VIDEO FRAME INTERPOLATION METHOD AND APPARATUS, AND COMPUTER-READABLE STORAGE MEDIUM

摘要

The present disclosure relates to the technical field of information presentation, and particularly relates to a video frame interpolation method and apparatus, a computer-readable storage medium, and an electronic device. Said method comprises: obtaining, according to two input frames, two initial optical flow graphs corresponding to the two input frames; optimizing the initial optical flow graphs to obtain target optical flow graphs; obtaining a frame interpolation kernel, two depth graphs, and two context feature graphs according to the two input frames; and obtaining an output frame according to the target optical flow graphs, the depth graphs, the context feature graphs, and the frame interpolation kernel by using a frame synthesis method. The described steps at least meet one of the following conditions: performing iterative residual optical flow estimation on the two input frames, to obtain two initial optical flow graphs; optimizing, according to the two input frames, the initial optical flow graphs by using pixel adaptive convolution in combination with upsampling, to obtain target optical flow graphs; obtaining, according to the two input frames, two depth graphs by using a target depth estimation model; and obtaining an output frame according to the target optical flow graphs, the depth graphs, the context feature graphs, and the frame interpolation kernel by using a pixel adaptive convolution frame synthesis method.