Research on Fully Automatic 2D to 3D Method Based on Deep Learning

Aiming at the problem that the traditional 2D to 3D method has excessive dependence on manual operation and complicated input, this paper proposed a fully automatic 2D to 3D method based on deep learning. First, a frame sequence is obtained according to the original 2D video, and a depth map corresponding to each frame image is obtained by using a convolutional neural network; then the left and right views are respectively drawn by using the original frame and the corresponding depth map according to the improved two-view DIBR algorithm. Finally, the improved Criminisi algorithm is used to fill the holes in the left and right views and to synthesize the final 3D video. The experimental results show that this method can not only obtain the ideal effect of 2D to 3D, but also the PSNR and SSIM value of the virtual viewpoint images are improved compared with other algorithms, and greatly improve the automation of 2D to 3D.