Abstract: Techniques for ranking the importance of various attributes associated with various product categories are described. According to various embodiments, product category information identifying various products in a particular product category in the inventory of a marketplace website is accessed. The product category information may further identify a set of attributes associated with the products in the particular product category. An importance value associated with each of the attributes is then calculated, the importance values indicating an importance of each of the attributes for uniquely identifying the products in the product category. Thereafter, each of the attributes are ranked, based on the importance value associated with each of the attributes.

Abstract: In one embodiment of the present invention, a graphics processing unit (GPU) is configured to detect an object in an image using a random forest classifier that includes multiple, identically structured decision trees. Notably, the application of each of the decision trees is independent of the application of the other decision trees. In operation, the GPU partitions the image into subsets of pixels, and associates an execution thread with each of the pixels in the subset of pixels. The GPU then causes each of the execution threads to apply the random forest classifier to the associated pixel, thereby determining a likelihood that the pixel corresponds to the object. Advantageously, such a distributed approach to object detection more fully leverages the parallel architecture of the PPU than conventional approaches. In particular, the PPU performs object detection more efficiently using the random forest classifier than using a cascaded classifier.

Abstract: An image processing method includes a step of acquiring a measured image G1B measured from a semiconductor device S and a first pattern image G2B showing a pattern of the semiconductor device S, a step of acquiring a reference measured image G3B measured from a reference semiconductor device SR being the semiconductor device S or a semiconductor device different from the semiconductor device S and a second pattern image G4B showing a pattern of the reference semiconductor device SR, a step of acquiring matching information indicating a correlation of the first pattern image G2B and the second pattern image G4B, and a step of determining a difference of the measured image G1B and the reference measured image G3B based on the matching information to acquire a comparative image G5B.

Abstract: A method of categorizing data points is described which, when combined with a clustering algorithm, provides groupings of data points that have an improved confidence interval. The method can be used to find an optimal number of groupings for a dataset, which in turn allows a user to categorize a group of data points for processing. In some examples, a dataset containing a number of data points may be accessed. Additionally, in some aspects, groupings of data points within the dataset may be grouped based at least in part on similarities between the data. Further, a number of groupings of data points may be adjusted so that the distance between the data points within one or more groupings of data points may fit within a confidence level.

Abstract: A matching system and related methods for matching an input item or input item set with stored items and/or stored item sets, and methods for manufacturing and using same.

Abstract: An object-recognition method and system employing Bayesian fusion algorithm to reiteratively improve probability of correspondence between captured object images and database object images by fusing probability data associated with each of plurality of object image captures.

Abstract: A system may use a configurable detected to identify a feature in a received image and an associated candidate tag based on user-defined items of interest, and to determine an associated accuracy metric. Moreover, based on the accuracy metric, costs of requesting the feedback from one or more individuals and a feedback threshold, the system may use a scheduler to selectively obtain feedback, having a feedback accuracy, about the candidate tag from the one or more individuals. Then, the system may generate a revised tag based on the feedback when the feedback indicates the candidate tag is incorrect. Next, the system presents a result with the feature and the candidate tag or the revised tag to another electronic device. Furthermore, based on a quality metric, the system may update labeled data that are to be used to retrain the configurable detector.

Abstract: A method and system for detecting anomalies in video footage. A training dictionary can be configured to include a number of event classes, wherein events among the event classes can be defined with respect to n-dimensional feature vectors. One or more nonlinear kernel function can be defined, which transform the n-dimensional feature vectors into a higher dimensional feature space. One or more test events can then be received within an input video sequence of the video footage. Thereafter, a determination can be made if the test event(s) is anomalous by applying a sparse reconstruction with respect to the training dictionary in the higher dimensional feature space induced by the nonlinear kernel function.

Abstract: An image processing device 20 acquires a plurality of captured images obtained by imaging a sample including a target cell at a plurality of different focal distances, sets an object region for detecting the target cell in at least two of the plurality of captured images, and performs determination about whether or not the target cell is included in the object region based on whether or not an image feature quantity based on change in image information in an image in-plane direction and a focal distance direction of an image in the object region in the plurality of captured images satisfies conditions defined for the target cell.

