpixels in each different set of superpixels of the input image may be assessed by comparing descriptors of the superpixels in each different set of superpixels of the input image with descriptors of reference superpixels in labeled reference images. An object label may then be assigned for a pixel of the input image based at least in part on the assessing of the potential object labels.

Abstract: Classification of image regions comprises: recursively partitioning an image into a tree of image regions having the image as a tree root and at least one image patch in each leaf image region of the tree, the tree having nodes defined by the image regions and edges defined by pairs of nodes connected by edges of the tree; assigning unary classification potentials to nodes of the tree; assigning pairwise classification potentials to edges of the tree; and labeling the image regions of the tree of image regions based on optimizing an objective function comprising an aggregation of the unary classification potentials and the pairwise classification potentials.

Abstract: A method of summarizing a document includes a step of extracting one or more sections of the document. The method also includes a step of separating at least one of the one or more extracted sections into one or more subsections based at least in part on a conjunctive structure of the section, wherein each subsection comprises one or more terms. The method also includes steps of determining whether one or more terms within a designated set of terms are present within at least one of the one or more subsections and, responsive to a determination that one or more terms within the designated set of terms are present within at least one of the one or more subsections, removing the one or more terms from the one or more subsections. The method also includes a step of aggregating at least a portion of the one or more sections into a summary of the document.

Abstract: A social analytic system may collect social signals from different social network accounts. The social signals may be associated with different ecosystems. Time series data may be generated from the social signals and the time series data may be filtered to remove at least some generic or unrelated trends. Different data sets from the time series data may be associated with different ecosystem metrics. The social analytic system may compare different filtered time series data sets to identify different ecosystem events. For example, the comparisons may be used to identify highly correlated ecosystem metrics and ecosystem anomalies, and predict ecosystem events.

Abstract: A method of generating an attribute map for processing an image. An image encoded into a plurality of segments is received. Each segment is characterized by a position within a base region associated with the image and is encoded according to a corresponding encoding format. A segment from the plurality of segments is decoded into a sub-image. The sub-image is associated with a combination operation and an attribute value derived based on at least one of a segment decoding format and a segment encoding format. The combination operation determines a process of combining the sub-image with the underlying base region. An information-carrying pixel value in the sub-image is determined based on the sub-image pixel content and the attribute value associated with the sub-image. The attribute map for processing the image is generated using the attribute value associated with the sub-image for at least one pixel of the attribute map.

Abstract: A system and methods for progressive feature evaluation of an electronic document image to identify user supplied elements is disclosed. The system includes a controller in communication with a storage device configured to receive and accessibly store a generated plurality of candidate images. The controller is operable to analyze the electronic document image to identify a first feature set and a second feature set, wherein each of the first and second feature sets represent a different form feature, compare the first feature set to the second feature set, and define a third feature set based on the intersection of the first and second feature sets, wherein the third feature sets represents the user provided elements.

Abstract: Techniques are disclosed for virtual scene generation. An image depicting a scene and annotated by a sparse set of labels is received. A dense set of labels annotating the image and a density map associated with the image are generated based on the sparse set of labels. A virtual scene is generated based on the dense set of labels and the density map, and the virtual scene is output.

Abstract: Embodiments provide a filtering method and apparatus for recovering an anti-aliasing edge. The method includes: calculating, for a nonlinearly filtered image, a variance value of an area surrounding a current pixel used for identifying an edge of the image; and re-filtering the current pixel when the variance value is greater than a predefined first threshold value, so as to recover an anti-aliasing edge. With the embodiments of the present application, aliasing noises resulted from nonlinear transform may be alleviated, thereby recovering an anti-aliasing edge of a natural object.

Abstract: The disclosure relates to the field of image processing technologies and particularly to a method and device for determining an image offset and a storage medium. A method according to an embodiment of the disclosure includes: dividing equally an acquired target image into a number M*N of first sub-areas and dividing equally an acquired image to be aligned into a number M*N of second sub-areas; determining an area offset between each first sub-area and a corresponding second sub-area; and determining an image offset between the target image and the image to be aligned according to the determined multiple area offsets.

Abstract: An apparatus includes a protrusion determination unit that determines a protrusion of print data with respect to a data arrangement area, an attribute determination unit that determines whether the attribute of the print data is text or non-text, a segmentation unit that segments text data into block data that can be arranged on at least two or more of data arrangement areas adjacent to each other when the print data is text data, and an arrangement order determination unit that determines an arrangement order of the plurality of block data after segmentation processing with respect to the plurality of data arrangement areas based on the arrangement order of text data prior to segmentation processing.

Abstract: An image processing device has an affinity calculating unit that handles each pixel in an image or each region including a plurality of pixels joined together, as a unit component, and determines a first affinity between each of the unit components and one of the unit components which is located in a periphery region of the image, a foreground region identifying unit that identifies, as a foreground region, a region made up of one or more of the unit components whose first affinities calculated by the affinity calculating unit are lower than a preset threshold, and a saliency measure output unit that determines a second affinity between each of the unit components in the image and the foreground region identified by the foreground region identifying unit, and outputs the second affinity as a saliency measure for each unit component.

Abstract: A device and method for dynamically adapting spatial resolution for imager Fourier transform spectrometers makes it possible to acquire data in interferogram mode and image mode on survey points for an observed scene, each survey point being associated a matrix of macro-pixels and defined by a plurality of zones. For each survey point, analysis of the content of each zone is carried out on the basis of data of the image mode. Classification into clear zone or non-clear zone is carried out as a function of proportion of cloud, and clear-pixel data are generated on the basis of the sum of the data of the macro-pixels of the clear zone class. Survey point data are generated on the basis of the sum of the data of all the macro-pixels of the matrix associated with the survey point. The survey point and clear pixel data streams are transmitted to the ground.

Abstract: Various embodiments of methods and apparatus for feature point localization are disclosed. An object in an input image may be detected. A profile model may be applied to determine feature point locations for each object component of the detected object. Applying the profile model may include globally optimizing the feature points for each object component to find a global energy minimum. A component-based shape model may be applied to update the respective feature point locations for each object component.