Abstract: Systems and methods are disclosed to recognize human action from one or more video frames by performing 3D convolutions to capture motion information encoded in multiple adjacent frames and extracting features from spatial and temporal dimensions therefrom; generating multiple channels of information from the video frames, combining information from all channels to obtain a feature representation for a 3D CNN model; and applying the 3D CNN model to recognize human actions.

Abstract: A method for associating a three-dimensional surface representing a real object and a three-dimensional reference surface, said reference surface being represented by a set of reference points, the method comprising: obtaining a set of real points representing the real surface, determining the normal vector of each point of said obtained set of real points, selecting, among the set of real points, control points according to the determined normal vector by converting the set of real points to a bi-dimensional space of normal vectors, generating sets of points having similar normal vector among the points of the set of real points and selecting, for each set of points with similar normal vector, one point that is a control point of the real surface, determining correspondence points close to the set of reference points that are determined to correspond to the control points of the real surface, and determining the motion that minimizes the distances between the control points of the real surface and the corres

Abstract: A method and system for image recognition in a collection of digital images includes training image classifiers and retrieving a sub-set of images from the collection. For each image in the collection, any regions within the image that correspond to a face are identified. For each face region and any associated peripheral region, feature vectors are determined for each of the image classifiers. The feature vectors are stored in association with data relating to the associated face region. At least one reference region including a face to be recognized is/are selected from an image. At least one classifier on which said retrieval is to be based is/are selected from the image classifiers. A respective feature vector for each selected classifier is determined for the reference region. The sub-set of images is retrieved from within the image collection in accordance with the distance between the feature vectors determined for the reference region and the feature vectors for face regions of the image collection.

Abstract: A method, a biometric identifier collection device, and a set of instructions are disclosed. A memory 208 may store a digital image having a biometric identifier. A processor 204 may execute a lighting compensation to remove a lighting effect from the biometric identifier. The processor 204 may process the biometric identifier to create an identification profile.

Abstract: The present invention provides a system for generating images of multi-views.

Abstract: An optical character recognition process characterizes text lines in a textual image by their base-line, mean-line and x-height. The base-line for at least one text line in the image is determined by finding a parametric curve that maximizes a first fitness function that depends on the values of pixels through which the parametric curve passes and pixels below the parametric curve. The base-line corresponds to the parametric curve for which the first fitness function is maximized. The first fitness function is designed so that it increases with increasing lightless or brightness of pixels immediately below the parametric curve while also increasing with decreasing lightness of pixels through which the parametric curve passes. The mean-line is determined by incrementally shifting the base-line upward by predetermined amounts (e.g., a single pixel) until a second fitness function for the shifted base-line is maximized. The second fitness function is essentially the inverse of the first fitness function.

Abstract: A vehicle travel support device determines presence of a recognition inhibiting factor of a lane mark on a road on which a vehicle is traveling with high accuracy irrespective of an imaging history by a vehicular camera from the same position. The vehicle travel support system generates an edge image by extracting an edge or actualizing an edge in an image obtained through the vehicular camera. When Hough transform of the edge image is performed, votes for a specified vote value of a linear component is evaluated in a ?-? space (Hough space). Presence of a recognition inhibiting factor of a lane mark on a road is determined by determining whether or not the votes of a specified vote value in a specified region denoting a standard travel lane of a vehicle in the real space is ?a threshold in the ?-? space.

Abstract: Processing for judging whether a face is included in a frame is performed, in a predetermined interval, on each of frames included in a moving image of a subject, displayed on a monitor, until the judgment becomes positive. If it is judged that a face is included in a frame, the facial position is detected in the frame, and stored. Then, judgment is made as to whether a face is included in the next frame after predetermined time. If the judgment is positive, the facial position is detected. The previously stored facial position is replaced by the newly detected facial position, and the newly detected facial position is stored. These processes are repeated until photographing operation is performed by operating a release unit.

Abstract: A transaction processing apparatus includes a display unit, a base unit, and an attaching and detaching unit that connects circuits of the display unit and the base unit. The display unit includes an authentication-information acquiring unit that acquires, in a state in which the display unit is detached from the base unit, identity authentication information of a customer and an authentication-information storing unit that stores the acquired identity authentication information.

Abstract: A facial image recognition system for a driver of a vehicle includes an image capturing unit, an image processing unit, and a warning unit. The image capturing unit is used for capturing facial images of the driver. The image processing unit is electrically coupled to the image capturing unit, has installed therein a facial frame selecting and position correcting method and an identification comparison algorithm, and receives the facial images of the driver from the image capturing unit. The warning unit is electrically coupled to the image processing unit, and emits a warning signal when the image processing unit determines that an identification of the driver has changed.

Abstract: An image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. By employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. The pattern recognizer can be a neural network including a plurality of stages of observers. The observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. Each of the observers includes a plurality of neurons. The input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units.