Abstract: A configured machine performs image matching and retrieval of natural images that may depict logos. The machine generates and uses color-localized spatial masks, which may be computationally less expensive than spatial verification techniques. Key points are detected within images that form a reference database of images. Local masks are defined by the machine around each key point based on the scale and orientation of the key point. To utilize color information presented in logo images, ordered color histograms may be extracted by the machine from locally masked regions of each image. A cascaded index may then be constructed for both visual descriptors and color histograms. For faster matching, the cascaded index maps the visual descriptors and color histograms to a list of relevant or similar images. This list may then be ranked to generate relevant matches for an input query image.

Abstract: A method and apparatus for determining a class reference data record for classifying documents of value includes creating a class reference image using a multiplicity of reference images of already classified documents of value in the same class; and creating a class reference data record having at least one class reference parameter using the class reference image. The pixel and intensity values of the respective pixel in the class reference image are a function of the pixel values of the relevant pixels in the multiplicity of reference images of already classified documents of value in the same class. The method includes determining at least one quantitative property of a document of value to be classified; and classifying the document of value to be classified on the basis of a comparison between the quantitative property of the document of value to be classified and the class reference data record.

Abstract: A position tracking apparatus includes: a depth image obtaining unit for obtaining a depth image; a database created by collecting depth images received from the depth image obtaining unit; a feature extracting unit for extracting features from each pixel of the depth image; a classifier training unit for training a classifier in order to determine the position of the parts of the body by receiving a feature set for each part of the body as inputs which are extracted by using the feature extracting unit from all of the depth images in the database; and a position determination unit for extracting features for each pixel of the depth image received by the depth image obtaining unit using the feature extracting unit in a state in which the classifier training unit trains the classifier, and for tracking the three-dimensional position of each part of the body through the classifier.

Abstract: The present invention relates to the method for updating scene model and video surveillance. A method is provided for updating a scene model in a video which is composed of a plurality of visual elements, comprising: a classifying step for classifying the visual elements in a scene into stationary visual elements and moving visual elements according to their appearance change rates; a border determining step for determining borders from the scene according to a spatial distribution information of the stationary visual elements and the moving visual elements; and an updating step for updating the scene model according to the determined borders in said scene.

Abstract: The present invention relates to method for processing an image, said image representing a rock sample. The method comprises receiving said image; for each current threshold value in a plurality of threshold values determining a threshold image, the threshold image having pixels associated with a first color value or a second color value, the determination of the threshold image being based on the received image and the current threshold; and for each current connected set of pixels associated with the first color value in the threshold image, determining a connected volume or a connected surface area of said current connected set of pixels. In addition, the method comprises determining numbers of connected sets of pixels associated with the first color value for a plurality of connected volumes or connected surface areas and for a plurality of thresholds.

Abstract: A method and system are provided. The method includes extracting shapelets from each of a plurality of time series dimensions of multi-dimensional time series data. The method further includes building a plurality of decision-tree classifiers, one for each time series dimension, responsive to the shapelets extracted therefrom. The method also includes generating a pairwise similarity matrix between respective different ones of the plurality of time series dimensions using the shapelets as intermediaries for determining similarity. The method additionally includes applying a feature selection technique to the matrix to determine respective feature weights for each of shapelet features of the shapelets and respective classifier weights for each of the decision-tree classifiers that uses the shapelet features.

Abstract: A method and a device for classifying samples of an image, each sample having a respective sample value the method comprising repartition of each of the samples into a plurality of index classes, wherein the index class allocated to each sample is determined in dependence upon the sample value of the said sample with respect to the sample values of first neighboring samples of said sample according to a given direction and further in respect to the sample values of the first neighboring samples with respect to the sample values of samples neighboring the first neighboring samples according to the said given direction. Processing each sample according to the respective allocated index class.

Abstract: Techniques are disclosed for extracting micro-features at a pixel-level based on characteristics of one or more images. Importantly, the extraction is unsupervised, i.e., performed independent of any training data that defines particular objects, allowing a behavior-recognition system to forgo a training phase and for object classification to proceed without being constrained by specific object definitions. A micro-feature extractor that does not require training data is adaptive and self-trains while performing the extraction. The extracted micro-features are represented as a micro-feature vector that may be input to a micro-classifier which groups objects into object type clusters based on the micro-feature vectors.

Abstract: A computer-implemented technique can receive a plurality of photos and automatically select a subset of the plurality of photos having a high degree of representativeness by jointly maximizing both photo quality and photo diversity to obtain a photo album. The technique can determine one or more clusters for the photo album using a hierarchical clustering algorithm, and store the photo album according to the one or more clusters. The technique can control the manner in which the photo album is displayed using the one or more clusters. The technique can adjust at least one of the one or more clusters and the automatic photo album generation based on user input. The user input can include at least one of adding, deleting, and moving a photo with respect to the one or more clusters. The technique can then re-cluster, automatically generate a new photo album, and/or adjust the presentation.

Abstract: A vision system of a vehicle includes a camera disposed at a vehicle and having a field of view exterior of the vehicle. The camera includes an imaging array having a plurality of photosensing elements arranged in a two dimensional array of rows and columns. The imaging array includes a plurality of sub-arrays comprising respective groupings of neighboring photosensing elements. An image processor is operable to perform a discrete cosine transformation of captured image data, and a Markov model compares at least one sub-array with a neighboring sub-array. The image processor is operable to adjust a classification of a sub-array responsive at least in part to the discrete cosine transformation and the Markov model.

Abstract: Neural network image curation techniques are described. In one or more implementations, curation is controlled of images that represent a repository of images. A plurality of images of the repository are curated by one or more computing devices to select representative images of the repository. The curation includes calculating a score based on image and face aesthetics, jointly, for each of the plurality of images through processing by a neural network, ranking the plurality of images based on respective said scores, and selecting one or more of the plurality of images as one of the representative images of the repository based on the ranking and a determination that the one or more said images are not visually similar to images that have already been selected as one of the representative images of the repository.

Abstract: According to one embodiment, an electronic device includes circuitry. The circuitry is configured to: receive first image data for displaying a first image that comprises a plurality of regions, modify a first region of the first image by adding at least a high-frequency component based on second image data, the first region of the first image including fewer noise components than a first threshold, and output third image data for displaying the first image that comprises the plurality of regions and the modified first region.

Abstract: An image recognition system includes a partial area determination unit for determining a plurality of learning partial areas in a learning image including an object to be recognized, a partial area set generation unit for generating a learning partial area set including the learning partial area and a plurality of peripheral areas included in a predetermined range with reference to the learning partial area, and a learning unit for selecting an area including an image suitable to be determined as an object to be identified included the object to be recognized from a plurality of areas included in the learning partial area set, to learn a classifier so as to determine a likelihood that the image included in the area is the object to be identified to be higher based on a feature amount related to the selected area.

Abstract: An apparatus includes a unit configured to acquire an object shape regression model, which comprises an average object shape, a plurality of regression functions and a plurality of feature selection maps; a unit configured to set an initial object shape for the object image based on the average object shape; a unit configured to calculate at least one feature vector with respect to a plurality of feature points; a unit configured, for each coordinate of the plurality of feature points, to select feature fragments from the calculated feature vector based on a corresponding one of the plurality of feature selection maps and assemble the feature fragments into a sub feature vector; and a unit configured, for at least one coordinate of at least one feature point, to predict a coordinate increment based on the corresponding sub feature vector and a corresponding one of the plurality of regression functions.

Abstract: Search systems and computer-implemented search methods are described. In one aspect, a search system includes a communications interface configured to access a plurality of data items of a collection, wherein the data items include a plurality of image objects individually comprising image data utilized to generate an image of the respective data item. The search system may include processing circuitry coupled with the communications interface and configured to process the image data of the data items of the collection to identify a plurality of image content facets which are indicative of image content contained within the images and to associate the image objects with the image content facets and a display coupled with the processing circuitry and configured to depict the image objects associated with the image content facets.

