The invention relates to the technical field of video dimension conversion, and provides a 2D-to-3D conversion method for a video, and the method comprises the steps: S1, collecting an open-source RGB-D image data set, carrying out the expansion, forming a depth estimation data set, constructing a depth estimation model through the depth estimation data set, and training the depth estimation model; S2, collecting 4K high-definition pictures to make an image restoration data set, expanding the image restoration data set, constructing an image restoration model through the image restoration dataset, and training the image restoration model; and S3, extracting an original image mask by using the pre-trained Mask-RCNN model, adjusting the resolution of the original image and the mask, sendingthe original image and the mask to a depth estimation model, calculating original left and right projection images according to the depth image, and sending the left and right projection images to animage restoration model to restore a black hole area. A deep learning algorithm and a traditional algorithm are combined, a deep learning model is used for replacing a depth map estimation algorithmand a black hole filling algorithm in a traditional DIBR method, and 2D / 3D conversion on an ultrahigh-resolution image is achieved.