Abstract: Provided is an image processing device including a global motion detection unit configured to detect a global motion indicating a motion of an entire image, a local motion detection unit configured to detect a local motion indicating a motion of each of areas of an image, and a main subject determination unit configured to determine a main subject based on the global motion and the local motion.

Abstract: An autonomous vehicle includes a travel vehicle main body, a model data storage, a photographic device, a search region determiner, an image feature point detector, a feature amount calculator and a position detector. The travel vehicle main body autonomously travels to a target position. The model data storage stores model data related to a geometric feature of an object. The photographic device photographs a periphery of the travel vehicle main body at the target position to acquire image data. The search region determiner predicts a position of the object based on the image data, and determines a search region of a predetermined range including the predicted position of the object. The image feature point detector detects a feature point of the image data with respect to the search region. The feature amount calculator calculates a feature amount of a matching candidate point extracted from the feature point.

Abstract: A system for analyzing scene traits in an object recognition ingestion ecosystem is presented. In some embodiment, a trait analysis engine analyzes a digital representation of a scene to derive one or more features. The features are compiled into sets of similar features with respect to a feature space. The engine attempts to discover which traits of the scene (e.g., temperature, lighting, gravity, etc.) can be used to distinguish the features for purposes of object recognition. When such distinguishing traits are found, an object recognition database is populated with object information, possibly indexed according to the similar features and their corresponding distinguishing traits.

Abstract: A method of detecting objects in three-dimensional (3D) point clouds and detecting differences between 3D point clouds and the objects therein is disclosed. A method includes receiving a first scene 3D point cloud and a second scene 3D point cloud, wherein the first scene 3D point cloud and the second scene 3D point cloud include first and second target objects, respectively; aligning the first scene 3D point cloud and the second scene 3D point cloud; detecting the first and second target objects from the first scene 3D point cloud and the second scene 3D point cloud, respectively; comparing the detected first target object with the detected second target object; and identifying, based on the comparison, one or more differences between the detected first target object and the detected second target object. Further aspects relate to detecting changes of target objects within scenes of multiple 3D point clouds.

Abstract: A method and apparatus for processing digital images are provided. The method includes: recognizing faces from a plurality of images of a jumping subject; determining respective priorities for the plurality of images of the jumping subject, wherein an index of the priorities is based on face recognition information; and aligning the plurality of images of the jumping subject based on the priorities.

Abstract: A system and a method for detecting and documenting the quality of straw, having the following steps: recording an image of a straw swath with a camera, generating a signal regarding the quality of the straw by means of an image-processing system based on parameters derived from the image of the camera, and storing and/or displaying the signal of the image processing system and/or controlling an actuator based on the signal of the image processing system.

Abstract: An image processing method includes steps of: providing a source face image and a target face image; extracting facial features from the source face image and the target face image respectively; detecting feature dimensions of the facial features from the source face image and the target face image respectively; pairing the facial features from the source face image with the facial features from the target face image; and, forming an output face image by adjusting the facial features from the source face image in at least one of the feature dimensions according to the paired features from the target face image in the corresponding feature dimensions.

Abstract: The invention discloses a direction-adaptive image deblurring method, comprising steps of: (1) defining a minimum cost function for deblurring an image by direction-adaptive total variation regularization; (2) converting the unconstrained minimization problem in step (1) to a constrained problem by auxiliary variables d1=Hu, d2=?xu and d3=?yu; (3) obtaining a new minimum cost function from the constrained problem in step (2) by introducing penalty terms; and (4) converting the minimization problem in step (3) to an alternating minimization problem about u, d1, d2 and d3, where a minimum of a variable is calculated as other variables are determined, and obtaining a deblurred image by solving the alternating minimization problem by an alternative and iterative minimization process.

Abstract: The present invention relates to a method of generating implicit terrain data and an electronic apparatus for performing the method, and more particularly to a method of generating implicit terrain data and an electronic apparatus which are capable of minimizing a size of data while maintaining geographical quality of original terrain data.

Abstract: In a lane line recognition apparatus, an abnormality determiner is configured to determine whether or not a width of a travel lane defined by left-side and right-side lane lines is abnormal, and a lane-line recognizer is configured to, when the width of the travel lane is not abnormal, recognize both of the left-side and right-side lane lines in a both-side line recognition mode, and when the width of the travel lane is abnormal, recognize one of the lane lines in a one-side line recognition mode. In the one-side line recognition mode, the lane-line recognizer is configured to, for each of the left-side and right-side lane lines, calculate two or more parameters of the lane line, and then integrate recognition results for the respective two or more parameters, and based on the integrated recognition results of the left-side and right-side lane lines, select one of the lane lines to be recognized.

Abstract: An automated method for the non-destructive testing of a woven preform for the manufacture of a turbine engine part and including a plurality of first marker threads intertwined with second marker threads, the first and second threads having light-reflecting properties that are different from those of the threads of the preform and being woven with the threads of the preform so as to form a surface grid on a given area of the preform. The method includes determining, with a plurality of consecutive steps, the spatial coordinates of the intersections between the first and second marker threads.

