Patents Examined by Jose L. Couso
-
Patent number: 11604975Abstract: A neural processor includes one or more neural engine circuits and a planar engine circuit. The neural engine circuits can perform convolution operations of first input data with one or more kernels to generate a first output. The planar engine circuit receives second input data that corresponds to a version of the first input data. The planar engine circuit also receives third input data that includes fourth input data and fifth input data stored together in a dimension of third input data. The planar engine circuit performs a first elementwise operation between a version of the second input data and a version of the fourth input data to generate intermediate data. The planar engine circuit performs a second elementwise operation between the intermediate data and a version of the fifth input data to generate a second output.Type: GrantFiled: April 9, 2020Date of Patent: March 14, 2023Assignee: Apple Inc.Inventors: Christopher L. Mills, Kenneth W. Waters, Youchang Kim
-
Patent number: 11599982Abstract: The present disclosure relates to systems and methods for image processing. The method may include receiving an original image including a plurality of pixels; determining a plurality of smoothed images by applying a plurality of filtering kernels to the plurality of pixels of the original image, wherein each of the plurality of filtering kernels may be associated with a respective kernel size; determining a plurality of enhanced images by comparing the original image with the plurality of smoothed images; and generating a target image based on at least one of the plurality of enhanced images and at least one of: the original image or at least one of the plurality of smoothed images.Type: GrantFiled: November 27, 2020Date of Patent: March 7, 2023Assignee: SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD.Inventors: Chunhua Jiang, Shurui Zhao, Hanyu Wang
-
Patent number: 11595655Abstract: A method and a device for encoding/decoding images are disclosed. The method for encoding images comprises the steps of: deriving a scan type of a residual signal for a current block according to whether or not the current block is a transform skip block; and applying the scan type to the residual signal for the current block, wherein the transform skip block is a block to which transform for the current block is not applied and is specified on the basis of information indicating whether or not transform for the current block is to be applied.Type: GrantFiled: December 22, 2020Date of Patent: February 28, 2023Assignees: Electronics and Telecommunications Research Institute, UNIVERSITY-INDUSTRY COOPERATION GROUP OF KYUNG HEE UNIVERSITYInventors: Hui Yong Kim, Sung Chang Lim, Jin Ho Lee, Jin Soo Choi, Jin Woong Kim, Gwang Hoon Park, Kyung Yong Kim
-
Patent number: 11586925Abstract: Disclosed is a recognition and training method and apparatus. The apparatus may include a processor configured to input data to a neural network, determine corresponding to a multiclass output a mapping function of a first class and a mapping function of a second class, acquire a result of a loss function including a first probability component that changes correspondingly to a function value of the mapping function of the first class and a second probability component that changes contrastingly to a function value of the mapping function of the second class, determine a gradient of loss corresponding to the input data based on the result of the loss function, update a parameter of the neural network based on the determined gradient of loss for generating a trained neural network based on the updated parameter. The apparatus may input other data to the trained neural network, and indicate a recognition result.Type: GrantFiled: April 6, 2018Date of Patent: February 21, 2023Assignee: Samsung Electronics Co., Ltd.Inventors: Hyun Sung Chang, Donghoon Sagong, Minjung Son
-
Patent number: 11580327Abstract: A training method for an image denoising model that can include collecting multiple sample image groups through a shooting device, each sample image group including multiple frames of sample images with a same photographic sensitivity and sample images in different sample image groups having different photographic sensitivities.Type: GrantFiled: January 13, 2021Date of Patent: February 14, 2023Assignee: Beijing Xiaomi Pinecone Electronics Co., Ltd.Inventor: Liang Zhang
-
Patent number: 11580631Abstract: Dynamic thin film interferometry is a technique used to non-invasively characterize the thickness of thin liquid films that are evolving in both space and time. Recovering the underlying thickness from the captured interferograms, unconditionally and automatically is still an open problem. A compact setup is provided employing a snapshot hyperspectral camera and the related algorithms for the automated determination of thickness profiles of dynamic thin liquid films. The technique is shown to recover film thickness profiles to within 100 nm of accuracy as compared to those profiles reconstructed through the manual color matching process. Characteristics and advantages of hyperspectral interferometry are discussed including the increased robustness against imaging noise as well as the ability to perform thickness reconstruction without considering the absolute light intensity information.Type: GrantFiled: February 15, 2021Date of Patent: February 14, 2023Assignee: The Board of Trustees of the Leland Stanford Junior UniversityInventors: Gerald G. Fuller, Vineeth Chandran Suja
-
