Image Matching for Motion Analysis and Stereovision

This research is invested on image matching methods, and their applications. It is divided into two parts: methodological and engineering. The first tackles decision methods used to match features between images. We found it on the “stable marriage” paradigm. Several algorithms are developed with constraints added, as global satisfaction or equity, for better adapting to application needs as controlling a certain local/global balance in the decision. Then, we introduce a generic matching system based on these algorithms where stable marriages become the key mechanism. The second part is devoted to applicative processes building on this matching core. First, level lines and their junctions are selected for primitive features to be paired in all applications. Then, the adapted implementation of the core is studied in the following applications : a registration system, a system for obstacle detection from stereo pairs, and a generic system for otion analysis. The matching quality is experimented on and tested in all three applications. It is compared with vote based and dynamic programming based matching results.

 str1str2 str3
 cyc1cyc3 cyc2

In stereovision, the method is implemented on PiCar platform (embedded electrical car) for obstacle detection. The RT-maps is used for real time analysis.

mvt1 mvt2 mvt3
route1 route3 route2

In the case of motion analysis, we present a general system for motion understanding: features extraction (‘level sets’ to ‘junctions’), features matching (‘flow’ for two associated junctions), motion classification (‘flows’), image segmentation (‘junctions’ to ‘level sets’, object). In that case, some additional algorithms are proposed, like the fuzzy c-mean with spatial constraint on level sets algorithm, snake algorithm for level sets segmentation and etc.

Motion segmentation, test on human motion sequence and car sequence.

efThe EFLAM is an efficient approach for the detection of level-line junctions in images. Potential junctions are exhibited independent from noise by their consistent local level-variation. Then, level-lines are tracked through junctions in descending the level-line flow. Flow junctions are extracted as image primitives to support matching in many applications.The primitive is robust against contrast changes and noise. It is easily made rotation invariant. As far as the image content allows, the spread of junctions can be controlled for even spatial distribution. We show some results and compare with the Harris detector.eflam




  • N. Suvonvorn, F. Le Coat, B. Zavidovique, (2007), Marrying level-line junctions for obstacle detection, In Proceedings of IEEE International Conference on Image Processing (ICIP), September 16-19, 2007, San Antonio, Texas, USA, p.305-308. ISBN: 1-4244-1437-7..
  • N. Suvonvorn and B. Zavidovique, (2006), EFLAM : A model to level-line junction extraction, In Proceedings of International Conference on Computer Vision Theory and Applications (VISAPP), February 25 – 28, 2006, INSTICC, Setubal, Portugal, p. 257-264.
  • N. Suvonvorn, S. Bouchafa and B. Zavidovique, (2005), Marrying level lines for stereo or motion, In Proceedings of International Conference on Image Analysis and Recognition (ICIAR), September 28-30, 2005, Toronto, Canada, p. 391-398.
  • N. Suvonvorn, S. Bouchafa and L. Lacassagne, (2004), Fast Reliable Level-Lines Segments Extraction, In Proceedings of IEEE’s International Conference on Information and Communication Technologies: from Theory to Applications (ICTTA), April 19-23, 2004, Damascus, Syria, p. 349- 350.
  • N. Suvonvorn, 2004, Algorithm for level-lines extraction, In Proceedings of the First Academic Conference of Thai Students in France and Europe, organized by the Association of Thai Students in France under royal patronage of Thailand, June 14, 2004, Montpelier, France.

Related posts: