Abstract: A method and apparatus for processing a depth image that removes noise of a depth image may include a noise estimating unit to estimate noise of a depth image using an amplitude image, a super-pixel generating unit to generate a planar super-pixel based on depth information of the depth image and the noise estimated, and a noise removing unit to remove noise of the depth image using depth information of the depth image and depth information of the super-pixel.

Abstract: In selected embodiments, one or more wearable mobile devices provide videos and other sensor data of one or more participants in an interaction, such as a customer service or a sales interaction between a company employee and a customer. A computerized system uses machine learning expression classifiers, temporal filters, and a machine learning function approximator to estimate the quality of the interaction. The computerized system may include a recommendation selector configured to select suggestions for improving the current interaction and/or future interactions, based on the quality estimates and the weights of the machine learning approximator.

Abstract: A system for oil storage tank monitoring, comprising an extraction module and an analysis module, wherein an extraction module is utilized to determine information from an oil storage tank image and an analysis module is utilized to perform operations on extracted information, such as for determining measurements or values of oil storage tanks.

Abstract: A process is provided for estimating a three-dimensional (3D) radiance field of a combustion process in an enclosure. An on-line intensity-temperature calibration is performed based on an association between an intensity of an image pixel and an actual temperature associated with a selected region in the enclosure. The intensity of the corresponding image is transformed to a radiance image based on settings of an image-capturing device and the on-line calibration. A registration and alignment estimation of the image is performed based on positional information of the enclosure. The radiance image is aligned based on the registration estimation. The 3D radiance field having voxels of the enclosure is estimated based on a two-dimensional to 3D transforming of the aligned radiance images.

Abstract: Annotating and classifying an image based on a user context includes determining a location data of an object captured in an image, determining an attribute data of the object, obtaining sensor data from sensors that are associated with the location data based on the attribute data, determining a recommended user context from one or more predefined user contexts based on a comparison of the location data, the attribute data, and the sensor data with location data, attribute data, and sensor data of one or more images associated with the one or more predefined user contexts, determining a recommended class of the captured image based on the recommended user context, selecting one or more annotation data from the location data, the attribute data, and the sensor data based on the recommended class or the recommended user context, and annotating the image with the one or more annotation data.

Abstract: Facial recognition algorithms may identify the faces of one or more people in a digital image. Multiple types of communication may be available for the different people in the digital image. A user interface may be presented indicating recognized faces along with the available forms of communication for the corresponding person. An indication of the total number of people available to be communicated with using each form of communication may be presented. The user may have the option to choose one or more forms of communication, causing the digital image to be sent to the recipients using the selected forms of communication. An individual may have provided information for facial recognition of the individual to a service. Based on the information, the service may recognize that the individual is in an uploaded picture and send the digital image to the user account of the individual.

Abstract: Disclosed are systems and methods for configuring a vision detector, wherein a training image is obtained from a production line operating in continuous motion so as to provide conditions substantially identical to those that will apply during actual manufacturing and inspection of objects. A training image can be obtained without any need for a trigger signal, whether or not the vision detector might use such a signal for inspecting the objects. Further disclosed are systems and methods for testing a vision detector by selecting, storing, and displaying a limited number of images from a production run, where those images correspond to objects likely to represent incorrect decisions.

Abstract: According to one embodiment, an electronics device generates first and second index information, the first index information including codes of character strings corresponding to strokes, the second index information including characteristic quantity of strokes. The device executes at least either one of a first search and a second search, according to character string likelihood of first strokes which is a search key. The first search is performed by using the second index information and characteristic quantity of the first strokes. The second search is performed by using the first index information and a code of a character string corresponding to the first strokes.

Abstract: A face is detected and identified in a digital image. A weight is calculated and assigned to the detected face based on characteristics of the face and social media connections between the person identified from the face and a target viewer of the digital image. One or more image effects are applied to the digital image to visually distinguish the detected face from other parts of the digital image and/or in relation to other faces detected in the image.

Abstract: Provided is an apparatus and method for extracting feature information of an image using a scale-invariant feature transform (SIFT) algorithm. The apparatus may include a first interface configured to generate one or more tile images from a first source image stored in a particular memory, such as a high-capacity short-term memory, and a feature information extractor configured to receive the generated one or more tile images and to respectively extract feature information from each of the one or more input tile images, where the first interface may be configured to generate the one or more tile images by selectively dividing the first source image into the one or more tile images based on a horizontal resolution of the first source image.

