Journal of Beijing University of Posts and Telecommunications

  • EI核心期刊

Journal of Beijing University of Posts and Telecommunications ›› 2021, Vol. 44 ›› Issue (5): 101-106.doi: 10.13190/j.jbupt.2020-255

• REPORTS • Previous Articles     Next Articles

A Depth Estimation Method for Multi View and High Precision Images

LI Jian, CHEN Yu-hang   

  1. School of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2020-12-01 Online:2021-10-28 Published:2021-09-06

Abstract: High-precision images are challenging to reconstruct effectively in dense multi view reconstruction. To solve the problem, a learning-based depth estimation method is proposed. In the method,the dilated convolution neural network is used to extract image features,and the long short-term memory network is applied to construct and optimize the cost volume. Besides,the supervised and unsupervised training methods are adopted. Experimental results on two real scene multi view image datasets show that the proposed method not only outperforms state-of-the-arts methods,but also is several times less in GPU memory application compared with traditional methods and other learning-based methods. Therefore, the proposed method can reconstuct high-precision images, while improving the accuracy and integrity of model depth prediction.

Key words: multi view reconstruction, recursive neural network, high-precision images

CLC Number: