Abstract: A head-mounted display, a method, and a non-transitory computer readable medium are provided. An embodiment of a method for obtaining training sample views of an object includes the step of storing, in a memory, multiple views of an object. The method also includes the step of deriving similarity scores between adjacent views and then a sampling density is varied based on the similarity scores.

Abstract: A method includes acquiring, from a camera, an image data sequence of a real object in a real scene and performing a first template-matching on an image frame in the image data sequence using intensity-related data sets stored in one or more memories to generate response maps. The intensity-related data sets represent an intensity distribution of a reference object from respective viewpoints. The reference object corresponds to the real object. A candidate region of interest is determined for the real object in the image frame based on the response maps, and second template-matching is performed on the candidate region of interest using shape-related feature data sets stored in one or more memories to derive a pose of the real object. The shape-related feature data sets represent edge information of the reference object from the respective viewpoints.

Abstract: An object attitude detection device includes a pick-up image acquisition unit, a template image acquisition unit, and an attitude decision unit. The pick-up image acquisition unit acquires a picked-up image of an object. The template image acquisition unit acquires a template image for each attitude of the object. The attitude decision unit decides an attitude of the object based on the template image having pixels. In the pixels, a distance between pixels forming a contour in the picked-up image and pixels forming a contour of the template image is shorter than a first threshold. Further, a degree of similarity between a gradient of the pixels forming the contour in the picked-up image and a gradient of the pixels forming the contour of the template image is higher than a second threshold.

Abstract: A method for one or more processors to implement includes acquiring a synthetic image of an object from a second orientation different from a first orientation, using a three-dimensional model. The method further includes identifying, in the synthetic image, second edge points that are located on an edge of the object that is not a perimeter. The method further includes identifying matched edge points, which are first edge points and second edge points at substantially a same location on the object. The method further includes storing the matched edged points in a memory that can be accessed by an object-tracking device, so that the object-tracking device can identify the object in a real environment by identifying the matched edge points in images of the object.

Abstract: A method for one or more processors to implement includes acquiring a synthetic image of an object from a second orientation different from a first orientation, using a three-dimensional model. The method further includes identifying, in the synthetic image, second edge points that are located on an edge of the object that is not a perimeter. The method further includes identifying matched edge points, which are first edge points and second edge points at substantially a same location on the object. The method further includes storing the matched edged points in a memory that can be accessed by an object-tracking device, so that the object-tracking device can identify the object in a real environment by identifying the matched edge points in images of the object.

Abstract: A head-mounted display, a method, and a non-transitory computer readable medium are provided. An embodiment of a method for obtaining training sample views of an object includes the step of storing, in a memory, multiple views of an object. The method also includes the step of deriving similarity scores between adjacent views and then a sampling density is varied based on the similarity scores.

Abstract: A method includes acquiring, from a camera, an image data sequence of a real object in a real scene and performing a first template-matching on an image frame in the image data sequence using intensity-related data sets stored in one or more memories to generate response maps. The intensity-related data sets represent an intensity distribution of a reference object from respective viewpoints. The reference object corresponds to the real object. A candidate region of interest is determined for the real object in the image frame based on the response maps, and second template-matching is performed on the candidate region of interest using shape-related feature data sets stored in one or more memories to derive a pose of the real object. The shape-related feature data sets represent edge information of the reference object from the respective viewpoints.

Abstract: A head-mounted display includes a camera that obtains an image of an object within a field of view. The head-mounted display further includes a processor configured to determine a plurality of feature points from the image and calculate a feature strength for each of the plurality of feature points. The processor is further configured to divide the image into a plurality of cells and select feature points having the highest feature strength from each cell and which have not yet been selected. The processor being further configured to detect and track the object within the field of view using the selected feature points.

Abstract: An object attitude detection device includes: a picked-up image acquisition unit which acquires a picked-up image of an object; a template image acquisition unit which acquires a template image for each attitude of the object; and an attitude decision unit which decides an attitude of the object, based on the template image having pixels such that a distance between pixels forming a contour in the picked-up image and pixels forming a contour of the template image is shorter than a first threshold and that a degree of similarity between a gradient of the pixels forming the contour in the picked-up image and a gradient of the pixels forming the contour of the template image is higher than a second threshold.

Abstract: A head-mounted display includes a camera that obtains an image of an object within a field of view. The head-mounted display further includes a processor configured to determine a plurality of feature points from the image and calculate a feature strength for each of the plurality of feature points. The processor is further configured to divide the image into a plurality of cells and select feature points having the highest feature strength from each cell and which have not yet been selected. The processor being further configured to detect and track the object within the field of view using the selected feature points.

Abstract: An adequate solution for computer vision applications is arrived at more efficiently and, with more automation, enables users with limited or no special image processing and pattern recognition knowledge to create reliable vision systems for their applications. Computer rendering of CAD models is used to automate the dataset acquisition process and labeling process. In order to speed up the training data preparation while maintaining the data quality, a number of processed samples are generated from one or a few seed images.

