Abstract: There is provided a discriminative framework for image alignment. Image alignment is generally the process of moving and deforming a template to minimize the distance between the template and an image. There are essentially three elements to image alignment, namely template representation, distance metric, and optimization method. For template representation, given a face dataset with ground truth landmarks, a boosting-based classifier is trained that is able to learn the decision boundary between two classes—the warped images from ground truth landmarks (e.g., positive class) and those from perturbed landmarks (e.g., negative class). A set of trained weak classifiers based on Haar-like rectangular features determines a boosted appearance model. A distance metric is a score from the strong classifier, and image alignment is the process of optimizing (e.g., maximizing) the classification score.

Abstract: In order to detect a specific detection object from an input image, a color serving as a reference is calculated in a reference image region. The difference for each color component between each pixel in the detection window and the reference color is calculated. Whether or not the detection object is included in the detection window is discriminated by a feature vector indicating how the difference is distributed in the detection window.

Abstract: A method and system for evaluating probabilistic boosting trees is disclosed. In an embodiment, input data is received at a graphics processing unit. A weighted empirical distribution associated with each node of the probabilistic boosting tree is determined using a stack implementation. The weighted empirical distribution associated with each node is added to a total posterior distribution value.

Abstract: A display control apparatus may include a calculating unit for calculating the degree of similarity of each of a plurality of second images to a selected first image, a display control unit for controlling displaying of the second images in the order of similarity, and a selecting unit for selecting, in response to an instruction from a user, one second image from among the plurality of second images displayed in the order of similarity. In response to the selection of the one second image, the display control unit may control the displaying of the second image and a third image belonging to the same group as the selected second image so that the third image and the second image are displayed in accordance with a predetermined order in the group.

Abstract: Images are classified as photos (e.g., natural photographs) or graphics (e.g., cartoons, synthetically generated images), such that when searched (online) with a filter, an image database returns images corresponding to the filter criteria (e.g., either photos or graphics will be returned). A set of image statistics pertaining to various visual cues (e.g., color, texture, shape) are identified in classifying the images. These image statistics, combined with pre-tagged image metadata defining an image as either a graphic or a photo, may be used to train a boosting decision tree. The trained boosting decision tree may be used to classify additional images as graphics or photos based on image statistics determined for the additional images.

Abstract: A method for training a classifier for classifying candidate regions in computer aided diagnosis of digital medical images includes providing a training set of images, each image including one or more candidate regions that have been identified as suspicious by a computer aided diagnosis system. Each image has been manually annotated to identify malignant regions. Multiple instance learning is applied to train a classifier to classify suspicious regions in a new image as malignant or benign by identifying those candidate regions that overlap a same identified malignant region, grouping each candidate region that overlaps the same identified malignant region into a same bag, and maximizing a probability P = ? i = 1 N ? p i y i ? ( 1 - p i ) 1 - y i , wherein N is a number of bags, pi is a probability of bag i containing a candidate region that overlaps with an identified malignant region, and yi is a label where a value of 1 indicates malignancy and 0 otherwise.

Abstract: A watermark information embedding apparatus generates a document image from electronic document data that has been input thereto, modifies the electronic document data based upon the document image and embeds information in the electronic document data. The apparatus includes a document image generator for generating a document image from the electronic document data; a document analyzer for detecting layout information of each constituent image in the generated document image; a normalization information calculation unit for calculating normalization information, which is for normalizing placement of each constituent image, based upon the detected layout information; a modification unit for modifying the electronic document data; and an embedding unit for embedding information in the modified electronic document data.

Abstract: A system and method for verifying the face of a user using a light mask are provided. The system includes a facial feature extraction unit for extracting a facial feature vector from a facial image received from a camera. A non-user Gaussian Mixture Model (GMM) configuration unit generates a non-user GMM from a facial image stored in a non-user database (DB). A user GMM configuration unit generates a user GMM by applying light masks to a facial image stored in a user DB. A log-likelihood value calculation unit inputs the facial feature vector both to the non-user GMM and to the user GMM, thus calculating log-likelihood values. A user verification unit compares the calculated log-likelihood values with a predetermined threshold, thus verifying whether the received facial image is a facial image of the user.