Abstract: Provided are a method and an apparatus for determining an intra prediction mode. The method includes: splitting a maximum coding unit into coding units of a first size; acquiring an edge direction of each of the coding units of the first size; grouping adjacent coding units of the first size based on uniformity of edge directions of the adjacent coding units of the first size; determining a size range of coding units to be applied to intra prediction on the intra prediction coding unit group; performing the intra prediction on the intra prediction coding unit group by using coding units having sizes included in the determined size range; and determining a coding unit for the intra prediction and an intra prediction mode of the coding unit by comparing costs according to the performed intra prediction on the coding units having the sizes included in the size range.

Abstract: In one embodiment, a method for generating textured graphics includes identifying border colors of pixels around two texture images and generating arrangements of border texels from the border colors that are positioned next to the two images in a texture atlas. The method includes generating mip-maps of the texture atlas with texels in the jump level assigned with the border color of the corresponding textures in the full-resolution texture atlas instead of the averaged color of the textures that would be assigned using a traditional mip-map process. The method includes storing the texture atlas including the two texture images and the border texels in a memory for use in generating repeated textures on an object in a virtual environment using at least one of the texture images with a mip-map without seam artifacts between the repeated textures.

Abstract: An image forming apparatus includes a character distinguishing unit, a character string extraction unit, a character specifying unit, a determination unit, an overwrite image data generation unit, and a control unit. The overwrite image data generation unit generates, if it is determined by the determination unit that there is a plurality of matches, overwrite image data that has been overwritten such that a character specified by the character specifying unit is indistinguishable as a character. The control unit performs control using the overwrite image data generated by the overwrite image data generation unit so as to cause a printing unit to perform printing on a printed material set in a printed material setting unit.

Abstract: Automatically classifying multimedia content that is specific to a video production includes obtaining, from a video sensor embedded in a video capturing device that captures a video, a time series of location data, direction data, orientation data, and a position of the first user and the second user, identifying for any given duration in the video, corresponding first user data, corresponding second user data, and corresponding video sensor data, annotating the video with the corresponding first user data, the corresponding second user data, and the corresponding video sensor data to obtain an annotated multimedia content, performing a comparison of a data pattern of the annotated multimedia content with data patterns of a script stored in a database to obtain a recommended section for the annotated multimedia content, and automatically classifying the annotated multimedia content by associating the annotated multimedia content with the recommended section.

Abstract: A device is configured to receive an image of a person. The image may include an image of a face. The device may create an emoticon based on the image of the face. The emoticon may represent an emotion of the person. The device may detect a facial expression of the person at a time that a message is being generated. The device may determine the emotion of the person based on the detected facial expression. The device may identify the emoticon as corresponding to the emotion of the person. The device may add the emoticon to the message based on the emoticon corresponding to the emotion of the person. The device may output the message including the emoticon.

Abstract: Image-processing apparatus and methods to adaptively control a size and/or location of a visual search window used for feature matching in a machine-vision system are described. A search window controller may receive motion vector data and image recognition rate data, and compute a search window size and/or search window location based on the received data. The computed search window size may be a portion of an image frame. The motion vector data and image recognition rate data may be computed from one or more images in a video image sequence. By adaptively controlling search window size and location, an appreciable reduction in data processing burden for feature matching may be achieved.

Abstract: An electronic writing solution server includes a query engine and a user interface engine. The query engine receives a search query from a user, the search query specifying a field in at least one form and including a first stroke or text, retrieves clusters from a database, retrieves example strokes from the clusters at vary levels of similarity to the first stroke, receives user selections for strokes of interest from the example strokes, and retrieves information associated with the strokes of interest from the database. The user interface engine generates graphical data for presenting the retrieved information to the user.

Abstract: An object recognition system and method is provided which uses automated algorithmically determined negative training. Negative training with respect to a particular object classifier allows for more streamlined and efficient targeted negative training, enabling time and cost savings while simultaneously improving the accuracy of recognition based on the targeted negative training. According to certain additional aspects, embodiments of the invention relate to a system and method for structuring an object recognition engine to maximize recognition accuracy and emulate human visual cortex.

