Abstract: The present disclosure provides an interactive image matting method, a computer readable memory medium, and a computer device. The interactive image matting method includes steps: obtaining an original image; collecting foreground sample points on a hair edge foreground region of the original image and collecting background sample points on a hair edge background region of the original image by a human-computer interaction method to correspondingly obtain a foreground sample space and a background sample space; receiving a marking operation instruction input by a user, and smearing a hair region of the original image according to the marking operation instruction to mark unknown regions; traversing the unknown regions to obtain a pixel of each unknown region, traversing all the sample pairs to select a sample pair with a minimum overall cost function value for the pixel of each unknown region.

Abstract: Methods, an apparatus, and software media are provided for removing unwanted information such as moving or temporary foreground objects from a video sequence. The method performs, for each pixel, a statistical analysis to create a background data model whose color values can be used to detect and remove the unwanted information. The method assumes that for each pixel the background is present in a majority of the frames. The camera that records the video sequence may move relative to the geometry of the video scene. A pixel in a first frame is matched to a location in the geometry. The method determines color values of pixels, matched to the location in the geometry, in successive frames and clusters color values to determine a background color value range. It may use quadratic or better interpolation and extrapolation to determine background color values for unavailable frames.

Abstract: The present disclosure provides geological linear body extraction method based on tensor voting coupled with Hough transformation, including pre-processing a remote sensing image to obtain a pre-processed remote sensing image; selecting three optimal wavebands from N multi-spectral wavebands of the pre-processed remote sensing image, so as to obtain a remote sensing image combined by the optimal wavebands, N being a natural number greater than or equal to 3; using Gaussian high-pass filtering to perform sharpening processing on the remote sensing image combined by the optimal wavebands, so as to enhance linearized edge information; performing edge detection on the remote sensing image having enhanced linearized edge information, so as to obtain all edge points in the remote sensing image; and converting all the edge points in the remote sensing image from an image coordinate system to a parameter coordinate system, and extracting a geological linear body from the parameter coordinate system.

Abstract: A system and method for monitoring a roadway by comparing captured images to image inconsistencies caused by lost or unnecessary motion and for minimizing lost or unnecessary motion of an axle of a vehicle and filtering erroneous pulses of an electrical signal to generate a trigger signal to capture images is presented. The system enables the minimization of image artifacts causing distortion in an image. The system can take images of a target, compute an object pixel size of the image to ensure calibration of longitudinal measurements, and calibrate the rotary encoder with the camera of the machine vision system. The system can enable calibration of the locomotive components in the field to compensate for the misalignment of the machine vision system and provide safe travels.

Abstract: An image processing apparatus includes a setting unit, a small region setting unit, a determination unit, a search unit, a noise reduction unit, and a synthesis unit. The setting unit sets a target patch in an input image. The small region setting unit sets a small region near the target patch. The determination unit determines a number of similar patches to be detected as a patch similar to the target patch based on a noise amount in the small region. The search unit searches the input image for the similar patch based on the number of similar patches. The noise reduction unit configured to perform noise reduction processing with respect to the target patch using the target patch and the similar patch. The synthesis unit generates an image by synthesizing a plurality of patch groups subjected to the noise reduction processing.

Abstract: The present technology relates to an image processing apparatus and method that allow to more accurately identify image processing target areas in an image. An image processing apparatus of the present technology includes a determination unit that determines, on the basis of reliability information indicating reliability of depth information, whether to make a determination on the basis of the depth information, the determination being as to whether an area is a target area for predetermined image processing, and the depth information indicating a depth of an image. The present technology can be applied to, for example, an imaging apparatus that captures a subject, a server that processes a captured image, etc.

Abstract: Methods and systems for assessing a unit load device (ULD) package density are disclosed herein. An example method includes capturing a primary and secondary image featuring the ULD. The method further includes segmenting the primary image and the secondary image into a plurality of subsections and determining a primary largest depth value and secondary largest depth value for each subsection. The method further includes determining a depth change by comparing the primary largest depth value for each subsection to the secondary largest depth value for each subsection. The method further includes clustering subsections in the secondary image including a respective secondary largest depth value that is substantially similar to a subsection corresponding to a depth change to create a plurality of clusters. The method further includes calculating a wasted volume, and calculating a ULD package density by subtracting the wasted volume from a filled volume of the ULD.