Abstract: Described is a system for anomaly detection to detect an anomalous object in an image, such as a concealed object beneath a person's clothing. The system is configured to generate a subspace model for a normal class using training images. The normal class represents normal objects in a common class. The system receives a novel image having an object in the common class. A set of geometric landmarks are identified in the object in the novel image for use in registering the image. The novel image is registered by warping the image so that the geometric landmarks coincide in the novel image and the training images, resulting in a warped novel image having an object. Thereafter, the system determines if the object in the warped novel image is anomalous by measuring the distance of the warped novel image from the subspace model. Finally, if anomalous, an operator is notified accordingly.

Abstract: The present invention provides a technique for automated selection of a parameterized operator sequence to achieve a pattern classification task. A collection of labeled data patterns is input and statistical descriptions of the inputted labeled data patterns are then derived. Classifier performance for each of a plurality of candidate operator/parameter sequences is determined. The optimal classifier performance among the candidate classifier performances is then identified. Performance metric information, including, for example, the selected operator sequence/parameter combination, will be outputted. The operator sequences selected can be chosen from a default set of operators, or may be a user-defined set. The operator sequences may include any morphological operators, such as, erosion, dilation, closing, opening, close-open, and open-close.

Abstract: A computer readable medium is provided embodying instructions executable by a processor to perform a method for sparse volume segmentation for 3D scan of a target. The method including learning prior knowledge, providing volume data comprising the target, selecting a plurality of key contours of the image of the target, building a 3D spare model of the image of the target given the plurality of key contours, segmenting the image of the target given the 3D sparse model, and outputting a segmentation of the image of the target.

Abstract: Face-based image clustering systems and methods are described. In one aspect, face regions are detected in images. At least one respective parameter value is extracted from each of the face regions. Ones of the face regions associated with parameter values satisfying a cluster seed predicate are classified as cluster seed face regions. The cluster seed face regions are clustered into one or more clusters. A respective face model is built for each of the clusters. The face models are stored. In another aspect, face regions are detected in images. At least one respective parameter value is extracted from each of the face regions. The face regions are ranked based on the extracted parameter values. The face regions are clustered in rank order into one or more clusters. Representations of ones of the clusters are rendered on a display.

Abstract: A method and system for searching a database of graphical data are described. Embodiments of the invention use accelerated image-comparing techniques based on an adaptation of the Levenshtein algorithm for matching or searching one-dimensional data strings for use with recognizing pre-selected targets in graphical contents of 2D images.

Abstract: A method for recognition between a first and second object represented by at least one first image and at least one second image, includes defining rectangular assemblies of random pixels; filtering the first image with first n filters obtained from the assemblies of pixels to obtain n first filtered matrices; classifying the n first filtered matrices by providing a first center and a first radius within a space of N dimensions; filtering the second image with the first n filters to obtain n second filtered matrices; classifying the n second filtered matrices by providing a second center within the space of N dimensions; and comparing the first center and first radius with the second center.

Abstract: An image analyzer for detecting a target object from image data, includes a reference detection unit, a primary detection unit, a secondary detection unit and an output unit. The reference detection unit detects a reference object included in the image data. The primary detection unit detects candidates for the target object from the image data. The secondary detection unit specifies a portion including the target object from among the candidates, by using a correlation between a feature of the detected reference object and a feature of the candidates. The output unit outputs information representing the portion including the target object specified by the secondary detection unit.

Abstract: A learning device includes: a generating unit configured to generate an image having different resolution from an input image; an extracting unit configured to extract a feature point serving as a processing object from an image generated by the generating unit; a calculating unit configured to calculate the feature amount of the feature point by subjecting the feature point to filter processing employing a predetermined filter; and an identifier generating unit configured to generate an identifier for detecting a predetermined target object from the image by statistical learning employing the feature amount; with the filter including a plurality of regions, and the calculating unit taking the difference value of difference within the regions as the feature amount.

