Image processing method and apparatus

1. An image processing method, comprising:

for each of a plurality of input images, extracting a content latent code of an input image based on a content encoder configured to extract a feature of a content of an image; and

extracting a style latent code of the input image based on a style encoder configured to extract a feature of a style of an image;

obtaining a content feature vector by calculating a weighted sum of content latent codes extracted from the input images based on a morphing control parameter;

obtaining a style feature vector by calculating a weighted sum of style latent codes extracted from the input images based on the morphing control parameter; and

generating a morphing image based on the content feature vector, the style feature vector, and a decoder configured to generate an image from an embedding vector;

wherein a morphing generator including the content encoder, the style encoder and the decoder is trained based on at least one of:

an adversarial loss associated with a discrimination between an image comprised in a training data and an output morphing image obtained from the morphing generator; or

a pixel-wise reconstruction loss associated with a difference between pixels of the image comprised in the training data and pixels of the output morphing image,

wherein, the training data includes a first image, a second image, and a ground truth morphing image,

wherein the ground truth morphing image is generated by linearly interpolating a first point corresponding to the first image sampled in a latent space by a pretrained model and a second point corresponding to the second image sampled in the latent space by the pretrained model.

2. The image processing method of claim 1 , wherein the morphing control parameter comprises a content morphing control parameter and a style morphing control parameter,

wherein the obtaining of the content feature vector comprises:

obtaining the content feature vector by calculating a weighted sum of content latent codes extracted from the input images based on the content morphing control parameter, and

the obtaining of the style feature vector comprises:

obtaining the style feature vector by calculating a weighted sum of style latent codes extracted from the input images based on the style morphing control parameter.

3. The image processing method of claim 1 , wherein the input images comprise a first image and a second image,

wherein the obtaining of the content feature vector comprises:

interpolating a first content latent code extracted from the first image and a second latent content code extracted from the second image based on the content morphing control parameter, and

the obtaining of the style feature vector comprises:

interpolating a first style latent code extracted from the first image and a second latent style code extracted from the second image based on the style morphing control parameter.

4. The image processing method of claim 1 , wherein the morphing control parameter comprises weights corresponding to the input images.

5. The image processing method of claim 1 , wherein the generating of the morphing image comprises:

obtaining a processed style feature vector by applying the style feature vector to a mapping network; and

generating the morphing image by decoding the processed style feature vector and the content feature vector based on the decoder.

6. The image processing method of claim 1 , wherein the content of an image comprises information associated with a pose of an object in the image, a position of the object in the image, and a composition of a camera capturing the image of the object, and

the style of an image comprises information associated with an appearance of an object in the image.

7. The image processing method of claim 1 , wherein the input images are different in at least one of the content or the style.

8. The image processing method of claim 1 , wherein the morphing image comprises a combination of one or more content features among content features of the input images, and a combination of one or more style features among style features of the input images.

9. The image processing method of claim 1 , wherein the input images comprise a plurality of frames of a video,

wherein the morphing image comprises an intermediate frame of the frames of the video.

10. 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 .

11. An image processing apparatus, comprising:

at least one processor;

wherein, for each of a plurality of input images, the processor is configured to:

extract a content latent code by encoding an input image;

extract a style latent code by encoding the input image;

obtain a content feature vector by calculating a weighted sum of content latent codes extracted from the input images based on a morphing control parameter;

obtain a style feature vector by calculating a weighted sum of style latent codes extracted from the input images based on the morphing control parameter; and

generate a morphing image based on the content feature vector and the style feature vector;

wherein the apparatus includes a morphing generator which is trained based on at least one of:

an adversarial loss associated with a discrimination between an image comprised in a training data and an output morphing image obtained from the morphing generator; or

a pixel-wise reconstruction loss associated with a difference between pixels of the image comprised in the training data and pixels of the output morphing image,

wherein, the training data includes a first image, a second image, and a ground truth morphing image,

wherein the ground truth morphing image is generated by linearly interpolating a first point corresponding to the first image sampled in a latent space by a pretrained model and a second point corresponding to the second image sampled in the latent space by the pretrained model.

12. The image processing apparatus of claim 11 , wherein the morphing control parameter comprises a content morphing control parameter and a style morphing control parameter,

wherein the processor is configured to:

for the obtaining of the content feature vector, obtain the content feature vector by calculating a weighted sum of content latent codes extracted from the input images based on the content morphing control parameter; and

for the obtaining of the style feature vector, obtain the style feature vector by calculating a weighted sum of style latent codes extracted from the input images based on the style morphing control parameter.

13. The image processing apparatus of claim 11 , wherein the input images comprise a first image and a second image,

wherein the processor is configured to:

for the obtaining of the content feature vector, interpolate a first content latent code extracted from the first image and a second content latent code extracted from the second image based on the content morphing control parameter; and

for the obtaining of the style feature vector, interpolate a first style latent code extracted from the first image and a second style latent code extracted from the second image based on the style morphing control parameter.

14. The image processing apparatus of claim 11 , wherein, for the generating of the morphing image, the processor is configured to:

obtain a processed style feature vector by applying the style feature vector to a mapping network; and

generate the morphing image by decoding the processed style feature vector and the content feature vector.

15. The image processing apparatus of claim 11 , wherein the morphing control parameter comprises weights corresponding to the input images.