Abstract: Introduced here are techniques for relighting an image by automatically segmenting a human object in an image. The segmented image is input to an encoder that transforms it into a feature space. The feature space is concatenated with coefficients of a target illumination for the image and input to an albedo decoder and a light transport detector to predict an albedo map and a light transport matrix, respectively. In addition, the output of the encoder is concatenated with outputs of residual parts of each decoder and fed to a light coefficients block, which predicts coefficients of the illumination for the image. The light transport matrix and predicted illumination coefficients are multiplied to obtain a shading map that can sharpen details of the image. Scaling the resulting image by the albedo map to produce the relight image. The relight image can be refined to denoise the relight image.

Abstract: In image segmentation it is usually assumed that all the surfaces of a scene are reflecting light in a perfectly diffuse or Lambertian manner. However, most existing surfaces are non-Lambertian surfaces. Examples of non-Lambertian surfaces are specular, glossy or mirror-like surfaces whose appearance may change depending on the point of view of the scene. The method according to an embodiment of the invention does not make the assumption that surfaces of objects in the view are perfect Lambertian reflectors by providing a surface color segmentation which takes into account the change in appearance due to the nature of the surfaces.

Abstract: The application discloses an automatic screen state detection robot, comprising a memory having an automatic screen state detection program stored thereon and a processor, and the automatic screen state detection program being executed by the processor to implement the operations of: controlling the robot to move into a preset area of each of service devices in a self-service branch respectively; detecting whether a service device has a circuit fault or not if the robot moves into the preset area of the service device; and controlling the service device to display an image according to preset display parameters if the display screen has no circuit fault, and analyzing the image displayed on the display screen to find whether the display screen of the service device has an abnormality of a preset type. The application also provides an automatic screen state detection method and a computer-readable storage medium.

Abstract: Disclosed herein are systems and methods for using focal stacks for image-based spoof detection. One embodiment takes the form of a method that includes obtaining a plurality of images of a biometric-authentication subject, the plurality of images captured by a camera system using a plurality of different focal distances; inputting, into an input-data-analysis module, a set of input data that includes the plurality of images; processing the set of input data using the input-data-analysis module to obtain, from the input-data-analysis module, a spoof-detection result for the biometric-authentication subject; and outputting the spoof-detection result for the biometric-authentication subject.

Abstract: A method for calibrating the position and/or orientation of a camera, in particular a camera mounted to a vehicle such as a truck, relative to a calibration pattern includes the steps of: A] acquiring an image of the calibration pattern by means of the camera; B] determining at least one parameter of the image and/or of the calibration pattern or a sub-pattern of the calibration pattern as it appears in the image; C] transforming the image based on the at least one parameter; D] identifying characteristic points or possible characteristic points of the calibration pattern within the transformed image of the calibration pattern; E] deriving the position and/or orientation of the camera relative to the calibration pattern from the identified characteristic points or possible characteristic points; F] in dependence of a confidence value of the derived position and/or orientation of the camera and/or in dependence of the number of iterations of steps B to F so far, repeating steps B to F, wherein in step B the de

Abstract: Systems and methods for spatial-aware information extraction from electronic source documents are disclosed. Exemplary implementations may: obtain an electronic source document in electronic format, including human-readable information; obtain extracted information that has been extracted from the electronic source document; generate a character-based representation of the electronic source document that uses a grid of character positions; and present a user interface to a user to search and/or select information in the electronic source document based on the character-based representation.