Abstract: Embodiments include methods, devices, software, and systems for identifying a person based on relatively permanent pigmented or vascular skin mark (RPPVSM) patterns in images. Locations of RPPVSMs in different images of people are point matched, and a correspondence probability that the point matched RPPVSMs are from different people is calculated. Other embodiments are also described. Other embodiments are also described and claimed.

Abstract: The invention provides a method of using machine vision to recognize text and symbols, and more particularly traffic signs.

Abstract: The system includes a 3D feature detection module and 3D recognition module 202. The 3D feature detection module processes 3D surface map of a biometric object, wherein the 3D surface map includes a plurality of 3D coordinates. The 3D feature detection module determines whether one or more types of 3D features are present in the 3D surface map and generates 3D feature data including 3D coordinates and feature type for the detected features. The 3D recognition module compares the 3D feature data with biometric data sets for identified persons. The 3D recognition module determines a match between the 3D feature data and one of the biometric data sets when a confidence value exceeds a threshold.

Abstract: Body art, such as tattoos, are integrated or extended onto clothing. Clothing patterns may be integrated or extended onto body art, such as temporary tattoos. In one embodiment, a computer analyzes an image of a body with a tattoo and generates an image suitable for application to clothing. When applied to clothing, the image displays the portion of the tattoo that is covered by the clothing, or may extend the appearance of the tattoo from the skin to the adjacent clothing.

Abstract: According to an aspect of the present invention, there is provided a pattern matching method of detecting an image of a detection target from a search image, comprising: obtaining a reference image of the detection target; generating the model edge image on a basis of the reference image; generating the edge extraction domain that is specified as a portion where the model edge image can exist by overlying a plurality of the model edge images obtained with at least one of rotating the model edge image within a predetermined range around a rotation center of the model edge image and translating the model edge image within a predetermined range; and performing pattern matching between the model edge image and the search edge image.

Abstract: An image processing apparatus includes a first acquiring unit that acquires an image to be processed; a setting unit that sets multiple partial image areas in the image to be processed; a second acquiring unit that acquires a first classification result indicating a possibility that an object of a specific kind is included in each of the multiple partial image areas; and a generating unit that generates a second classification result indicating a possibility that the object of the specific kind is included in the image to be processed on the basis of the first classification result of each of the multiple partial image areas.

Abstract: The present disclosure is directed towards methods and systems for capturing artifact-free biometric images of an eye. The eye may be in motion and in the presence of partially-reflective eyewear. The method may include acquiring, by a first sensor, a first image of an eye while the eye is illuminated by a first illuminator. The first image may include a region of interest. The first sensor may be disposed at a fixed displacement from the first illuminator and a second sensor. The second sensor may acquire, within a predetermined period of time from the acquisition of the first image, a second image of the eye. The second image may include the region of interest. An image processor may determine if at least one of the first and second images include artifacts arising from one or both of the first illuminator and eyewear, within the region of interest.

Abstract: Object recognition systems, methods, and devices are provided. Candidate objects may be detected. The candidate objects may be verified as depicting objects of a predetermined object type with verification tests that are based on comparisons with reference images known to include such objects and/or based on context of the candidate objects. The object recognition system may identify images in a social networking service that may include objects of a predetermined type.

Abstract: Inputs of a plurality of images constituting a group of images of items regarded as non-defective items are previously accepted and stored, and a defect threshold for detecting a defective portion of an inspection object is set based on the plurality of stored images. A defect amount to be compared with a determination threshold for making a non-defective/defective determination on the inspection object is calculated with respect to each of the plurality of stored images based on the set defective threshold, and whether or not each of the calculated defect amounts is an outlier is tested by use of at least one of a parametric technique and a non-parametric technique. Outlier information for specifying an image whose defect amount has been tested to be the outlier is displayed and outputted.

Abstract: A method of analyzing a depth image in a digital system is provided that includes detecting a foreground object in a depth image, wherein the depth image is a top-down perspective of a scene, and performing data extraction and classification on the foreground object using depth information in the depth image.