Artificial intelligence-based image encoding and decoding apparatus and method

1. A method of reconstructing an optical flow by using artificial intelligence (AI), the method comprising:

obtaining, from a bitstream, feature data of a current residual optical flow for a current image;

obtaining the current residual optical flow by applying the feature data of the current residual optical flow to a neural-network-based first decoder;

obtaining a second-order optical flow between a current predicted optical flow and a previous optical flow; and

obtaining the current predicted optical flow by modifying the previous optical flow according to the second-order optical flow; and

reconstructing a current optical flow based on the current residual optical flow and the current predicted optical flow.

2. The method of claim 1 , wherein the current image is reconstructed based on current residual image data and a current predicted image generated based on a previous reconstructed image and the reconstructed current optical flow.

3. The method of claim 1 , wherein the obtaining of the second-order optical flow comprises:

obtaining the second-order optical flow by applying at least one of the previous optical flow, feature data of the previous optical flow, and feature data of a previous residual optical flow to a second prediction neural network.

4. The method of claim 1 , wherein the obtaining of the second-order optical flow comprises:

obtaining, from the bitstream, feature data of the second-order optical flow;

obtaining the second-order optical flow by applying the feature data of the second-order optical flow to a neural-network-based third decoder.

5. The method of claim 1 , wherein the feature data of the current residual optical flow is obtained by performing entropy decoding and inverse quantization on the bitstream.

6. The method of claim 1 , wherein the neural-network-based first decoder is trained based on:

first loss information corresponding to a difference between a current training image and a current reconstructed training image related to the current training image; and

second loss information corresponding to entropy of the feature data of the current residual optical flow of the current training image.

7. The method of claim 1 , wherein the feature data of the current residual optical flow is obtained from the bitstream based on the current image corresponding to a predictive (P) frame, and based on the P frame following another P frame.

8. The method of claim 7 , wherein based on the P frame following an intra (I) frame, the method further comprises:

obtaining feature data of the current optical flow from the bitstream; and

reconstructing the current optical flow by applying the feature data of the current optical flow to a neural-network-based fourth decoder.

9. A non-transitory computer-readable recording medium having recorded thereon a program for executing the method of claim 1 .

10. An apparatus for reconstructing an optical flow by using artificial intelligence (AI), the apparatus comprising:

at least one processor configured to implement:

a bitstream obtainer configured to obtain feature data of a current residual optical flow from a bitstream for a current image; and

a prediction decoder configured to:

obtain the current residual optical flow by applying the feature data of the current residual optical flow to a neural-network-based first decoder,

obtain a second-order optical flow between a current predicted optical flow and a previous optical flow,

obtain the current predicted optical flow by modifying the previous optical flow according to the second-order optical flow, and

reconstruct a current optical flow based on the current residual optical flow and the current predicted optical flow.

11. A method of encoding an optical flow by using artificial intelligence (AI), the method comprising:

obtaining a second-order optical flow between a current predicted optical flow and a previous optical flow;

obtaining the current predicted optical flow by modifying the previous optical flow according to the second-order optical flow;

obtaining feature data of a current residual optical flow by applying a current image, a previous reconstructed image, and the current predicted optical flow to a neural-network-based first encoder; and

generating a bitstream corresponding to the feature data of the current residual optical flow,

wherein the current residual optical flow corresponds to a difference between a current optical flow and the current predicted optical flow.