Color-Guided Depth Map Super Resolution Using Convolutional Neural Network.

Min Ni,Jianjun Lei,Runmin Cong,Kaifu Zheng,Bo Peng,Xiaoting Fan
DOI: https://doi.org/10.1109/access.2017.2773141
IF: 3.9
2017-01-01
IEEE Access
Abstract:With the development of 3-D applications, such as 3-D reconstruction and object recognition, accurate and high-quality depth map is urgently required. Recently, depth cameras have been affordable and widely used in daily life. However, the captured depth map always owns low resolution and poor quality, which limits its practical application. This paper proposes a color-guided depth map super resolution method using convolutional neural network. First, a dual-stream convolutional neural network, which integrates the color and depth information simultaneously, is proposed for depth map super resolution. Then, the optimized edge map generated by the high resolution color image and low resolution depth map is used as additional information to refine the object boundary in the depth map. Experimental results demonstrate the effectiveness of the proposed method compared with the state-of-the-art methods.
What problem does this paper attempt to address?