Patent number: 11573432Abstract: A method for fitting OK lenses to a patient comprising the steps of: applying a corneal topography apparatus to a patient to capture baseline and post wear maps of a cornea of the patient in a computer to thereby derive a difference map; processing the difference map to fit Zernike polynomials thereto wherein weights of said fitted polynomials comprise features of a test feature vector for the difference map; applying the test feature vector to a classification machine trained to classify the difference map as one of a number of predetermined classes; and subsequently treating the patient taking into account the classification machine's classification.Type: GrantFiled: June 28, 2018Date of Patent: February 7, 2023Assignee: CAPRICORNIA CONTACT LENS PTY LTDInventor: Paul Gifford
-
Patent number: 11568516Abstract: A solution to the problem of image and video stitching is disclosed that compensates for the effects of lens distortion, camera misalignment, and parallax in combining multiple images. The disclosed image stitching technique includes depth or disparity estimation, alignment, and blending processes configured to be computationally efficient and produce quality results by limiting the presence of noticeable seams and artifacts in the final stitched image. An inter-frame approach applies image stitching to video frames to maintain temporal continuity between successive frames across a stitched video output having a 360-degree viewing perspective. A temporal adjustment is configured to improve temporal continuity between a subsequent frame and a previous frame in a sequence of video frames.Type: GrantFiled: September 7, 2020Date of Patent: January 31, 2023Assignee: Nikon CorporationInventors: Tanmay Nitin Bichu, Ikuya Saito, Hideo Hoshuyama
-
Patent number: 11568519Abstract: An image enhancement method, an image enhancement apparatus, and a method and apparatus for training the image enhancement apparatus are provided. The image enhancement apparatus may enhance a quality of a raw image captured by an under-display camera (UDC). The image enhancement apparatus may process the raw image for each channel according to characteristics of a photographing environment of the UDC. The image enhancement apparatus may further enhance the image by processing the raw image to reflect characteristics for each channel.Type: GrantFiled: March 9, 2021Date of Patent: January 31, 2023Assignee: Samsung Electronics Co., Ltd.Inventors: Eunhee Kang, Kinam Kwon, Sujin Lee, Hyong Euk Lee, Byung In Yoo, Sangwon Lee, Jae Seok Choi
-
Patent number: 11568581Abstract: The disclosure relates to PET imaging systems and methods. The systems may obtain a plurality of PET images of a subject and a CT image acquired by performing a spiral CT scan on the subject. Each gated PET image may include a plurality of sub-gated PET images. The CT image may include a plurality of sub-CT images each of which corresponds to one of the plurality of sub-gated PET images. The systems may determine a target motion vector field between a target physiological phase and a physiological phase of the CT image based on the plurality of sub-gated PET images and the plurality of sub-CT images. The systems may reconstruct an attenuation corrected PET image corresponding to the target physiological phase based on the target motion vector field, the CT image, and PET data used for the plurality of gated PET images reconstruction.Type: GrantFiled: November 16, 2020Date of Patent: January 31, 2023Assignee: SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD.Inventors: Tao Feng, Hongdi Li
-
Patent number: 11562177Abstract: A triple verification method is provided.Type: GrantFiled: November 27, 2019Date of Patent: January 24, 2023Assignee: Foundation of Soongsil University-Industry CooperationInventors: Young Tack Park, Jagvaral Batselem, Wan Gon Lee
-
Patent number: 11544828Abstract: A method includes producing a filter mask based on a blur mask and a saliency mask and identifying locations of a plurality of bounding boxes of a plurality of objects of interest in a received image. The method also includes applying the filter mask to the received image and to the locations of the plurality of bounding boxes in the received image to remove at least one object of interest from consideration. The method further includes performing a comparison of a location of a bounding box of the plurality of bounding boxes of an object of interest remaining in consideration to a predetermined safe region of the received image and generating a validation result based on the comparison.Type: GrantFiled: November 18, 2020Date of Patent: January 3, 2023Assignee: Disney Enterprises, Inc.Inventors: Miquel Angel Farré Guiu, Marc Junyent Martin, Pablo Pernias
-
Patent number: 11539955Abstract: An image compression method includes parsing, by a processor, an image file to be compressed; and compressing, by the processor, the image file by ignoring a portion of bits of each pixel from among a plurality of pixels of the image file.Type: GrantFiled: January 29, 2020Date of Patent: December 27, 2022Assignee: SNOW CORPORATIONInventor: Byung-Sun Park
-
Patent number: 11538136Abstract: A device includes a memory configured to store an adapted network that is configured to generate a modified image based on a single image. The device includes a processor configured to obtain, from a stream of video data, a first distorted image depicting an object, and to provide the first distorted image to the adapted network to generate a first modified image. The processor is configured to obtain, from the stream of video data, a second distorted image depicting the object, and to provide the second distorted image to the adapted network to generate a second modified image. The object is distorted differently in the second distorted image than in the first distorted image. The processor is configured to generate a video output including the first modified image and the second modified image without visible artifacts due to distortion differences between the first distorted image and the second distorted image.Type: GrantFiled: October 28, 2020Date of Patent: December 27, 2022Assignee: Qualcomm IncorporatedInventors: Stone Yun, Alireza Shoa Hassani Lashdan, Darren Gnanapragasam, David Hansen
