Improvement of Image Matching by using the Proximity Criterion: Application to Omnidirectional and Perspective Images | Science Publications

Journal of Computer Science

Ahmed Hammouch¹  Ibrahim Guelzim¹  Driss Aboutajdine¹ 

关键词: Images matching;    3D reconstruction;    omnidirectional vision;    proximity criterion;   

DOI  :  10.3844/jcssp.2011.1230.1236

学科分类：计算机科学（综合）

来源: Science Publications

【 摘 要 】

Problem statement: In computer vision, matching is an important phase for severalapplications (object reconstruction, robot navigation ...). The similarity measures used providedresults which could be improved. Approach: This research proposed to improve image matchingby using the proximity criterion. The similarity measures used mutual information and correlationcoefficient. The matching was done between neighborhoods of points of interest extracted from theimages. The second chance algorithm was also applied. We have worked in case which thesensor had a slight displacement between two images. The tests were performed onomnidirectional and perspective grayscale images. Results: The improvement by introducingthe proximity criterion reached 15.9% for non-noised perspective images, 32.1% for noisedperspective images, 47.69% for non-noised omnidirectional images and 58.5% for noisedomnidirectional images. Conclusion/Recommendations: The introduction of the proximitycriterion has significantly improved the performance of the matching. The method isrecommended in mobile robotics, knowing that a good matching leads to a better location andbetter movement of the robot.

【 授权许可】

Unknown