Abstract: An image searching apparatus which searches for an image that is similar to a query image from among a plurality of images is provided. The apparatus derives a color similarity that denotes a degree of similarity and a luminance similarity that denotes a degree of similarity between the query image and each of the images to be compared with. Then, the apparatus transforms the luminance similarity of each of the grayscale images to be compared with and the query image into an integrated similarity that is integrated in accordance with a correspondence between the color similarity and the luminance similarity of each of the color images to be compared with, and outputs the images to be compared in order by similarity, using the color similarity of the color images to be compared and the integrated similarity of the grayscale images to be compared.

Abstract: An efficient, effective and at times superior object detection and/or recognition (ODR) function may be built from a set of Bayesian stumps. Bayesian stumps may be constructed for each feature and object class, and the ODR function may be constructed from the subset of Bayesian stumps that minimize Bayesian error for a particular object class. That is, Bayesian error may be utilized as a feature selection measure for the ODR function. Furthermore, Bayesian stumps may be efficiently implemented as lookup tables with entries corresponding to unequal intervals of feature histograms. Interval widths and entry values may be determined so as to minimize Bayesian error, yielding Bayesian stumps that are optimal in this respect.

Abstract: A method of registering 3-dimensional digitized images to 2-dimensional digitized images during a medical procedure includes providing a pair of correctly-registered training images L={lr, lf} and their joint intensity distribution pl(ir, if), wherein ir and if are reference and floating images, respectively, providing a pair of observed images O={or, of} and their joint intensity distribution po(ir, if), mapping a marginal intensity distribution of the observed pair O={or, of} to a marginal intensity distribution of the training pair L={lr, lf}, and estimating a set of parameters T that registers image of to image or by maximizing a weighted sum of a Jensen-Shannon divergence (JSD) of a joint intensity distribution of the observed pair and a joint intensity distribution of the training pair and a similarity measure between the observed images.

Abstract: A method of computing global-to-local metrics for recognition. Based on training examples with feature representations, the method automatically computes a local metric that varies over the space of feature representations to optimize discrimination and the performance of recognition systems. Given a set of points in an arbitrary features space, local metrics are learned in a hierarchical manner that give low distances between points of same class and high distances between points of different classes. Rather than considering a global metric, a class-based metric or a point-based metric, the proposed invention applies successive clustering to the data and associates a metric to each one of the clusters.

Abstract: Described herein is a framework for automatically classifying a structure in digital image data are described herein. In one implementation, a first set of features is extracted from digital image data, and used to learn a discriminative model. The discriminative model may be associated with at least one conditional probability of a class label given an image data observation Based on the conditional probability, at least one likelihood measure of the structure co-occurring with another structure in the same sub-volume of the digital image data is determined. A second set of features may then be extracted from the likelihood measure.

Abstract: A method includes making a first determination as to whether a current pixel has a value which reflects a mosquito noise artifact, and determining whether to apply a filtering process at the current pixel based on a result of the first determination. In addition, or alternatively, a method includes making a second determination as to whether a current pixel has a value which reflects a ringing artifact, and determining whether to apply a filtering process at the current pixel based on a result of the second determination.

Abstract: Techniques are disclosed for a video surveillance system to learn to recognize complex behaviors by analyzing pixel data using alternating layers of clustering and sequencing. A video surveillance system may be configured to observe a scene (as depicted in a sequence of video frames) and, over time, develop hierarchies of concepts including classes of objects, actions and behaviors. That is, the video surveillance system may develop models at progressively more complex levels of abstraction used to identify what events and behaviors are common and which are unusual. When the models have matured, the video surveillance system issues alerts on unusual events.

Abstract: A video fingerprint insertion apparatus is provided. A frequency domain selection unit selects a frequency domain into which a fingerprint is inserted, from frequency domains of a video; a fingerprint insertion position determination unit determines a position where the fingerprint is to be inserted, based on a fingerprint insertion determination value in the selected frequency domain; and a fingerprint insertion unit inserts a fingerprint bit stream whose insertion strength is controlled, to the position where the fingerprint is to be inserted.

