Single Image Depth Estimation Using Edge Extraction Network and Dark Channel Prior

The key to the depth estimation from a single image lies in inferring distance of various objects without copying texture while maintaining clear object boundaries. In this paper, we propose using edge extraction network and dark channel prior (DCP). We build an based on generative adversarial networks (GANs) select valid edges number image. use DCP generate transmission map that is able represent camera. Transmission generated by conducting minimum value filtering DCP. First, concatenate with original RGB form tensor, i.e. + T. Second, initial tensor through generator stacked residual blocks. Third, compare input edges. Both maps are network. Finally, distinguish real fake discriminator, enhances performance generator. Various experiments NYU, Make3D MPI Sintel datasets demonstrate proposed generates images as well outperforms state-of-the-art methods terms visual quality quantitative measurements.