公开(公告)号：US20230132630A1

公开(公告)日：2023-05-04

申请号：US17862779

申请日：2022-07-12

申请人： SAMSUNG ELECTRONICS CO., LTD. ,  Seoul National University R&DB Foundation

发明人： Eunhee KANG ,  Minsoo KANG ,  Bohyung HAN ,  Sehwan KI ,  HYONG EUK LEE

IPC分类号： G06N3/04 ,  G06N3/08

摘要： A method includes: generating, based on a student network result of an implemented student network provided with an input, a sample corresponding to a distribution of an energy-based model based on the student network result and a teacher network result of an implemented teacher network provided with the input; training model parameters of the energy-based model to decrease a value of the energy-based model, based on the teacher network result and the student network result; and training the implemented student network to increase the value of the energy-based model, based on the sample and the student network result.

公开(公告)号：US20240153035A1

公开(公告)日：2024-05-09

申请号：US18330654

申请日：2023-06-07

申请人： SAMSUNG ELECTRONICS CO., LTD.

发明人： Younghyun JO ,  Sehwan KI ,  Eunhee KANG ,  Hyong Euk LEE

IPC分类号： G06T3/40

CPC分类号： G06T3/4076 ,  G06T3/4046

摘要： A processor-implemented method includes generating an adjusted reference patch by adjusting a position of a reference patch in a reference image based on a pixel value of a ground truth (GT) patch of a GT image and a pixel value of the reference patch, wherein the GT patch corresponds to a specific region of an input image; generating a super-resolution (SR) image of the input image using a SR model provided an input that is based on the generated adjusted reference patch; and training the SR model based on the SR image and the GT image.

公开(公告)号：US20240223917A1

公开(公告)日：2024-07-04

申请号：US18350240

申请日：2023-07-11

申请人： Samsung Electronics Co., Ltd.

发明人： Sehwan KI ,  Younghyun JO ,  Eunhee KANG ,  Hyong Euk LEE

IPC分类号： H04N25/46 ,  G06T3/40 ,  H04N9/78 ,  H04N23/69 ,  H04N23/85 ,  H04N25/10

CPC分类号： H04N25/46 ,  G06T3/4046 ,  G06T3/4053 ,  H04N9/78 ,  H04N23/69 ,  H04N23/85 ,  H04N25/10

摘要： A method and apparatus for generating a high-resolution digital zoom image may obtain a raw image captured by an image sensor from the image sensor, obtain an RGB image corresponding to the raw image, separate a luminance component and a chrominance component from the RGB image, extract a first feature of the RGB image, of which color information is enhanced, by applying the chrominance component to a first neural network, extract a second feature of the RGB image, of which texture information is enhanced, by applying the raw image and the luminance component to a second neural network, and generate the high-resolution digital zoom image, which corresponds to the RGB image, by upscaling the color information and the texture information based on the first feature or the second feature.

公开(公告)号：US20240054606A1

公开(公告)日：2024-02-15

申请号：US18337774

申请日：2023-06-20

申请人： Samsung Electronics Co., Ltd.

发明人： Geonseok SEO ,  Sehwan KI ,  Jae Seok CHOI ,  Insoo KIM ,  Hyong Euk LEE

IPC分类号： G06T5/00 ,  G06T5/50

CPC分类号： G06T5/001 ,  G06T5/50 ,  G06T2207/20084

摘要： A processor-implemented method includes: obtaining a plurality of image frames acquired for a scene within a predetermined time; determining loss values respectively corresponding to the plurality of image frames; determining a reference frame among the plurality of image frames based on the loss values; and generating a final image of the scene based on the reference frame.

公开(公告)号：US20230071693A1

公开(公告)日：2023-03-09

申请号：US17869165

申请日：2022-07-20

申请人： Samsung Electronics Co., Ltd.

发明人： Eunhee KANG ,  Sehwan KI ,  Nahyup KANG ,  Kinam KWON ,  Hyong Euk LEE ,  Jae Seok CHOI

IPC分类号： G06T5/00

摘要： A processor-implemented method with degraded image restoration includes: receiving a degraded training image; training a first teacher network of an image restoration network and a second teacher network of the image restoration network to infer differential images corresponding to the degraded training image, wherein each of the first teacher network and the second teacher network comprises a differentiable activation layer and a performance of the first teacher network is greater than that of the second teacher network; initially setting a student network of the image restoration network based on the second teacher network; and training the student network to infer a differential image corresponding to the degraded training image by iteratively backpropagating, to the student network, a contrastive loss that decreases a first difference between a third output of the student network and a first output of the first teacher network and increases a second difference between the third output and a second output of the second teacher network.