Abstract: A point cloud intensity completion method and system based on semantic segmentation are provided. The point cloud intensity completion method includes: acquiring an RGB image and point cloud data of a road surface synchronously by a photographic camera and a lidar; performing spatial transformation on the point cloud data by using a conversion matrix to generate a two-dimensional reflection intensity projection map and a two-dimensional depth projection map; performing reflection intensity completion on the RGB image and the two-dimensional reflection intensity projection map to obtain a single-channel reflection intensity projection map; performing depth completion on the RGB image and the two-dimensional depth projection map to obtain a single-channel depth projection map; and performing coarse-grained completion on the RGB image, the single-channel reflection intensity projection map and the single-channel depth projection map to obtain a two-dimensional coarse-grained reflectance intensity projection map.

Abstract: An electronic device and method are disclosed. The electronic device includes a memory and processor, which implements the method, including: obtaining image data for an image, set in the image a region of interest and a background region, using depth information, both including one or more sub-regions, determining respective bokeh characteristics for the one or more first sub-regions and the one or more second sub-regions based on a first characteristic criterion and a second characteristic criterion, respectively, and processing the image to apply a bokeh effect to the plurality of pixels based on the respective bokeh characteristics for the one or more first sub-regions and the one or more second sub-regions.

Abstract: A smartphone may be freely moved in three dimensions as it captures a stream of images of an object. Multiple image frames may be captured in different orientations and distances from the object and combined into a composite image representing an image of the object. The image frames may be formed into the composite image based on representing features of each image frame as a set of points in a three dimensional point cloud. Inconsistencies between the image frames may be adjusted when projecting respective points in the point cloud into the composite image. Quality of the image frames may be improved by processing the image frames to correct errors. Further, operating conditions may be selected, automatically or based on instructions provided to a user, to reduce motion blur. Techniques, including relocalization such that, allow for user-selected regions of the composite image to be changed.

Abstract: An adhered substance detection apparatus determines whether a substance is adhered to an image-capturing apparatus. The adhered substance detection apparatus includes a controller configured to function as: an extractor that extracts candidate regions for being an adhered substance region in which the substance is adhered to the image-capturing apparatus, the candidate regions being extracted based on an edge that is detected from pixels in an image captured by the image-capturing apparatus; and a final determiner that finally determines whether the candidate regions are the adhered substance region, based on i) an area of the candidate region located in a first region of the captured image, the first region including a road surface region in the captured image and ii) an area of the candidate region located in a second region of the captured image, the second region including a region other than the road surface region in the captured image.

Abstract: A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform: obtaining an image of an item; removing background pixels from the image by removing white pixels from the image up to a first threshold; determining an item outline of the item in the image, wherein the item outline comprises aliased pixels along a periphery of the item in the image; removing grey pixels from the item outline in the image up to a second threshold to create a first updated image; removing shadows from the first updated image to create a second updated image based on a saliency map and a third threshold for shadow-like grey pixels; mapping each pixel in the second updated image to a respective mapped color in a predetermined color palette; and determining one or more dominant colors of the respective mapped colors based on one or more highest respective percentages of the respective mapped colors.

Abstract: A method and system to generate training data for a deep learning model in memory instead of loading pre-generated data from disk storage. A corpus may be stored as lines of text. The lines of text can be manipulated in the memory of a central processing unit (CPU) of a computing system, using asynchronous multi-processing, in parallel with a training process being conducted on the system's graphics processing unit (GPU). With such an approach, for a given line of text, it is possible to take advantage of different fonts and different types of image augmentation without having to put the images in disk storage for subsequent retrieval. Consequently, the same line of text can be used to generate different training images for use in different epochs, providing more variability in training data (no training sample is trained on more than once). A single training corpus may yield many different training data sets.

Abstract: Disclosed herein are systems and methods for using machine learning for image-based spoof detection. One embodiment takes the form of a method that includes obtaining an input-data set that includes a plurality of images captured of a biometric-authentication subject by a plurality of cameras of a camera system. The method also includes inputting the input-data set into a trained machine-learning module, and processing the input-data set using the machine-learning module to obtain, from the machine-learning module, a spoof-detection result for the biometric-authentication subject. The method also includes outputting the spoof-detection result for the biometric-authentication subject.