Abstract: An apparatus and method are provided including a first segmenter and a second segmenter. The first segmenter is configured to generate a first segmentation result from a medical image using a first segmentation parameter for a candidate lesion. The second segmenter is configured to determine a target lesion to segment from among the candidate lesion based on the first segmentation result, and generate a second segmentation result using a second segmentation parameter to segment the target lesion.

Abstract: A method of de-blocking video data is disclosed. The video data encoding color channels in a 4:2:2 format is received. The video data is encoded in a quad-tree. A plurality of transform units is generated for one of the color channels, each of the transform units including at least one transform. A distance from an edge of one of the transform units to a boundary of a transform of the transform unit is determined. An edge flag for the transform unit is determined, the edge flag indicating the determined distance. De-blocking is applied to the transform units of the video data according to the determined edge flag.

Abstract: There is provided a display control apparatus including an external memory accommodating unit for accommodating a removable external memory; a database recognizing unit for recognizing a database stored in the external memory, the database being recorded with an image stored in the external memory and information related to the image in correspondence to each other; a display method setting unit for setting either display method of a stored first display method or a second display method of displaying the image stored in the external memory without using the database based on a recognition result of the database recognizing unit; and a display controlling unit for displaying the image stored in the external memory by the display method based on the set display method.

Abstract: Disclosed herein are a system, method and architecture for media content enrichment. A visual object detector is trained using a training data set and an existing visual object detector. The newly-adapted visual object detector may be used to detect a visual object belonging to a class of visual object. The existing object detector that is used to train the adapted object detector detects a class of visual objects different from the visual object class detected by the adapted object detector. A media content item depicting a visual object detected using the adapted object detector may be associated with metadata, tag or other information about the detected visual object to enrich the media content item.

Abstract: Techniques for setting depth values for invalid measurement regions of depth images are described herein. A computing device may set the depth values based on evaluations of depth values of neighboring pixels and of corresponding pixels from time-adjacent depth images. Alternately or additionally, the computing device may utilize a texture image corresponding to the depth image to identify objects and may set depth values for pixels based on depth values of other pixels belonging to the same object. After setting the depth values, the computing device may normalize the depth values of the pixels. Further, the computing device may generate reduced representations of the depth images based on a depth reference model or a depth error model and may provide the reduced representations to an encoder.

Abstract: In one aspect, the description relates to identifying complementary data objects, including providing a plurality of data objects, applying a clustering algorithm for grouping at least some of the data objects into two or more clusters, for each of the clusters, calculating a cluster center, calculating, for at least a first one of the cluster centers, a complementary cluster center, determining a second cluster center of a second cluster, the second cluster center being determined as the one of the cluster centers having the smallest distance in respect to the complementary cluster center, selecting at least one data object of the determined second cluster. Other features and aspects may be realized, depending upon the particular application.

Abstract: Present embodiments contemplate systems, apparatus, and methods to improve feature generation for object recognition. Particularly, present embodiments contemplate excluding and/or modifying portions of images corresponding to dispersed pixel distributions. By excluding and/or modifying these regions within the feature generation process, fewer unfavorable features are generated and computation resources may be more efficiently employed.

Abstract: Various implementations of the invention perform facial recognition on a target image compared against a color image from an image gallery, where the target image was acquired by an infrared-sensitive camera of a target that was illuminated with infrared light. According to various implementations of the invention, a blue component and a green component of the pixels in the color image are suppressed or eliminated, and facial recognition is performed between the target image and the color-suppressed image.

Abstract: A method and system for estimating a fingerprint pose are provided. The method includes following steps. A fingerprint image to be estimated is input, features are extracted from the fingerprint image to be estimated in all possible center positions and directions, each feature is scored via a classifier and a center position and direction corresponding to a feature of a highest score is selected as a fingerprint pose estimation result.

Abstract: A method reduces processing time required to identify locations burned by fire by receiving a feature value for each pixel in an image, each pixel representing a sub-area of a location. Pixels are then grouped based on similarities of the feature values to form candidate burn events. For each candidate burn event, a probability that the candidate burn event is a true burn event is determined based on at least one further feature value for each pixel in the candidate burn event. Candidate burn events that have a probability below a threshold are removed from further consideration as burn events to produce a set of remaining candidate burn events.