Abstract: Methods, apparatuses, systems, and software for extended phase correction in phase sensitive magnetic resonance imaging utilizing an optimized region-growing based phase correction algorithm. Phase correction is formulated as selecting a vector for each pixel of an image from two input candidate vectors so that the orientation of the output vector is spatially smooth. In certain embodiments, the optimized region growing algorithm uses automated quality guidance for determining the sequence of region growing and jointly considers the two input candidate vectors during region growing. Further, the algorithm tracks the quality and the mode at each step of the processing. Spatially isolated tissue regions are automatically segmented and processed with different threads of region growing and the correct vector is reliably identified as the output vector for each thread of region growing. Final phase correction was performed by pixel level optimization.

Abstract: An apparatus is provided for generating perceptually lossless image data or perceptually enhanced image data including: a color transform unit for color transforming a received image to a desired color space, forming converted color-space (CCS) images; a transform unit for receiving CCS images and transforming the CCS images into transformed CCS images each containing a set of transform coefficients; a quantization unit to quantize the transformed CCS images to form quantized CCS images; an image processor for receiving and processing transformed or quantized CCS images to produce optimized CCS images that are perceptually lossless or perceptually enhanced in quality; and an encoder for compressing the optimized CCS images; wherein the compressed optimized CCS images are subsequently prepared for storage and/or transmission.

Abstract: System, apparatus, method, and computer readable media for texture enhanced non-local means (NLM) image denoising. In embodiments, detail is preserved in filtered image data through a blending between the noisy input target pixel value and the NLM pixel value that is driven by self-similarity and further informed by an independent measure of local texture. In embodiments, the blending is driven by one or more blending weight or coefficient that is indicative of texture so that the level of detail preserved by the enhanced noise reduction filter scales with the amount of texture. Embodiments herein may thereby denoise regions of an image that lack significant texture (i.e. are smooth) more aggressively than more highly textured regions. In further embodiments, the blending coefficient is further determined based on similarity scores of candidate patches with the number of those scores considered being based on the texture score.

Abstract: The image reconstruction method includes performing a forward projection of a first image of an object using a line integral-based value and a point spread function (PSF)-based value to acquire a forward projection value, and performing a back projection of raw image data using the forward projection value to acquire a second image of the object.

Abstract: Discerning small, weak features (“specks”) from imagery can be critical in early-stage cancer detection and other applications. In one aspect, the presence of a speck is judged from a large set of “votes” about whether a point in the imagery has a value above or below its neighbors. Hundreds or thousands or more votes can be gathered from one or more images, to thereby—in the aggregate—tend to confirm or refute the presence of a speck at a particular location. Votes can be based on simple pixel differences, e.g., between a pixel at the center of a 7×7 pixel image excerpt, and each of 48 other pixels in the excerpt. More sophisticated methods can employ analyses of triads, quads, and rings, and kernel-based Markov Random Field frameworks, to roll-up to a final conclusion. A great number of other features and arrangements are also detailed. The technology is particularly illustrated in the context of detecting exoplanets from astronomical imagery of remote star systems.

Abstract: Methods, apparatuses, systems, and software for extended phase correction in phase sensitive Magnetic Resonance Imaging. A magnetic resonance image or images may be loaded into a memory. Two vector images A and B associated with the loaded image or images may be calculated either explicitly or implicitly so that a vector orientation by one of the two vector images at a pixel is substantially determined by a background or error phase at the pixel, and the vector orientation at the pixel by the other vector image is substantially different from that determined by the background or error phase at the pixel. A sequenced region growing phase correction algorithm may be applied to the vector images A and B to construct a new vector image V so that a vector orientation of V at each pixel is substantially determined by the background or error phase at the pixel.

Abstract: According to one embodiment, a handwritten character retrieval apparatus is provided with an acquisition unit, a feature extraction unit, an segmentation unit, a attribute append unit and a retrieval unit. The acquisition unit acquires a handwritten character string in units of a stroke. The feature extraction unit extracts a first feature value unique to each of the strokes from the handwritten character string. The segmentation unit segments the strokes into a plurality of sets. The attribute append unit appends a second feature value based on the sets to each of the strokes. The retrieval unit executes retrieval based on the first feature value and the second feature value.

Abstract: A blurred image having a spatially invariant motion blur resulting from camera motion during image capture is deblurred based on one or more light streaks identified and extracted from the blurred image. A blur kernel for the blurred image is estimated by performing an optimization procedure having a blur kernel constraint based at least in part on the light streak. One or more light streaks can in some embodiments be posed as the blur kernel constraint. A modeled light streak may be defined as a convolution between the blur kernel and a simulated light source, with the optimization procedure being to minimize a distance between the modeled light streak and the corresponding identified light streak from the blurred image.

Abstract: A method using optical coherence tomography to capture the microvascular network of the superficial layer of the finger skin for the purpose of fingerprint authentication and liveness detection. At the dermal papilla region, the vascular pattern follows the same pattern of the fingerprint and this vascular pattern forms a live vascular fingerprint. This live vascular fingerprint provides for ultrahigh security and a unique way for fingerprint-based personal verification. Because the system is based on blood flow, which only exists in a living person, the technique is robust against spoof attaching. After performing non-contact in-vivo imaging of a human fingertip, a three dimensional vasculature image is reconstructed from a plurality of vasculature tomography images and at least one vasculature fingerprint image which corresponds to the fingertip is extracted from the three dimensional vasculature image.