earticle

논문검색

Adaptive Weight Stereo Matching Algorithm based on Neural Network and Improved Mean-Shift

초록

영어

In order to increase the effect of matching for local stereo matching method and decrease the amount of computation, a new adaptive weight based local stereo matching method is proposed in this paper. In this method, two methods are mainly employed to construct weight model: (1) Neural Network is used to establish the spatial weight model, which makes good use of the pixels in support window; (2) An edge hold off Mean-Shift method is proposed to distribute the intensity weight accurately. For decreasing the matching cost error, the census transform is introduced to calculate the matching cost. The influence of the parameters on the performance of our method is also discussed at last. Simulation results indicate that the performance of our method is better than that of Yoon’s method under low support window.

목차

Abstract
 1. Introduction
 2. Improved Cost Aggregation Function
  2.1. Spatial Weight Algorithm Based on Neural Network
  2.2. Intensity Weight Function based on Edge Hold off Mean-Shift Algorithm
  2.3. Aggregation Cost
 3. LBP Algorithm
 4. Simulation Experiment and Discussions
  4.1. Effect of Parameter a on Performances of Our Algorithm
  4.2. Influence of Segmentation Algorithm in Performance
  4.3. Comparison with Yoon Algorithm
  4.4. LBP Algorithm Energy Change
  4.5. Performance Evaluation
 5. Conclusions
 Acknowledgements
 References

저자정보

  • Zheng Sun School of Mechatronic Engineering, Zaozhuang University, Zaozhuang 277160
  • Rong Yang Department of Mechanical Engineering, McMaster University, ON L8S 1A1
  • Hui Li School of Mechatronic Engineering, China University of Mining and Technology, Xuzhou 221116

참고문헌

자료제공 : 네이버학술정보

    함께 이용한 논문

      ※ 원문제공기관과의 협약기간이 종료되어 열람이 제한될 수 있습니다.

      0개의 논문이 장바구니에 담겼습니다.