Abstract: The disclosed technology includes techniques for improved content coverage in automatically-generated content summaries. The technique may include clustering a set of input content, determining diffusion for each cluster, and selecting representatives of each cluster to optimize other secondary metrics. Various types of input content may be used, including groups of images, video clips, or other multimedia content. Contiguous content may be manually or programmatically divided into discrete portions before clustering, for example, a lengthy video divided into a number of short clips. In some implementations, the disclosed technique may be implemented effectively on a mobile device. In other words, the processing required may be computationally feasible for execution on a smartphone or similar device.

Abstract: Systems and methods for displaying information on a wearable user device include determining calibration information based on user actions and information from images captured by a camera on a wearable user device. A field of view calibration may then be performed on a display engine in the wearable user device using the calibration information. Graphical information may then be displayed on a display device on the wearable user device according to a user field of view using the display engine in the wearable user. Field of view calibrations may be performed using calibration information that is based on: a user's head range of motion such that displayed graphical information conforms to that head range of motion, calibration objects by themselves that indicate a user's perspective in a user field of view, and calibration objects that allow the size of a user's hands to be determined and used to measure objects.

Abstract: Features are disclosed for detecting an event in input data using a cascade-based detection system. Detection of the event may be triggered at any stage of the cascade, and subsequent stages of the cascade are not reached in such cases. Individual stages of the cascade may be associated with detection thresholds for use in triggering detection of the event. The sequence of stages may be selected based on some observed or desired operational characteristic, such as latency or operational cost. In addition, the cascade may be modified or updated based on data received from client devices. The data may relate to measurements and determinations made during real-world use of the cascade to detect events in input data.

Abstract: A computer-implemented method and system for determining documents that are nearest to a query are provided herein. The method includes constructing a vantage point tree based on a number of document vectors. The method also includes searching the vantage point tree to determine a number of nearest neighbor document vectors to a query vector by removing a portion of the document vectors from the vantage point tree based on one or more vantage points for each of a number of nodes in the vantage point tree and a specified search radius centered about the query vector.

Abstract: The computational complexity of a denoising filter is reduced while suppressing a reduction in coding efficiency. In a video encoding/decoding apparatus which uses a loop filter, a deviation degree calculating unit calculates the degree of deviation between a denoising target pixel and a surrounding pixel using a decoded image. A template shape setting unit limits a template shape so that its size is proportional to the degree of deviation relative to the maximum value of the degree of deviation in the decoded image. When noise of the target pixel is removed, weighted in accordance with the degree of similarity between a template of the target pixel and a template of each of search points in a search shape and the weighted sum of pixel values at the search points, the loop filter performs template matching using the limited template shape and removes noise of the target pixel.

Abstract: A system and method for generating an ordered set of suggested profile images. The profile image application includes a scorer, a recognition engine, a trend analyzer, a suggestion module and a user interface engine. The scorer scores user images based on metadata, such as the number of people tagged in the image and the recency of the image. The recognition engine identifies people and objects in the image. The trend analyzer identifies trends. The suggestion module generates the ordered set of suggested profile images based at least in part on the score. The user interface engine generates a user interface that includes the ordered set of suggested profile images.

Abstract: Aspects of the subject matter described herein relate to functions used for retrieving image results based on search queries. More specifically, image search queries can be pre-grouped or classified based on visual and semantic similarity. For example, a pairwise image similarity value for a pair of queries can be computed based on one or more of the sum of all of the overlapping the image results, the sum of the image distances between all of the pairs of images in the image results, and the rank of each of the images in the image results. The pairwise image similarity values can then be used to generate image query clusters. Each image query clusters can include a set of queries with high pairwise image similarity values. In some examples, a distance function can be determined for each image query cluster. This data can be used to provide image results.

Abstract: A method of providing video feeds from a plurality of cameras to a plurality of screens including determining a plurality of constraints on a centralized processor processing the video feeds, determining a camera semantic classification for each of the plurality of cameras, determining historical events captured by each of the plurality of cameras, and providing at least one video feed to at least one of the screens according to the plurality of constraints on the centralized processor, the camera semantic classifications and the historical events.