Abstract: While locating a license plate of a moving vehicle on consecutive images, motion detection is first performed on the consecutive images to detect a moving vehicle image, which is segmented using edge detection, and the segmented moving vehicle image is analyzed to retrieve characteristics for locating a license plate image and determining characters on the located license plate image. As a result, a precise location of the license plate is thus precisely located for further recognition no matter what weathers in which the consecutive images are recorded. The above-mentioned technique requires merely few calculations, is easily implemented, and may be applied on an intelligent digital video recording (DVR) system including many computer-vision functions.

Abstract: At least one approximation image is created of the image at one or multiple scales. Translation difference images are created by pixel-wise subtracting the values of an approximation image at scale s and the values of a translated version of the approximation image. A non-linear modification is applied to the values of the translation difference image (s) and at least one enhanced center difference image at a specific scale is computed by combining the modified translation difference images at that scale or a smaller scale with weights wi,,j. An enhanced image is computed by applying a reconstruction algorithm to the enhanced center difference images. The non-linear modification of the values of the translation difference images is steered by (a) characteristic (s) computed out of the approximation image (s) at least one scale.

Abstract: A system for one dimensional segmentation of an iris of an eye into a map of the iris and classification of the map into unaffected areas and affected areas. Also, the system may provide for regular shape fitting of the areas for normalization and identifying the unaffected areas as symmetric segments. Further, the system may assign weights to the unaffected areas and the affected areas of the map of the iris and an enrolled map of an iris and their corresponding bins for matching purposes.

Abstract: A system, circuit and methods for target detection from hyper-spectral image data are disclosed. Filter coefficients are determined using a modified constrained energy minimization (CEM) method. The modified CEM method can operate on a circuit operable to perform constrained linear programming optimization. A filter comprising the filter coefficients is applied to a plurality of pixels of the hyper-spectral image data to form CEM values for the pixels, and one or more target pixels are identified from the CEM values. The process may be repeated to enhance target recognition by using filter coefficients determined by excluding the identified target pixels from the hyper-spectral image data.

Abstract: Display suitable to an actual three-dimensional model or a recognition-target object is performed when stereoscopic display of a three-dimensional model is performed while correlated to an image used in three-dimensional recognition processing. After a position and a rotation angle of a workpiece are recognized through recognition processing using the three-dimensional model, coordinate transformation of the three-dimensional model is performed based on the recognition result, and a post-coordinate-transformation Z-coordinate is corrected according to an angle (elevation angle f) formed between a direction of a line of sight and an imaging surface. Then perspective transformation of the post-correction three-dimensional model into a coordinate system of a camera of a processing object is performed, and a height according to a pre-correction Z-coordinate at a corresponding point of the pre-coordinate-transformation three-dimensional model is set to each point of a produced projection image.

Abstract: A system for iris recognition using a set of quality metrics, which may include eye image validation, blur assessment, offset, gazing, obscuration, visibility, and the like. These metrics may be established as quantitative measures which can automatically assess the quality of eye images before they are processed for recognition purposes. Quadrant iris analysis, histograms, map processing enhancements, and multi-band analysis may be used in aiding in the iris recognition approach.

Abstract: A method for determining feature point locations in an image performs a first search in a predetermined first search area to search for locations of plural feature points in the image, corrects the locations of the plural feature points based on a geometric layout relationship among the plural feature points searched for, sets a second search area based on the corrected location of each of the feature points, and performs a second search in the second search area to search for the location of each of the feature points. Then, the method determines reliability of the location of each feature point searched for by the second search and selects one of the corrected location and the location searched for by the second search, as a location of the feature point.

Abstract: An image encoding device that receives a plurality of component values of a color input signal inputted to an object pixel, vector quantizes, and encodes each of the plurality of component values includes: a difference vector quantization unit that vector quantizes and encodes each difference value obtained by subtracting prediction component values based on a previous object pixel from the component values of the object pixel respectively; an absolute value vector quantization unit that vector quantizes and encodes each different value of the object pixel; a determination unit that compares each difference value with respective comparison ranges to determine a code selection; and a code selector that selects one of an output from the difference vector quantization unit and an output from the absolute value vector quantization unit based on an output from the determination unit.

Abstract: Hands may be tracked before, during, and after occlusion, and a gesture may be recognized. Movement of two occluded hands may be tracked as a unit during an occlusion period. A type of synchronization characterizing the two occluded hands during the occlusion period may be determined based on the tracked movement of the occluded hands. Based on the determined type of synchronization, it may be determined whether directions of travel for each of the two occluded hands change during the occlusion period. Implementations may determine that a first hand and a second hand are occluded during an occlusion period, the first hand having come from a first direction and the second hand having come from a second direction. The first hand may be distinguished from the second hand after the occlusion period based on a determined type of synchronization characterizing the two hands, and a behavior of the two hands.