Image processing method and image processing apparatus

1. An image processing method, comprising:

receiving an input image and a guide image corresponding to the input image;

extracting informative features from the input image and the guide image to enhance the input image;

selectively obtaining a first feature for the input image from among the informative features; and

processing the input image based on the first feature,

wherein the selectively obtaining of the first feature comprises selectively obtaining the first feature for each channel of the input image from among the informative features by applying the informative features to a first attention network that uses channel attention.

2. The image processing method of claim 1 , wherein the input image comprises a depth image, and

the guide image comprises an infrared (IR) image.

3. The image processing method of claim 2 , wherein the guide image comprises a position of an object in the guide image is same as the position of the object in the input image.

4. The image processing method of claim 3 , wherein the guide image comprises any one or any combination of features a normal direction of a surface of the object, a direction of light incident to the object, and a distance to the object.

5. The image processing method of claim 1 , wherein the extracting of the informative features comprises at least one of:

determining the informative features from the input image and the guide image by supervised learning; or

determining the informative features from the input image and the guide image by unsupervised learning.

6. The image processing method of claim 5 , wherein the extracting comprises:

estimating a portion of the informative features by the supervised learning; and

estimating a remaining portion of the informative features, other than the portion, by the unsupervised learning.

7. The image processing method of claim 1 , wherein the informative features comprise any one or any combination of an edge of the input image, an edge of the guide image, a gradient of the input image, a gradient of the guide image, a first surface normal map having a magnitude of each direction vector of a surface normal estimated from the input image as an intensity, an image having an angle of a normal estimated from the input image as an intensity, a depth intensity estimated from the input image, a second surface normal map having a magnitude of each direction vector of a surface normal estimated from the guide image as an intensity, an image having an angle of a normal estimated from the guide image as an intensity, and a segmentation map corresponding to the guide image.

8. The image processing method of claim 1 , wherein the selectively obtaining of the first feature comprises selectively obtaining the first feature from among the informative features using the first attention network.

9. The image processing method of claim 8 , wherein the selectively obtaining of the first feature comprises selectively obtaining the first feature for at least one of a space or a channel of the input image from among the informative features using the first attention network.

10. The image processing method of claim 1 , further comprising:

extracting image features from the input image; and

selectively obtaining a second feature for the input image from among the image features, wherein the processing of the input image comprises processing the input image based on the first feature and the second feature.

11. The image processing method of claim 10 wherein the processing of the input image comprises processing on the input image by aggregating the first feature and the second feature.

12. The image processing method of claim 10 , wherein the selectively obtaining of the second feature comprises selectively obtaining the second feature for each region of the input image by applying the image features to a second attention network that uses spatial attention.

13. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the image processing method of claim 1 .

14. An image processing apparatus, comprising:

a communication interface configured to receive an input image and a guide image corresponding to the input image; and

a processor configured to extract informative features from the input image and the guide image to enhance the input image, to selectively obtain a first feature for the input image from among the informative features, and to process the input image based on the first feature,

wherein the selectively obtaining of the first feature comprises selectively obtaining the first feature for each channel of the input image from among the informative features by applying the informative features to a first attention network that uses channel attention.

15. The image processing apparatus of claim 14 , wherein the processor is further configured to perform at least one of determine the informative features from the input image and the guide image by supervised learning, or to determine the informative features from the input image and the guide image by unsupervised learning.

16. The image processing apparatus of claim 15 , wherein the processor is further configured to:

estimate a portion of the informative features by the supervised learning, and to estimate a remaining portion of the informative features, other than the portion, by the unsupervised learning.

17. The image processing apparatus of claim 14 , wherein the informative features comprise any one or any combination of an edge of the input image, an edge of the guide image, a gradient of the input image, a gradient of the guide image, a first surface normal map having a magnitude of each direction vector of a surface normal estimated from the input image as an intensity, an image having an angle of a normal estimated from the input image as an intensity, a depth intensity estimated from the input image, a second surface normal map having a magnitude of each direction vector of a surface normal estimated from the guide image as an intensity, an image having an angle of a normal estimated from the guide image as an intensity, and a segmentation map corresponding to the guide image.

18. The image processing apparatus of claim 14 , wherein the processor is further configured to selectively obtain the first feature for at least one of a space or a channel of the input image from among the informative features using the first attention network.

19. The image processing apparatus of claim 14 , wherein the processor is further configured to extract image features from the input image, to selectively obtain a second feature for the input image from among the image features, and to process the input image by aggregating the first feature and the second feature.

20. The image processing apparatus of claim 19 , wherein the processor is further configured to selectively obtain the second feature for each region of the input image by applying the image features to a second attention network that uses spatial attention.

21. An image processing method, comprising:

receiving an input image and a guide image corresponding to the input image;

extracting informative features from the input image and the guide image;

selectively obtaining a first feature for each channel of the input image from among the informative features by applying the informative features to a first attention network that uses channel attention;

selectively obtaining a second feature for the input image from among the informative features using a second attention network that uses spatial attention; and

processing the input image based on the first feature and the second feature.

22. The image processing method of claim 21 , further comprising denoising the input image, and the extracting of the informative features comprises extracting the informative features from the denoised input image and the guide image.

23. The image processing method of claim 21 , further comprising weighing the input image based on the second feature, and the extracting of the informative features comprises extracting the informative features from the weighted input image and the guide image.