Abstract: The present invention is a complete end-to-end video face recognition system. The invention performs a track-by-track labeling of all tracks within a video. A novel algorithm Mean Sequence SRC (MSSRC) is applied to a face track to perform video face recognition using a joint optimization to leverage all of the available video data and the knowledge that the face track frames belong to the same individual. Additionally the system constructs a probabilistic affinity graph combining appearance and co-occurrence similarities to model the relationship between face tracks in a video. Finally, using this relationship graph, random walk analysis is employed to propagate strong class predictions among similar face tracks, while dampening weak predictions.

Abstract: An image evaluation device (101) properly evaluating the graininess (or roughness) of an image is provided. The blurer (102) creates a second image b by blurring a first image a. The differentiator (103) creates a third image c presenting the difference between the first image a and second image b. The pixel value of each pixel of the third image c presents the difference in pixel value between the pixels at the same position in the first image a and second image b. The scanner (104) scans the pixels contained in the third image c, obtains the differences in pixel value between adjoining pixels, and obtains the respective probabilities of occurrence of the differences. The calculator (105) calculates the entropy from the obtained, respective probabilities of occurrence of the differences. The outputter (106) outputs the entropy as the evaluation value of the graininess (or roughness) of the first image.

Abstract: A system and method of video-based chew counting by receiving image frames from a video camera, determining feature points within the image frames from the video camera, generating a motion signal based on movement of the feature points across the image frames from the video camera, and determining a chew count based on the motion signal.

Abstract: A method for detecting the presence of a logo in a current picture of a video is disclosed. The method comprises: determining a current candidate logo bounding box in the current picture and a preceding candidate logo bounding box in a picture preceding the current picture; detecting the presence of a logo in the current picture if at least one of the following conditions is fulfilled: the distance between the centers of the current and the preceding candidate logo bounding boxes is below a first threshold or an overlap between the current and the preceding candidate logo bounding boxes is above a second threshold; or an overlap between the current candidate logo bounding box and a saliency mask of the current picture is above a third, wherein the saliency mask identifies regions around saliency peaks.

Abstract: A device may request, and a server may provide, access to information associated with an image based on trusted face recognition. The server may receive a first image, registration, and associated information from a first device. The first image may include feature data, which the server may extract to create a first feature template. The server may store, and correlate, the first feature template, registration and associated information for the first image. A second device may extract feature data from a second image, create, and provide to the server, a second feature template. The server may authenticate the second device. The server may compare the second feature template with stored feature templates to identify a match. The server may determine whether to share associated information for a match with, and return an access indication to, the second device based, in part, on the authentication.

Abstract: A method for ascertaining the position of an edge in or on an object surface region of interest by optical scanning. The reflectivity of the object surface region is evaluated. Light is emitted onto the object surface region under different illumination conditions, in particular different light incidence directions, and in each illumination condition a sequence S1 to Sn of camera images B is recorded. Each camera image B of a sequence S1 to Sn is recorded at another illumination intensity I. Subsequently in each case one reflectivity image R1 to Rn is produced from a plurality of or from all camera images B of a sequence S1 to Sn. Thereafter, a resulting reflectivity image E is produced from a plurality of or from all reflectivity images R1 to Rn by weighted addition, in which resulting reflectivity image E the position of an edge is determined.

Abstract: Provided are methods and systems for generating a binarized representation of a document using a mobile device configured to provide a Flashed and No-flashed image of the document. According to an exemplary method, the Flash and No-flash images are aligned and blended, where the resulting image includes the No-flash image where any hot spots are included in the Flash image. After blending, the blended image is binarized.

Abstract: The present disclosure proposes a method and an electronic device for detecting glare pixels of an image including an object and an electronic device using the same. The method includes the following steps. First, the image is processed to remove overexposed pixels from the image to generate a first image. The first image is processed to remove non-object pixels from the first image to generate an object pixel image. The glare pixels are detected from the object pixel image according to a saturation distribution threshold and an intensity distribution threshold. The present disclosure is able to adjust the saturation distribution threshold and the intensity distribution threshold dynamically and adaptively to satisfy various kinds of objects and ambient lighting conditions so as to improve the probability of successful detection and reduce the probability of false alarms on the glare pixels.

