US 12,096,131 B2
Method of generating noise-reduced image data and electronic device for performing the same
Quockhanh Dinh, Suwon-si (KR); Kyonghwan Jin, Suwon-si (KR); Youngo Park, Suwon-si (KR); and Kwangpyo Choi, Suwon-si (KR)
Assigned to SAMSUNG ELECTRONICS CO., LTD., Suwon-si (KR)
Filed by SAMSUNG ELECTRONICS CO., LTD., Suwon-si (KR)
Filed on Oct. 27, 2021, as Appl. No. 17/512,338.
Application 17/512,338 is a continuation of application No. PCT/KR2021/012433, filed on Sep. 13, 2021.
Claims priority of application No. 10-2020-0140681 (KR), filed on Oct. 27, 2020; and application No. 10-2020-0149889 (KR), filed on Nov. 11, 2020.
Prior Publication US 2022/0132001 A1, Apr. 28, 2022
Int. Cl. G06T 7/90 (2017.01); G06F 18/214 (2023.01); G06F 18/40 (2023.01); G06N 20/00 (2019.01); G06T 3/4015 (2024.01); G06T 3/4038 (2024.01); G06T 5/20 (2006.01); G06T 5/70 (2024.01); H04N 23/81 (2023.01); H04N 23/90 (2023.01)
CPC H04N 23/81 (2023.01) [G06F 18/2148 (2023.01); G06F 18/40 (2023.01); G06N 20/00 (2019.01); G06T 3/4015 (2013.01); G06T 3/4038 (2013.01); G06T 5/20 (2013.01); G06T 5/70 (2024.01); G06T 7/90 (2017.01); H04N 23/90 (2023.01); G06T 2207/10016 (2013.01); G06T 2207/10024 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20182 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A method, performed by an electronic device including a camera module, of generating noise-reduced image data of an object, the method comprising:
obtaining, by a second processor of the electronic device, from a first processor of the electronic device, a control signal for obtaining image data;
storing a learning model into a memory;
obtaining, by using the camera module, raw image data of the object obtained from light received by the camera module, the raw image data having a color array of a plurality of colors with respect to a plurality of pixels;
obtaining, by using the stored learning model, a color data set of the plurality of pixels from the obtained raw image data, the color data set comprising a plurality of pieces of color data representing the plurality of colors; and
obtaining the noise-reduced image data of the object by using the obtained color data set,
wherein the learning model is trained by using input data and target data,
wherein the target data is obtained based on a standard color data set comprising a plurality of pieces of standard color data classified according to the plurality of colors, the plurality of pieces of standard color data being obtained by applying a down-sampling filter to sample raw image data, and
wherein the input data is obtained based on training raw image data and random noise generated by a noise generator, the training raw image data being obtained by performing mosaicing based on the standard color data set.