Abstract: Systems and methods are provided for the denoising of images in the presence of broadband noise based on the detection and/or estimation of in-band noise. According to various example embodiments, an estimate of broadband noise that lies within the imaging band is made by detecting or characterizing the out-of-band noise that lies outside of the imaging band. This estimated in-band noise may be employed for denoise the detected imaging waveform. According to other example embodiments, a reference receive circuit that is sensitive to noise within the imaging band, but is isolated from the imaging energy, may be employed to detect and/or characterize the noise within the imaging band. The estimated reference noise may be employed to denoise the detected in-band imaging waveform.

Abstract: A computer-implemented method of training a convolutional neural network configured to morph content features of an input image with style features of a style image is provided. The computer-implemented method includes selecting a training style image; extracting style features of the training style image; selecting a training content image; extracting content features of the training content image; processing the training content image through the convolutional neural network to generate a training output image including the content features of the training content image morphed with the style features of the training style image; extracting content features and style features of the training output image; computing a total loss; and tuning the convolutional neural network based on the total loss including a content loss, a style loss, and a regularization loss.

Abstract: A method for analyzing a rock sample includes segmenting a digital image volume corresponding to an image of the rock sample, to associate voxels in the digital image volume with a plurality of rock fabrics of the rock sample. The method also includes identifying a set of digital planes through the digital image volume. The set of digital planes intersects with each of the plurality of rock fabrics. The method further includes machining the rock sample to expose physical faces that correspond to the identified digital planes, performing scanning electron microscope (SEM) imaging of the physical faces to generate two-dimensional (2D) SEM images of the physical faces, and performing image processing on the SEM images to determine a material property associated with each of the rock fabrics.

Abstract: An image-recognition method is provided. The method includes the following steps: receiving structured data, wherein the structured data includes training-set data and testing-set data, and the structured data includes a plurality of groups, and each group includes one or more types, and each type includes a plurality of check-point images; training an artificial-intelligence (AI) model using the training-set data; inputting the testing-set data into the AI model to obtain a model evaluation of the AI model; and determining one or more first types with a lower overall recognition rate or a lower confidence level in the structured data, and deletes or corrects the check-point images in the one or more first types to update the structured data.

Abstract: Provided are a damage diagram creation method, a damage diagram creation device, a damage diagram creation system, and a recording medium capable of detecting damage with high accuracy based on a plurality of images acquired by subjecting a subject to split imaging. In a damage diagram creation method, damage of a subject is detected from each image (each image in a state of being not composed) constituting a plurality of images (a plurality of images acquired by subjecting the subject to split imaging), and thus, damage detection performance is not deteriorated due to deterioration of image quality in an overlapping area. Therefore, it is possible to detect damage with high accuracy based on a plurality of images acquired by subjecting the subject to split imaging. Detection results for the respective images can be composed using a composition parameter calculated based on correspondence points between the images.

Abstract: Methods, systems, and apparatus including computer programs encoded on a computer storage medium, for generating convex decomposition of objects using neural network models. One of the methods includes receiving an input that depicts an object. The input is processed using a neural network to generate an output that defines a convex representation of the object. The output includes, for each of a plurality of convex elements, respective parameters that define a position of the convex element in the convex representation of the object.

Abstract: The present disclosure provides a method for processing an event data flow and a computing device. The method includes: reading a plurality of pieces of event data with a first duration sequentially from the event data flow; with respect to each piece of event data with the first duration, analyzing the event data to acquire time-difference information about each event within the first duration; and generating an image frame presenting a change in movement within the first duration in accordance with the time-difference information about each event within the first duration.

Abstract: Provided are an under-screen biometric identification apparatus and an electronic device. The under-screen biometric identification apparatus includes: a lens disposed under a display screen for receiving an optical signal formed by reflection of a human finger on the display screen, where the optical signal is used to detect biometric information of the finger; a lens barrel, where the lens is fixed in the lens barrel; and a support, where the support is connected to the lens barrel by means of threaded connection for supporting the lens barrel. An under-screen biometric identification apparatus and an electronic device provided in embodiments of the present application can improve the efficiency of under-screen biometric identification.

Abstract: A method (and structure and computer product) for an audiovisual source separation processing includes receiving video data showing images of a plurality of sound sources into a video encoder, while concurrently receiving into the video encoder optical flow data of the video data, the optical flow data indicating motions of pixels between frames of the video data. The video encoder encodes the received video data into video localization data comprising information associating pixels in the frames of video data with different channels of sound and encodes the received optical flow data into video separation data comprising information associating motion information in the frames of video data with the different channels of sound.

Abstract: Example embodiments described herein therefore relate to an object-model based event detection system that comprises a plurality of sensor devices, to perform operations that include: generating sensor data at the plurality of sensor devices; accessing the sensor data generated by the plurality of sensor devices; detecting an event, or precursor to an event, based on the sensor data, wherein the detected event corresponds to an event category; accessing an object model associated with the event type in response to detecting the event, wherein the object model defines a procedure to be applied by the event detection system to the sensor data; and streaming at least a portion of a plurality of data streams generated by the plurality of sensor devices to a server system based on the procedure, wherein the server system may perform further analysis or visualization based on the portion of the plurality of data streams.