Abstract: Methods for constructing classifiers for binary classification of quantitative brain electrical activity data is described. The classifier building methods are based on the application of one or more evolutionary algorithms. In one embodiment, the evolutionary algorithm used is a genetic algorithm. In another embodiment, the evolutionary algorithm used is a modified Random Mutation Hill Climbing algorithm. In yet another embodiment, a combination of a genetic algorithm and a modified Random Mutation Hill Climbing algorithm is used for building a classifier. The classifier building methods are fully automated, and are adapted to generate classifiers (for example, Linear Discriminant Functions) with high sensitivity, specificity and classification accuracy.

Abstract: A method of producing an enhanced Active Appearance Model (AAM) by combining images of multiple resolutions is described herein. The method generally includes processing a plurality of images each having image landmarks and each image having an original resolution level. The images are down-sampled into multiple scales of reduced resolution levels. The AAM is trained for each image at each reduced resolution level, thereby creating a multi-resolution AAM. An enhancement technique is then used to refine the image landmarks for training the AAM at the original resolution level. The landmarks for training the AAM at each level of reduced resolution is obtained by scaling the landmarks used at the original resolution level by a ratio in accordance with the multiple scales.

Abstract: A multi-class sampling component (MCSC) is described for selecting samples associated with two or more sampling classes to produce output information. The overall set of samples in the output information exhibits a desirable Poisson distribution. Further, each subset of samples associated with each respective class exhibits a Poisson distribution. The MCSC selects samples based on intra-class radius information (describing the minimum allowed distances between same-class samples) and inter-class radius information (describing the minimum allowed distances between different-class samples). The MCSC can be applied to different applications, such as an object placement application, a color stippling application, a sensor design application, and so on.

Abstract: The present invention is directed to a method for detecting or predicting (302, 602) whether a test image is blurred. In one embodiment, the method includes extracting a training statistical signature (366) that is based on a plurality of data features (362, 364) from a training image set (14, 16), the training image set (14, 16) including a sharp image (14) and a blurry image (16); training a classifier (368) to discriminate between the sharp image (14) and the blurry image (16) based on the training statistical signature; and applying (302, 602) the trained classifier to a test image that is not included in the training image set (14, 16) to predict whether the test image is sharp (18) or blurry (20). The step of extracting can include measuring one or more statistical moments (576, 776) for various levels (L0-L5), estimating a covariance (577, 777) between adjacent levels (L0-L5), and/or extracting various metadata features (364, 664) from the images (14, 16).

Abstract: In a line by line image forming apparatus, line switching information which depends on scan line curve and an overlap data length to be read that overlaps across a plurality of lines before and after a switching position instructed by line switching information are set in a register. If switching to the line above or below is instructed by the line switching information, an address generating unit, when reading image data from an image memory, generates the read address and read data length of the image memory in accordance with the line switching information and the overlap data length, and reads image data corresponding to a current line and the line above or below the current line in accordance with the generated data.

Abstract: A method and apparatus are provided for adapting an exponential probability model. In a first stage, a general-purpose background model is built from background data by determining a set of model parameters for the probability model based on a set of background data. The background model parameters are then used to define a prior model for the parameters of an adapted probability model that is adapted and more specific to an adaptation data set of interest. The adaptation data set is generally of much smaller size than the background data set. A second set of model parameters are then determined for the adapted probability model based on the set of adaptation data and the prior model.

Abstract: A method adapts a boosted classifier to new samples. A boosted classifier is trained using initial samples. The boosted classifier is a combination of weak classifiers. Each weak classifier of the boosted classifier is updated adaptively by adding contributions of new samples and deleting contributions old samples.

Abstract: A method of managing large scale biometric data identity matching includes identifying a plurality of biometric modalities to be used in conducting identity matches, determining one of the biometric modalities to be a primary biometric modality and determining whether at least one matching algorithm is compatible with matrix entries. When the at least one matching algorithm is compatible with a matrix entry the method includes assigning a plurality of matching systems to the matrix entry. The matching systems are configured to conduct 1:1 or 1:few matching after conducting a 1:N identity matching. Furthermore, the method includes comparing a probe against enrollment data records stored in a corresponding matching system, and storing information regarding a matching enrollment data record in a queue when the probe matches an enrollment data record.

Abstract: A computer-implemented method for non-rigid multi-modal registration, includes providing trained data corresponding to previously registered images captured by different modalities, receiving two or more images captured by the different modalities, and registering the two or more images according to context information determined based on the trained data.

