Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a second model to approximate the output of a first model to classify, according to a classification scheme, image data received as input, and after the second model is trained accessing map data that specifies a plurality of geographic locations, and for each geographic location associated with an entity for each image of the one or more images that depict the entity located at the geographic location, providing the image to the second model to generate an embedding for the image, associating each of the one or more embeddings generated by the second model with the geographic location, and storing, in a database, location data specifying the geographic location, the associated one or more embeddings, and data specifying the entity, as an associated entity entry for the entity.

Abstract: An acquiring unit includes a first acquiring unit, a second acquiring unit, and an extracting unit. The first acquiring unit acquires first data from contract data indicating a contract. The second acquiring unit acquires, from an information source other than contract data, second data different from the first data by using the first data acquired by the first acquiring unit. The extracting unit extracts features of the contract from first data acquired by the first acquiring unit and second data acquired by the second acquiring unit. The features include at least second data. A calculating unit calculates an evaluation value of the contract according to a user, by using the features of the contract acquired by the acquiring unit and a calculation criterion read from a calculation criterion DB.

Abstract: Systems and methods are described for colorizing a point cloud comprising a set of point cloud data points associated with a particular object using a plurality of candidate images. For each respective data point in the set of point cloud data points, an image is selected, from a subset of the plurality of candidate images that include an un-occluded point of the particular object corresponding to the respective data point, that has a shortest subject distance to the un-occluded point of the particular object. The respective data point is colorized based on a color in the selected image of the un-occluded point of the particular object corresponding to the respective data point.

Abstract: In order to provide a more preferable white balance adjustment, an image processing apparatus obtains an input image, determines a representative point in the input image and a white balance coefficient of the representative point, obtains first information representing the inclination of a surface that includes the representative point and second information representing the inclination of a surface that includes a pixel of interest in the input image, and adjusts white balance of the pixel of interest based on the first information, the second information, and the white balance coefficient.

Abstract: The invention relates to a method for background removal from documents. The method includes obtaining an image of a document, performing a clustering operation on the image to obtain a plurality of image segments, and performing, for each image segment, a foreground/background classification to determine whether the image segment includes foreground. The method further includes obtaining an augmented image by combining the image segments that include foreground, and obtaining a background-treated image by cropping the image of the document, based on the foreground in the augmented image.

Abstract: A method including normalizing an original image into an intermediate image, the intermediate image including multiple local blocks; calculating image feature data of the local blocks; calculating weight distribution data corresponding to the local blocks in the intermediate image according to the image feature data, the weight distribution data representing a degree of possibility that the local blocks include part or all of an object; and determining a location area of the object in the original image based on the weight distribution data obtained by calculation. By using the technical solutions in this present disclosure, an object in an image is localized rapidly and efficiently, and a subject area is determined, thereby saving a large amount of work for manually labeling images.

Abstract: In one example a management system for an autonomous vehicle, comprises a first image sensor to collect first image data in a first geographic region proximate the autonomous vehicle and a second image sensor to collect second image data in a second geographic region proximate the first geographic region and a controller communicatively coupled to the first image sensor and the second image sensor and comprising processing circuitry to collect the first image data from the first image sensor and second image data from the second image sensor, generate a first reliability index for the first image sensor and a second reliability index for the second image sensor, and determine a correlation between the first image data and the second image data. Other examples may be described.

Abstract: A total internal reflection based imaging system may include a light transmitting member having an imaging surface. A pressure sensitive membrane may be arranged on the imaging surface. The pressure sensitive membrane may include a top surface, a bottom surface, and an elastic deformable film forming at least a portion of the top surface. The imaging system may include a textured surface disposed between the pressure sensitive membrane and the imaging surface. An application of pressure on the top surface may deform the deformable film to reduce a distance between the deformable film and the imaging surface. A light source may be configured to emit a light towards the imaging surface such that reduced total internal reflection of the light occurs where the bottom surface of the pressure sensitive membrane contacts the imaging surface. A sensor may be configured to capture the light reflected from the imaging surface.

Abstract: Techniques are disclosed for generating logical sensors for an image driver. The image driver monitors values corresponding to at least a first feature in one or more regions of a first image in a stream of images received by a first sensor. The image driver identifies at least a first correlation between at least a first and second value of the monitored values. The image driver generates a logical sensor based on the identified correlations. The logical sensor samples one or more features corresponding to the identified correlation from a second image in the stream of images.

Abstract: Embodiments of this specification disclose a method and an apparatus for automotive part recognition based on a vehicle damage assessment image. The method includes: determining a damaged automotive part of a vehicle by recognizing a damage assessment image of the vehicle based on an image recognition algorithm; obtaining an automotive part list of the vehicle, wherein the automotive part list comprises a plurality of automotive parts corresponding to a plurality of automotive part identification serial numbers; matching the damaged automotive part with the plurality of automotive parts in the automotive part list to determine an automotive part corresponding to the damaged automotive part; and outputting an automotive part identification serial number of the determined automotive part.

Abstract: A system for processing a seller provided input image depicting an object. The system may process the input images to be in compliance with an online shopping website's image requirements. The image processing service may execute multiple modules on the input image. An edge detection module may detect one or more edges of the image of the object and determine border data indicative of a border around the object. An object of interest module may extract the image of the object within the border to generate an object of interest image. A background extraction module may create one or more closed shapes to separate the image of the object from background pixels. A hole detection module may traverse the one or more closed shapes to determine a presence of background pixels, remove the background pixels, and generate a resulting image.

