US 12,322,073 B2
Image processing model generation method, processing method, storage medium, and terminal
Songnan Li, Huizhou (CN); Yu Zhang, Huizhou (CN); and Dahai Yu, Huizhou (CN)
Assigned to TCL TECHNOLOGY GROUP CORPORATION, Huizhou (CN)
Filed by TCL TECHNOLOGY GROUP CORPORATION, Huizhou (CN)
Filed on Jul. 14, 2022, as Appl. No. 17/865,340.
Application 17/865,340 is a continuation of application No. PCT/CN2020/141932, filed on Dec. 31, 2020.
Claims priority of application No. 202010162684.0 (CN), filed on Mar. 10, 2020; application No. 202010162709.7 (CN), filed on Mar. 10, 2020; and application No. 202010163472.4 (CN), filed on Mar. 10, 2020.
Prior Publication US 2022/0398698 A1, Dec. 15, 2022
Int. Cl. G06K 9/00 (2022.01); G06T 5/50 (2006.01); G06T 5/70 (2024.01); G06T 5/73 (2024.01); G06T 5/77 (2024.01)
CPC G06T 5/77 (2024.01) [G06T 5/50 (2013.01); G06T 5/70 (2024.01); G06T 5/73 (2024.01); G06T 2207/20081 (2013.01)] 25 Claims
OG exemplary drawing
 
1. An image processing method comprising:
obtaining a to-be-processed image captured by an under-screen camera;
performing an image processing operation on the to-be-processed image by inputting the to-be-processed image into a corresponding image processing model to obtain a processed image, the image processing operation including at least ghost effect removal processing, wherein the image processing model corresponding to color deviation removal processing is a first image processing model, and the image processing model corresponding to the ghost effect removal processing is a second image processing model; and
obtaining an output image according to the processed image of the corresponding image processing model,
wherein each image processing model is generated by:
generating a generated image corresponding to a first image by a predetermined network model according to the first image of a training image set, the training image set including a plurality of training image groups, each training image group including the first image and a second image, wherein the first image is a color deviation image of the second image for the first image processing model, or the first image is a ghost effect image of the second image for the second image processing model; and
correcting a model parameter by the predetermined network model according to the second image corresponding to the first image and the generated image corresponding to the first image, and continuing to generate an updated generated image corresponding to a next first image according to the next first image in a next training image group of the training image set until a training condition of the predetermined network model satisfies a predetermined condition,
wherein the first image is generated based on an under-screen imaging system, and the second image describes a same content as the first image and is acquired without the under-screen imaging system
wherein: the predetermined network model for generating the second image processing model corresponding to the ghost effect removal processing includes an encoder and a decoder; and
generating the generated image corresponding to the first image by the predetermined network model according to the first image of the training image set includes:
inputting the first image of the training image set into the encoder to obtain a feature image of the first image through the encoder, an image size of the feature image being smaller than an image size of the first image; and
inputting the feature image into the decoder to output the generated image through the decoder, an image size of the generated image being equal to the image size of the first image.