Abstract: A markerless motion capture system is provided for measurements accurate enough for biomechanical, clinical, sport, entertainment, animation, game and movie, design, ergonomics, surveillance applications. The system has multiple cameras distributed around a viewing volume. The cameras allow for the creation of three-dimensional mesh representations of an object dynamically moving within the viewing volume. A model of the object that incorporates specific morphological and kinematic model information (including soft joint constraints) is then matched to the captured three-dimensional mesh representations. The matching routine aims to embed the model into each of the three-dimensional representations using (i) iterative closest point or simulated annealing algorithms and (ii) using soft joint constraints. This unique combination of routines offers a simple, time-efficient, accurate and thus more meaningful assessment of movements.

Abstract: In a method of detection of different objects in an input image by application, to partial images cut at different positions in the input image, of a plurality of weak classifiers that evaluate whether a detection target image is an image of a predetermined object based on a histogram of values of characteristic quantities calculated from a plurality of sample images representing the predetermined object, the histogram is extended to multi-dimensions and a criterion for the evaluation by the weak classifiers is a multi-dimensional histogram representing histograms for the different objects in the form of vectors.

Abstract: A method of locating a contour of a structure in an image by processing said image including the structure is provided. A starting set of digital data representative of the image including the structure is taken, the structure in said image having annotated on it from three to ten landmark positions. A statistical model of said structure to the landmark positions annotated on the image is fitted and an initial estimate of the contour of the structure made. Using grey level information derived from points adjacent the estimated contour the contour boundary is iteratively moved to produce a final estimate of the contour of the structure.

Abstract: An object detecting device includes a comparing unit to extract feature amounts for two regions on a determining object image and compare a feature amount based on the two feature amounts extracted; and a computing unit to select one of two values having different absolute values according to the comparison result, and compute an evaluation value to determine whether or not an object is included in the determining object image, by performing computation with the selected value.

Abstract: This disclosure describes various exemplary method and computer program products for transductive multi-label classification in detecting video concepts for information retrieval. This disclosure describes utilizing a hidden Markov random field formulation to detect labels for concepts in a video content and modeling a multi-label interdependence between the labels by a pairwise Markov random field. The process groups the labels into several parts to speed up a labeling inference and calculates a conditional probability score for the labels, the conditional probability scores are ordered for ranking in a video retrieval evaluation.

Abstract: A face recognition apparatus and method using Principal Component Analysis (PCA) learning per subgroup, the face recognition apparatus includes: a learning unit which performs Principal Component Analysis (PCA) learning on each of a plurality of subgroups constituting a training data set, and then performs Linear Discriminant Analysis (LDA) learning on the training data set, thereby generating a PCA-based LDA (PCLDA) basis vector set of each subgroup; a feature vector extraction unit which projects a PCLDA basis vector set of each subgroup to an input image and extracts a feature vector set of the input image with respect to each subgroup; a feature vector storing unit which projects a PCLDA basis vector set of each subgroup to each of a plurality of face images to be registered, thereby generating a feature vector set of each registered image with respect to each subgroup, and storing the feature vector set in a database; and a similarity calculation unit which calculates a similarity between the input image

Abstract: Methods, systems, devices and/or storage media for passwords. An exemplary method tiles an image, associates an index with each tile and optionally determines offsets for select tiles. Further, the tiling optionally relies on probability and/or entropy. An exemplary password system includes an image; a grid associated with the image, the grid composed of polygons; an index associated with each polygon; and an offset associated with each polygon wherein password identification relies on one or more indices and one or more offsets.

Abstract: Embodiments provide for an image capture device having a learning function. Briefly described, one embodiment comprises at least one manual control adapted to generate a control signal that controls image capture operation, the control operable by a user of the image capture device; at least one sensor adapted to sense an operating condition during image capture; a first element adapted to analyse inputs from the sensor to learn at least one associated operating condition, and adapted to analyse the control signal to learn a corresponding preference of the user; a second element adapted to predict a value corresponding to the control signal and the associated operating condition; and a third element adapted to compare the predicted value and the control signal, and adapted to determine a confidence level from the compared predicted value and control signal, wherein the confidence level corresponds to a degree of confidence in the predicted value.