Abstract: A non-transitory computer-readable medium stores a program. The program causes the computer to: acquire photographed image data which is generated by photographing, in an imaging apparatus, a photographed object indicating multiple reference points; acquire pieces of position information which indicate positions of the multiple reference points in the photographed image; map pieces of relative information to the respective reference points by analyzing relative positional relations between the multiple reference points in accordance with a first method; map the pieces of relative information to the respective reference points by analyzing relative positional relations between the multiple reference points in accordance with a second method; and determine a certain reference point as a reference point to be excluded, in a case that a first mapping relation is different from a second mapping relation.

Abstract: Technologies are presented herein in support of a system and method for performing iris identification and verification using mobile devices. Embodiments of the present invention concern a method for authenticating a user with a mobile device that includes steps for capturing imagery of a user's face, analyzing the imagery to identify the portion of the image depicting the user's eyes and segmenting the eye image to isolate the iris image data and encoding the iris image data in iris template. The present disclosure further describes authenticating the user based on the encoded iris template and additional techniques for preventing erroneous authentication caused by spoofing. In some examples, the anti-spoofing techniques may include capturing one or more images of a user's eyes and analyzing the captured images for indications of liveness based on reflection characteristics depicted in the imagery.

Abstract: Encoding and decoding occupancy information is disclosed. A method includes determining row sums for the region, determining column sums for the region, encoding, in a compressed bitstream, at least one of the row sums and the column sums, and encoding, in the compressed bitstream and based on a coding order, at least one of the rows and the columns of the region. The coding order is based on the encoded at least one of the row sums and the column sums. The row sums include, for each row of the region, a respective count of a number of locations in the row having a specified value. The column sums include, for each column of the region, a respective count of a number of locations in the column having the specified value. A location having the specified value is indicative of the occupancy information at the location.

Abstract: A contact image sensor can illuminate a portion of a banknote with one or more light emitting diodes (LEDs), capture an image of the portion of the banknote with illumination reflected from or transmitted through the banknote, and generate a digital output signal that represents the captured image. A controller board can receive the digital output signal, generate one or more LED control signals, each LED control signal including changes in voltage to indicate times to switch a light emitting diode on or off, direct the one or more LED control signals to the contact image sensor, bypass digital LED control circuitry that is included with the contact image sensor, and switch the one or more LEDs in the contact image sensor on or off at the times indicated by the respective light emitting diode control signals.

Abstract: An image segmentation method is disclosed. The method includes receiving a plurality of atlases and a subject image, each atlas including an atlas image showing a structure of interest and associated structure delineations, the subject image being acquired by an image acquisition device and showing the structure of interest. The method further includes calculating, by an image processor, mapped atlases by registering the respective atlases to the subject image, and determining, by the image processor, a first structure label map for the subject image based on the mapped atlases. The method also includes training, by the image processor, a structure classifier using a subset of the mapped atlases, and determining, by the image processor, a second structure label map for the subject image by applying the trained structure classifier to one or more subject image points in the subject image.

Abstract: Methods and apparatus, including computer program products, for creating a quality annotated training data set of images for training a quality estimating neural network. A set of images depicting a same object is received. The images in the set of images have varying image quality. A probe image whose quality is to be estimated is selected from the set of images. A gallery of images is selected from the set of images. The gallery of images does not include the probe image. The probe image is compared to each image in the gallery and a match score is generated for each image comparison. Based on the match scores, a quality value is determined for the probe image. The probe image and its associated quality value are added to a quality annotated training data set for the neural network.

Abstract: Provided are an image encoding method, an image decoding method and an image processing system including image encoding/decoding apparatus. The image encoding method includes steps of: acquiring a first image and a plurality of second images; updating features of each second image of the plurality of second images to obtain corresponding update features; superposing the first image with the update features of each second image of the plurality of second images to generate superposed images; generating a plurality of prediction images according to the superposed images; determining difference features between each second image of the plurality of second images and a corresponding prediction image; outputting the superposed images and the difference features; wherein the updating and/or predicting adopts a convolutional neural network.

Abstract: A method of encapsulating an encoded bitstream representing one or more images, the encapsulated bitstream comprising a data part and a metadata part. The method includes providing image item information identifying a portion of the data part representing a sub-image or an image of a single image and/or a set of single images; providing image description information comprising parameters including display parameters and/or transformation operators relating to one or more images and outputting said bitstream together with said provided information as an encapsulated data file. Said image item information comprises one or more properties including at least part of the image description information dedicated to the considered sub-image or single image or set of single images, said image description information being defined in one or more boxes.

Abstract: Image processing generates reference gain information based on a first tone characteristic by using an input image, the reference gain information including a gain value corresponding to a luminance value for each pixel, the first tone characteristic including a luminance range in which a gain signal monotonically increases; generates a correction signal to correct the reference gain information based on a second tone characteristic in which the gain signal monotonically decreases or monotonically increases regardless of the luminance range by using the input image and a low-frequency image generated from the input image; and performs gain processing on the input image based on gain information obtained after the reference gain information is corrected by the correction signal.