Abstract: To improve the performance and accuracy of an image segmentation neural network, a cascaded robust learning framework for the segmentation of noisy labeled images includes two stages: a sample selection stage, and a joint optimization stage with label correction. In the first stage, the clean annotated samples are selected for network updating, so that the influence of noisy sample can be interactively eliminated. In the second stage, the label correction module works together with the joint optimization scheme to revise the imperfect labels. Thus, the training of the whole network is supervised by the corrected labels and the original ones.

Abstract: Method, computer readable medium, and apparatus of recognizing character zone in a digital document. In an embodiment, the method includes classifying a segment of the digital document as including text, calculating at least one parameter value associated with the classified segment of the digital document, determining, based on the calculated at least one parameter value, a zonal parameter value, classifying the segment of the digital document as a handwritten text zone or as a printed text zone based on the determined zonal parameter value and a threshold value, the threshold value being based on a selection of an intersection of a handwritten text distribution profile and a printed text distribution profile, each of the handwritten text distribution profile and the printed text distribution profile being associated with a zonal parameter corresponding to the determined zonal parameter value, and generating, based on the classifying, a modified version of the digital document.

Abstract: Some embodiments relate to sharpening segments of an image differently based on content in the image. Content based sharpening is performed by a content image processing circuit that receives luminance values of an image and a content map. The content map identifies categories of content in segments of the image. Based on one or more of the identified categories of content, the circuit determines a content factor associated with a pixel. The content factor may also be based on a texture and/or chroma values. A texture value indicates a likelihood of a category of content and is based on detected edges in the image. A chroma value indicates a likelihood of a category of content and is based on color information of the image. The circuit receives the content factor and applies it to a version of the luminance value of the pixel to generate a sharpened version of the luminance value.

Abstract: The present disclosure relates to a semantic segmentation method and system for a remote sensing image fusing GIS data. The method includes: obtaining a first remote sensing data training set and first GIS data; preprocessing the first remote sensing data training set to obtain a second remote sensing data training set; preprocessing the first GIS data based on the second remote sensing data training set to obtain second GIS data; performing data enhancement on the second remote sensing data training set to obtain a third remote sensing data training set; training a semantic segmentation model based on the third remote sensing data training set and the second GIS data; and performing semantic segmentation on the remote sensing image to be segmented based on the first GIS data and the trained semantic segmentation model to obtain a semantic set.

Abstract: Aspects of the disclosure provide a training method and device for an image enhancement model and a storage medium. The method can include inputting each training input image group into the image enhancement model to obtain a predicted image output by the image enhancement model, and training the image enhancement model until convergence through each loss function respectively corresponding to each training pair. Each loss function can include a plurality of gray scale loss components corresponding to a plurality of frequency intervals one to one, and each gray scale loss component is determined based on a difference between a gray scale frequency division image of each predicted image and a gray scale frequency division image of the corresponding target image in each frequency interval, and different gray scale loss components correspond to different frequency intervals.

Abstract: An image processing circuit for correcting a distorted image includes an internal memory and a correction circuit. The internal memory of the image processing circuit is configured to store a radial look-up table (LUT), a set of tangential LUTs, and co-ordinates of a correction center of the distorted image. The radial LUT and the set of tangential LUTs include first and second sets of parameters to correct radial and tangential distortion of the distorted image, respectively. The correction circuit is configured to reconstruct portions of the correction LUT on-the-fly based on the radial LUT, the set of tangential LUTs, and the co-ordinates of the correction center, and correct portions of the distorted image based on the reconstructed portions of correction LUT to generate the portions of the corrected image.

Abstract: Embodiments of this disclosure provide an image encoding method and apparatus and image decoding method and apparatus. The image encoding includes performing convolutional neural network (CNN) encoding on image data to generate feature vectors or feature maps; quantizing the feature vectors or feature maps to generate discrete symbols to be encoded; and estimating probabilities of the symbols to be encoded by using a multi-scale context model including multiple mask convolution layers of different scales. An entropy encoding of the image data is performed according to the probabilities of the symbols to be encoded.

Abstract: A first worker node of a distributed system computes a first set of gradients using a first neural network model and a first set of weights associated with the first neural network model. The first set of gradients are transmitted from the first worker node to a second worker node of the distributed system. The second worker node computes a first set of synchronized gradients based on the first set of gradients. While the first set of synchronized gradients are being computed, the first worker node computes a second set of gradients using a second neural network model and a second set of weights associated with the second neural network model. The second set of gradients are transmitted from the first worker node to the second worker node. The second worker node computes a second set of synchronized gradients based on the second set of gradients.

Abstract: A super-resolution method according to the inventive concept may include receiving an input image of low resolution, separating the input image into low resolution (LR) unit patches, classifying a texture type of each of pixels included in the input image using a local binary pattern for the LR unit patches, generating high resolution (HR) unit patches corresponding to each of the pixels based on a mapping kernel corresponding to the texture type, and combining the HR unit patches based on a predetermined setting to generate an output image of high resolution.

Abstract: A method includes obtaining, using at least one processor of an electronic device, multiple video frames of a video stream and multiple depth frames corresponding to the multiple video frames. The method also includes generating, using the at least one processor, multiple blur kernel maps based on the multiple depth frames. The method further includes reducing, using the at least one processor, depth errors in each of the multiple blur kernel maps. The method also includes performing, using the at least one processor, temporal smoothing on the multiple blur kernel maps to suppress temporal artifacts between different ones of the multiple blur kernel maps. In addition, the method includes generating, using the at least one processor, blur effects in the video stream using the multiple blur kernel maps.