Image processing method and electronic device

An electronic device, comprising:

• at least one processor;
• and
• a storage device storing one or more programs which when executed by the at least one processor, causes the at least one processor to:
• obtain a content image, a texture image, and a to-be-optimized image, contents of the to-be-optimized image is consistent with contents of the content image;
• extract a first object mask of the texture image and a second object mask of the to-be-optimized image;
• input the content image, the texture image, the to-be-optimized image, the first object mask, and the second object mask into an image recognition model;
• obtain a first content matrix corresponding to the content image, a first texture matrix corresponding to the texture image, a second content matrix and a second texture matrix corresponding to the to-be-optimized image, a first mask matrix corresponding to the first object mask, and a second mask matrix corresponding to the second object mask by the image recognition model;
• divide the first texture matrix into a plurality of first sub-matrices, and the second texture matrix into a plurality of second sub-matrices;
• calculate object loss and background loss according to the plurality of first sub-matrices, the plurality of second sub-matrices, the first mask matrix, and the second mask matrix, and calculate texture loss according to the object loss and the background loss;
• calculate content loss according to the first content matrix and the second content matrix;
• determine a total loss according to the texture loss and the content loss;
• transmit the total loss to the image recognition model by using a gradient descent algorithm, and obtain pixel loss of each pixel of the to-be-optimized image;
• adjust a value of each pixel of the to-be-optimized image to minimize the total loss according to the pixel loss, and obtain an optimized image.

A non-transitory storage medium having stored thereon instructions that, when executed by a processor of an electronic device, causes the electronic device to perform an image processing method, the method comprising:

• obtaining a content image, a texture image, and a to-be-optimized image, wherein contents of the to-be-optimized image is consistent with contents of the content image;
• extracting a first object mask of the texture image and a second object mask of the to-be-optimized image;
• inputting the content image, the texture image, the to-be-optimized image, the first object mask, and the second object mask into an image recognition model;
• obtaining a first content matrix corresponding to the content image, a first texture matrix corresponding to the texture image, a second content matrix and a second texture matrix corresponding to the to-be-optimized image, a first mask matrix corresponding to the first object mask, and a second mask matrix corresponding to the second object mask by the image recognition model;
• dividing the first texture matrix into a plurality of first sub-matrices, and the second texture matrix into a plurality of second sub-matrices;
• calculating object loss and background loss according to the plurality of first sub-matrices, the plurality of second sub-matrices, the first mask matrix, and the second mask matrix, and calculating texture loss according to the object loss and the background loss;
• calculating content loss according to the first content matrix and the second content matrix;
• determining a total loss according to the texture loss and the content loss;
• transmitting the total loss to the image recognition model by using a gradient descent algorithm, and obtaining pixel loss of each pixel of the to-be-optimized image;
• adjusting a value of each pixel of the to-be-optimized image to minimize the total loss according to the pixel loss, and obtaining an optimized image.