Abstract: A method for recognizing markers printed on a learning material, includes sampling an image of the learning material; grouping the sampled image of pixels into a first image group and a second image group based on a threshold; and calculating medians of the first image group and the second image group to update the threshold with a first average value of the calculated medians. Further, the method for recognizing markers printed on the learning material includes repeating the above until a difference between a previous threshold and an updated threshold is equal to or smaller than a reference value; binarizing an image captured by a camera based on the updated threshold; and detecting the markers based on the binary image.

Abstract: Described herein is a technology for facilitating coordinated description in image analysis. In one implementation, the technology includes receiving image data including at least first and second descriptors (204) describing portions of the image data. The first and second descriptors are coordinated by determining at least one conditional probability of observing the first descriptor in the image data given an occurrence of the second descriptor (206). A classifier may then be trained based on the conditional probability (208).

Abstract: A method for multiple image restoration includes receiving a plurality of images corrupted by noise, and initializing a reduced noise estimate of the plurality of images. The method further includes estimating a probability of distributions of noise around each pixel and the probability of the signal, estimating mutual information between noise on the plurality of images based on the probabilities of distributions of noise around each pixel and the joint distribution of noise, and updating each pixel within a search range to determine a restored image by reducing the mutual information between the noise on the plurality of images.

Abstract: The present invention provides a method and apparatus for detecting fire in a monitored area even if the flames are hidden behind obstructing objects.

Abstract: Methods of statistical learning for Bayesian inference in the context of efficient optimization schemes for image restoration are presented. Second and third order priors that may be learned while maintaining graph representability are identified. A framework to learn and impose prior knowledge on the distribution of pairs and triplets of labels via graph cuts is presented. The disclosed methods optimally restore binary textures from very noisy images with runtimes in the order of seconds while imposing hundreds of statistically learned constraints per node.

Abstract: The present invention is directed to a system and method for on-road vehicle detection. A video sequence is received that is comprised of a plurality of image frames. A potential vehicle appearance is identified in an image frame. Known vehicle appearance information and scene geometry information are used to formulate initial hypotheses about vehicle appearance. The potential vehicle appearance is tracked over multiple successive image frames. Potential motion trajectories for the potential vehicle appearance are identified over the multiple image frames. Knowledge fusion of appearance, scene geometry and motion information models are applied to each image frame containing the trajectories. A confidence score is calculated for each trajectory. A trajectory with a high confidence score is determined to represent a vehicle appearance.

Abstract: Disclosed is a distortion invariant system, method and computer readable medium for detecting the presence of one or more predefined targets in an input image. The input image and a synthetic discriminant function (SDF) reference image are correlated in a shift phase-encoded fringe-adjusted joint transform correlation (SPFJTC) correlator yielding a correlation output. A peak-to-clutter ratio (PCR) is determined for the correlation output and compared to a threshold value. A predefined target is present in the input image when the PCR is greater than or equal to the threshold value.

Abstract: A method of processing a mammogram image to derive a value for a parameter useful in detecting differences in breast tissue in subsequent images of the same breast or relative to a control group of such images, said derived parameter being an aggregate probability score reflecting the probability of the image being a member of a predefined class of mammogram images, comprises computing for each of a multitude of pixels within a large region of interest within the image a pixel probability score assigned by a trained statistical classifier according to the probability of said pixel belonging to an image belonging to said class, said pixel probability being calculated on the basis of a selected plurality of features of said pixels, and computing said parameter by aggregating the pixel probability scores over said region of interest. Saud features may include the 3-jet of said pixels.

Abstract: An image processing apparatus is provided that is capable of executing identification of an object by using a two-dimensional-image having a relatively low resolution. In the image processing apparatus, image data obtained by photo capturing a prescribed object is used as a processing target; with respect to at least one plane relating to a part of the object, an N-dimensional estimated feature value (N ?3) defining the plane is operated; the N-dimensional estimated feature value and information identifying the original object are associated with each other and stored as a recognition database in a storage unit; and the recognition database is applied to a recognition process of the object.