Abstract: An adequate solution for computer vision applications is arrived at more efficiently and, with more automation, enables users with limited or no special image processing and pattern recognition knowledge to create reliable vision systems for their applications. Computer rendering of CAD models is used to automate the dataset acquisition process and labeling process. In order to speed up the training data preparation while maintaining the data quality, a number of processed samples are generated from one or a few seed images.

Abstract: Methods and systems for detecting and compensating for motion depicted in a sequence of frames are disclosed. One example method includes converting video frames to monochrome. Conversion of an image to monochrome includes sampling luminance data of a portion of pixels in the image to identify the image as being of a particular image type, selecting a bit plane of the image based on the identified image type, and converting the image to a monochrome image using the selected bit plane. After conversion of video frames to monochrome, a pixel subset in a monochrome image corresponding to a first frame is compared to candidate matching pixel subsets within a search area of a monochrome image corresponding to a second, consecutive frame. The frames are processed to compensate for perceived motion blur based on the detected inter-frame motion.

Abstract: An image of a scanned book is segmented using a feature image to map pixels corresponding to a page area and to create page objects and detect borders of the page. A book spine region is detected by locating a plain background area between two of the page objects, analyzing the page borders to detect their shape, and analyzing their shape to detect the book spine end points. Using the page borders, the feature image is examined to detect top-to-bottom and bottom-to-top declines in pixel values to determine the corners of a shadow distortion in the original scanned image. Squeeze and curvature distortion are also detected. A Bezier curve is used to model each of the three distortions detected on the page. The detected distortion is corrected by first defining a trapezoidal correction area. The intensity, squeeze, and curvature corrections are then applied along lines within the trapezoidal correction area.

Abstract: Due to an accumulated error from the pair-wise registration, the stitched image may be blurred or have a gap when a loop is encountered. In order to remove the accumulated error, we identify a closed loop where a first image frame overlaps with a second image frame, the second image frame being captured earlier in a scanning sequence than the first image frame; register the first image frame with the second image frame; and apply a global optimization to adjust registration parameters for the plurality of pair-wise registrations of image frames within the closed loop using global constraints.

Abstract: An image of a scanned book is segmented using a feature image to map pixels corresponding to a page area and to create page objects and detect borders of the page. A book spine region is detected by locating a plain background area between two of the page objects, analyzing the page borders to detect their shape, and analyzing their shape to detect the book spine end points. Using the page borders, the feature image is examined to detect top-to-bottom and bottom-to-top declines in pixel values to determine the corners of a shadow distortion in the original scanned image. Squeeze and curvature distortion are also detected. A Bezier curve is used to model each of the three distortions detected on the page. The detected distortion is corrected by first defining a trapezoidal correction area. The intensity, squeeze, and curvature corrections are then applied along lines within the trapezoidal correction area.

Abstract: Two images are stitched together through minimization of a cost function that consists of registration errors from image data of the two images, as well as the estimated errors from a set of sensors. The weight function in the cost function is derived from the confidence value of sensor estimation that considers the sensor errors including lift and off page as well as a measure of accuracy of the sensor readings. Weights are used to adjust image registration accuracy against sensor accuracy to produce a set of registration parameters that would best stitch the two images together. In order to handle large errors for initial registration parameters and to avoid local minima in the minimization process, the image pair may be registered in a lower resolution and then refined in a higher resolution.

Abstract: The present invention is a system and method for decoding an image of a bar code. Decoding the barcode includes tokenizing a plurality of pixels in the image of the barcode based upon a plurality of thresholds to form a first set of tokens. Decoding the barcode also includes re-tokenizing the plurality of pixels in the image of the barcode based upon the intensity of the pixels in the plurality of tokens, and the relativity intensity of neighboring tokens, to form a second set of tokens.

Abstract: An off page condition or invalid sensors position data is detected by checking the errors from an initial transformation parameter estimation. If an abnormally large error is encountered, a sensor's reading (position data) may be invalid or the sensor was off page. Then the invalid sensor data will be identified and removed. Finally the transformation parameters will be re-estimated using valid sensor position data only. A weighted least-square minimization is used by considering the sensor lift situation. If a sensor is lifted, the weight for the error related to the sensor will be set to a small weight or zero. Also considered are the geometric properties of sensor locations in weighting the sensor error. A confidence measurement of the sensor data and associated error is performed. The confidence measurement is derived from an error ellipse at 95% confidence level.

Abstract: Due to an accumulated error from the pair-wise registration, the stitched image may be blurred or have a gap when a loop is encountered. In order to remove the accumulated error, we identify a closed loop where a first image frame overlaps with a second image frame, the second image frame being captured earlier in a scanning sequence than the first image frame; register the first image frame with the second image frame; and apply a global optimization to adjust registration parameters for the plurality of pair-wise registrations of image frames within the closed loop using global constraints.