-
Patent number: 11531897Abstract: Disclosed is a recognition and training method and apparatus. The apparatus may include a processor configured to input data to a neural network, determine corresponding to a multiclass output a mapping function of a first class and a mapping function of a second class, acquire a result of a loss function including a first probability component that changes correspondingly to a function value of the mapping function of the first class and a second probability component that changes contrastingly to a function value of the mapping function of the second class, determine a gradient of loss corresponding to the input data based on the result of the loss function, update a parameter of the neural network based on the determined gradient of loss for generating a trained neural network based on the updated parameter. The apparatus may input other data to the trained neural network, and indicate a recognition result.Type: GrantFiled: April 6, 2018Date of Patent: December 20, 2022Assignee: Samsung Electronics Co., Ltd.Inventors: Hyun Sung Chang, Donghoon Sagong, Minjung Son
-
Patent number: 11528426Abstract: Disclosed are an image sensing device and an operating method thereof, and the image sensing device may include: an image sensor including a plurality of pixels and suitable for generating an image based on incident light; and an image processor suitable for generating a high dynamic range (HDR) image based on the image and two or more pieces of tone mapping information, which are divided according to luminance.Type: GrantFiled: September 18, 2020Date of Patent: December 13, 2022Assignee: SK hynix Inc.Inventors: Su Ram Cha, Seong Hee Park
-
Patent number: 11526980Abstract: The invention relates to a method for processing images obtained by a diffraction detector, of a crystalline or polycrystalline material, in which a first image of the material is acquired in a state of reference as well as a second image of the material in a deformed state. The invention is characterised in that, in a calculator, during a first step (E6, E12), a current elastic deformation gradient tensor Fe is given a value determined by calculation, during a second step (E7), the current displacement field induced by the tensor Fe is calculated, during a third step (E8), third digital values of a deformed image {hacek over (g)}(x)=g(x+u(x)) corrected by the current displacement field are calculated, and during an iterative algorithm, iterations of the second and third steps (E12, E7, E8) are carried out on modified values of the tensor r Fe until a convergence criterion is met in relation to the correction to the current value of Fe.Type: GrantFiled: December 7, 2018Date of Patent: December 13, 2022Inventors: Félix Latourte, Qiwei Shi, François Hild, Stéphane Roux
-
Patent number: 11521305Abstract: Provided are an image processing method and device, a mobile terminal, and a storage medium. The method includes: acquiring a first image output by an image sensor; responsive to that a first signal value of the first image after being processed exceeds a saturation value of a first bit width at a preset image processing stage, recording the first signal value of the processed first image as a second signal value with a second bit width, the second bit width being greater than the first bit width; and mapping the second signal value recorded with the second bit width to a third signal value recorded with the first bit width to obtain a second image. Through the method, an operation of increasing a bit width can be used to retain more information of an image and improve the quality of the image, and an implementation is simple and effective.Type: GrantFiled: August 31, 2020Date of Patent: December 6, 2022Assignee: Beijing Xiaomi Mobile Software Co., Ltd.Inventor: Jing Xu
-
Patent number: 11521131Abstract: A machine learning model can be trained using a first set of degraded images for each of a plurality of combinations and corresponding reference images, where a number of degraded images in the first set corresponding to a particular combination of the plurality of combinations is selected in accordance with a probability value associated with the particular combination. A validation process can be used to determine a loss value for each of the plurality of combinations of degradations. Updates to the probability values associated with the plurality of combinations can be calculated based on the loss values. The machine learning model can be updated using a second set of degraded images for each of the plurality of combinations, and the corresponding reference images, where a number of degraded images in the second set corresponding to the particular combination is selected based on the updated probability value.Type: GrantFiled: January 24, 2019Date of Patent: December 6, 2022Assignee: Jumio CorporationInventors: Sai Narsi Reddy Donthi Reddy, Qiqin Dai
-
Patent number: 11514371Abstract: In one embodiment, a computing system may receive an uncompressed image from a camera. The computing system may generate a compressed image by performing a compression process on the uncompressed image, wherein a decompressed image may be generated as a byproduct during the compression process. The computing system may send the decompressed image to a machine-learning model that was trained using decompressed images. The computing system may generate, by the machine-learning model, an output based on the decompressed image. The computing system may provide operational instructions to a vehicle based on the output.Type: GrantFiled: March 11, 2019Date of Patent: November 29, 2022Assignee: Woven Planet North America, Inc.Inventors: Forrest Samuel Briggs, James Allen-White Hoffacker, Dhruv Lamba, Yi Lu, Phillip Sawbridge