Image processing device and operating method therefor
An image processing apparatus includes: a memory storing one or more instructions; and a processor configured to execute the one or more instructions to: obtain similarity information, based on a pixel difference between each pixel included in a first image and an adjacent pixel of each pixel included in the first image; determine weight information based on the similarity information; obtain first feature information by performing a convolution operation between the first image and a first kernel; obtain second feature information by applying the weight information to the first feature information; and generate a second image based on the second feature information.
1 . An image processing apparatus comprising:
a memory storing one or more instructions; and
a processor configured to execute the one or more instructions to:
obtain first similarity information, based on a pixel difference between each pixel included in a first image and a first adjacent pixel immediately to the right of each pixel included in the first image, wherein the first image comprises w×h pixels and the first similarity information comprises w×h elements;
obtain second similarity information, based on a pixel difference between each pixel included in the first image and a second adjacent pixel immediately to the left of each pixel included in the first image, wherein the second similarity information comprises w×h elements;
determine weight information based on the first similarity information and the second similarity information, wherein the weight information comprises w×h elements;
obtain first feature information by performing a convolution operation between the first image and a first kernel, wherein the first feature information comprises w×h elements;
obtain second feature information by performing an elementwise multiplication operation of the weight information and the first feature information; and
generate a second image based on the second feature information.
2 . The image processing apparatus of claim 1 , wherein the processor is further configured to execute the one or more instructions to determine the weight information such that a weight value increases as the pixel difference between each pixel included in the first image and the first adjacent pixel decreases, and the weight value decreases as the pixel difference between each pixel included in the first image and the first adjacent pixel increases.
3 . The image processing apparatus of claim 1 , wherein the processor is further configured to execute the one or more instructions to determine the weight information by applying a natural exponential function or a softmax function to the first similarity information and the second similarity information.
4 . The image processing apparatus of claim 1 , wherein a number of channels of the first kernel is determined such that a number of channels of the weight information is equal to a number of channels of the first feature information.
5 . The image processing apparatus of claim 1 , wherein the processor is further configured to execute the one or more instructions to:
obtain third feature information by performing a convolution operation between the second feature information and a second kernel, and
generate the second image based on the third feature information.
6 . The image processing apparatus of claim 1 , wherein the processor is further configured to execute the one or more instructions to:
obtain second weight information through a convolution operation between the weight information and a third kernel,
obtain fourth feature information by applying the second weight information to the second feature information, and
generate the second image based on the fourth feature information.
7 . The image processing apparatus of claim 6 , wherein the processor is further configured to execute the one or more instructions to obtain the fourth feature information by performing the elementwise multiplication operation of the second feature information and the second weight information.
8 . The image processing apparatus of claim 1 , wherein the processor is further configured to execute the one or more instructions to generate the second image by processing the first image by using a neural network including a plurality of layers,
wherein the first image is input to an n-th layer among the plurality of layers, and
wherein the second feature information is input to a (n+1)th layer among the plurality of layers.
9 . An operating method of an image processing apparatus, the operating method comprising:
obtaining first similarity information, based on a pixel difference between each pixel included in a first image and a first adjacent pixel immediately to the right of each pixel included in the first image, wherein the first image comprises w×h pixels and the first similarity information comprises w×h elements;
obtaining second similarity information, based on a pixel difference between each pixel included in the first image and a second adjacent pixel immediately to the left of each pixel included in the first image, wherein the second similarity information comprises w×h elements;
determining weight information based on the first similarity information and the second similarity information, wherein the weight information comprises w×h elements;
obtaining first feature information by performing a convolution operation between the first image and a first kernel, wherein the first feature information comprises w×h elements;
obtaining second feature information by performing an elementwise multiplication operation of the weight information and the first feature information; and
generating a second image based on the second feature information.
10 . The operating method of claim 9 , wherein the determining the weight information comprises determining the weight information by applying a natural exponential function or a softmax function to the first similarity information and the second similarity information.
11 . The operating method of claim 9 , wherein a number of channels of the first kernel is determined such that a number of channels of the weight information is equal to a number of channels of the first feature information.
12 . A non-transitory computer-readable recording medium having stored thereon a program, which when executed by a computer, performs a method comprising:
obtaining first similarity information, based on a pixel difference between each pixel included in a first image and a first adjacent pixel immediately to the right of each pixel included in the first image of each pixel included in the first image, wherein the first image comprises w×h pixels and the first similarity information comprises w×h elements;
obtaining second similarity information, based on a pixel difference between each pixel included in the first image and a second adjacent pixel immediately to the left of each pixel included in the first image, wherein the second similarity information comprises w×h elements;
determining weight information based on the first similarity information and the second similarity information, wherein the weight information comprises w×h elements;
obtaining first feature information by performing a convolution operation between the first image and a first kernel, wherein the first feature information comprises w×h elements;
obtaining second feature information by performing an elementwise multiplication operation of the weight information and the first feature information; and
generating a second image based on the second feature information.