Holographic image processing method and holographic image processing apparatus
Provided is a holographic image processing apparatus including a memory configured to store at least one instruction, and a processor configured to execute the at least one instruction stored in the memory to generate a corrected holographic image by correcting an original holographic image captured by a holographic camera based on a neural network configured to learn hologram correction in advance.
1. A holographic image processing apparatus comprising:
a memory configured to store at least one instruction; and
a processor configured to execute the at least one instruction stored in the memory to generate a corrected holographic image by correcting an original holographic image captured by a holographic camera based on a neural network configured to learn hologram correction in advance,
wherein the processor is further configured to generate a training data set by propagating N original images captured by the holographic camera to a preset depth, and
wherein the neural network is further configured to receive the training data set to generate a corrected hologram set comprising corrected holographic images, generate a reproduction image set by propagating corrected holographic images to depths corresponding to original depths of the N original images, and obtain a loss value by comparing the reproduction image set with an original image set.
2. The holographic image processing apparatus of claim 1 , wherein the processor is further configured to propagate the original holographic image to the preset depth and input the original holographic image to the neural network.
3. The holographic image processing apparatus of claim 1 , wherein the N holograms are obtained as the holographic camera captures each of the N original images located at different depths corresponding to the original depths from the holographic camera.
4. The holographic image processing apparatus of claim 3 , wherein the N original images comprise two-dimensional planar images.
5. The holographic image processing apparatus of claim 1 ,
wherein the neural network is further configured to learn the hologram correction in a direction that the loss value reduces.
6. The holographic image processing apparatus of claim 5 , wherein the processor is further configured to generate the reproduction image set by numerically reproducing the corrected hologram set.
7. The holographic image processing apparatus of claim 1 , wherein the neural network comprises a resnet block.
8. A holographic image processing method comprising:
training a neural network based on a training data set generated based on propagating N original images captured by a holographic camera to a preset depth; and
generating a corrected holographic image by correcting an original holographic image captured by the holographic camera based on the neural network configured to learn hologram correction in advance,
wherein the training of the neural network comprises generating a corrected hologram set comprising corrected holographic images by inputting the training data set to the neural network, generating a reproduction image set by propagating corrected holographic images to depths corresponding to original depths of the N original images, and obtaining a loss value by comparing the reproduction image set with an original image set.
9. The holographic image processing method of claim 8 , further comprising propagating the original holographic image to a specific depth and inputting the original holographic image to the neural network.
10. The holographic image processing method of claim 8 , wherein the N holograms are obtained as the holographic camera captures each of the N original images located at different depths corresponding to the original depths from the holographic camera.
11. The holographic image processing method of claim 10 , wherein the N original images comprise two-dimensional planar images.
12. The holographic image processing method of claim 8 ,
wherein the training of the neural network comprises learning the hologram correction in a direction that the loss value reduces.
13. The holographic image processing method of claim 12 , wherein, in the generating of the reproduction image set, the reproduction image set is generated by numerically reproducing the corrected hologram set.
14. A holographic image processing system comprising:
a holographic camera configured to capture an original holographic image;
a holographic image processing apparatus comprising:
a memory configured to store at least one instruction; and
a processor configured to execute the at least one instruction stored in the memory to generate a corrected holographic image by correcting the original holographic image based on a neural network configured to learn hologram correction in advance,
wherein the processor is further configured to generate a training data set by propagating N original images captured by the holographic camera to a preset depth, and
wherein the neural network is further configured to receive the training data set to generate a corrected hologram set comprising corrected holographic images, generate a reproduction image set by propagating corrected holographic images to depths corresponding to original depths of the N original images, and obtain a loss value by comparing the reproduction image set with an original image set.
15. The holographic image processing apparatus of claim 14 , wherein the processor is further configured to propagate the original holographic image to the preset depth and input the original holographic image to the neural network.
16. The holographic image processing apparatus of claim 14 , wherein the N holograms are obtained as the holographic camera captures each of the N original images located at different depths corresponding to the original depths from the holographic camera.
17. The holographic image processing apparatus of claim 16 , wherein the N original images comprise two-dimensional planar images.