Abstract: According to one embodiment, a plurality of moving objects is detected from a plurality of frames acquired in time series. Each of the moving objects is corresponded among the frames. A tracklet of each moving object corresponded is extracted and stored. A frame to calculate a position of a moving object is set to a notice frame. The frames are grouped into a first block including at least the notice frame, a second block positioned before the first block in time series, and a third block positioned after the first block in time series. A secondary tracklet included in the second block is acquired from the stored tracklets. The secondary tracklet is corresponded with tracklets included in the first block and the third block, based on a similarity between the secondary tracklet and each of the tracklets. The secondary tracklet is associated with the corresponded tracklets, as a tertiary tracklet.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for selecting training images. One of the methods includes determining, for each of a plurality of labels that each designate a respective food class of a plurality of food classes, a respective measure of importance. A respective sample size is determined for the label based on the respective measure of importance of the label. A number of training images are selected for each respective label according to the determined sample size for the label. A predictive model is trained using the selected training images as training data.

Abstract: A content selection apparatus selects a target content item which is a content item to be presented to a user, from a plurality of content items. The plurality of content items are sorted into a plurality of management units according to a creation date and time of each content item. The content selection apparatus includes a selection unit that selects a predetermined number of management units in ascending order of the number of sorted content items, and selects a content item that belongs to the selected management units as the target content item.

Abstract: A contour/shape detection model may use relatively simple and efficient kernels to detect target edges in an object within an image or video. A co-occurrence probability may be calculated for two or more edge features in an image or video using an object definition. Edge features may be differentiated between in response to measured contextual support, and prominent edge features may be extracted based on the measured contextual support. The object may then be identified based on the extracted prominent edge features.

Abstract: Approaches for deciding what individuals in a population of visual system “neurons” are looking for using sparse overcomplete feature dictionaries are provided. A sparse overcomplete feature dictionary may be learned for an image dataset and a local sparse representation of the image dataset may be built using the learned feature dictionary. A local maximum pooling operation may be applied on the local sparse representation to produce a translation-tolerant representation of the image dataset. An object may then be classified and/or clustered within the translation-tolerant representation of the image dataset using a supervised classification algorithm and/or an unsupervised clustering algorithm.

Abstract: This disclosure presents methods, systems, computer-readable media, and apparatuses for optically tracking the location of one or more objects. The techniques may involve accumulation of initial image data, establishment of a dataset library containing image features, and tracking using a plurality of modules or trackers, for example an optical flow module, decision forest module, and color tracking module. Tracking outputs from the optical flow, decision forest and/or color tracking modules are synthesized to provide a final tracking output. The dataset library may be updated in the process.

Abstract: A computing device configured to gather utility metering information. In embodiments, the computing device may comprise a transceiver and an image processing module. The image processing module may be coupled to the transceiver and configured to receive an image utilizing the transceiver. The image may include one or more utility meters. The image processing module may be further configured to extract metering information from the one or more utility meters of the image and transmit this metering information to a utility provider associated with the utility meter Other embodiments may be described and/or claimed.

Abstract: An electronic device may include an array of finger sensing pixels and data acquisition circuitry coupled to the array. The data acquisition circuitry may be capable of acquiring finger biometric data from each sub-array of the array, and acquiring spoof detection data from at least one of the sub-arrays in an interleaved fashion with the finger biometric data.

Abstract: A bad pixel correction method for an image is provided. The bad pixel correction method includes the steps of determining whether each of pixels of a predetermined block in the image is a bad pixel sequentially; when a current pixel of the block is determined to be bad and a temperature value is higher than a threshold, re-determining that at least one previously determined pixel in the predetermined block to be bad; and correcting the current pixel and the previously determined pixel, wherein the previously determined pixel is adjacent to the current pixel.

Abstract: A system and a method for iris detection in a face region of a digital image are disclosed. In one embodiment, iris is detected in the face region of the digital image using eye features, such as a shape of the iris and/or luminance values of a sclera region.

Abstract: A system for enhancing an input image including receiving an input image and filtering the input image with a plurality of non-linear smoothing filters providing a respective plurality of filtered outputs. The system processes a plurality of the filtered outputs with respect to at least one of another of the filtered outputs and the input image to determine a plurality of detail layers. The system filters the plurality of detail layers with a plurality of non-linear smoothing filters providing a respective plurality of smoothed layers. The system adjusts the plurality of smoothed layers in such a manner that regions closer to an edge are enhanced to a lesser extent than regions farther from an edge and combining the adjusted the smoothed layers to provide an enhanced output image.

Abstract: Based on an image imaged by an imaging unit (2), feature points of a plurality of objects present around a vehicle are detected by a feature point detecting unit (35). A first distance from the feature point to the imaging unit (2) is calculated based on a temporal change of the feature point. A second distance from a certain object included in the plurality of objects present in the image to the imaging unit is calculated using some pixels of the certain object. The first distance of the plurality of feature points, other than the certain feature point, simultaneously detected by the feature point detecting unit (35) is modified based on a ratio of the first distance and the second distance.