Approches SIMD, SMP et MIMD-DM pour la stabilisation 2D d’images en temps réel
SIMD, SMP and MIMD-DM approaches for real-time 2D image stabilization
OPEN ACCESS
We present a real-time image stabilization method, based on a 2D motion model, and exploiting different levels of parallelism in its implementation. This stabilization method is decomposed into three parts. First, the image matching is determined by a feature-based technique. In the second part, the motion between consecutive frames is estimated and filtered to extract the unwanted motion component. Finally, these component is used to correct (warp) the images, resulting in a stable sequence. To validate our stabilization approach in a real-time on-board system context, the algorithm was implemented and tested over different hardware platforms, allowing a performance evaluation in function of the adopted architecture. In this paper, we present some results concerning the parallel implementation of the algorithm, using the SIMD ALTIVEC instructions set, a symmetric multi-processor architecture (SMP) and a MIMD-DM architecture.
Résumé
Nous présentons une méthode de stabilisation de séquence d'images en temps réel, basée sur un modèle de mouvement 2D, avec mise en correspondance par détection et suivi de primitives. Le déplacement estimé entre deux images est filtré afin d'isoler la composante non voulue du mouvement, et finalement utilisé pour corriger les images rendant la séquence stable. Afin de valider l'approche dans un contexte de système temps réel embarqué, l'algorithme a été implanté et testé sur plusieurs plateformes matérielles différentes, permettant l'évaluation des performances selon l'architecture adoptée. Nous montrons ici quelques résultats obtenus, concernant notamment la parallélisation de l'algorithme au moyen des instructions SIMD ALTIVEC, l'adoption d'une architecture symétrique multiprocesseur (SMP) et l'implantation sur une architecture de type MIMD-DM.
2D image stabilization, real-time application, SIMD instructions, SMP, MIMD-DM
Mots clés
Stabilisation 2D d'images, application temps réel, instructions SIMD, architecture SMP, architecture MIMD-DM
[1] Nicolas ALLEZARD, Lionel DAMEZ, and Jean-Pierre DÉRUTIN, Real time image stabilization by visual features matching. Rapport de recherche, LASMEA – Université Blaise Pascal, Mar 2003.
[2] S. BALARSKIRSKY and R. CHELLAPPA, Performance characterization of image stabilization algorithms. Rapport de recherche, Center for Automation Research – University of Maryland, 1996.
[3] J. BARRON, D. FLEET, S. BEAUCHEMIN, and T. BURKITT, Performance of optical flow techniques. International Journal of Computer Vision, 12(1):43-77, 1994.
[4] F. DIAS REAL DE OLIVEIRA, Analyse, développement et qualification d'un algorithme de stabilisation électronique d'images. Thèse de master, Université Blaise Pascal, 2004.
[5] Z. DURIC and A. ROSENFELD, Shooting a smooth video with a shaky camera. Machine Vision and Applications, 13(5-6):303-313, 2003.
[6] J. FALCOU and J. SEROT, E.v.e., an object oriented simd library. In International Conference and Computation Science – ICCS'2004, pages 323-330, 2004.
[7] C. HARRIS and M. STEPHENS. A combined corner and edge detector. In Proceeding of the 4th Alvey Vision Conference, pages 147-151, 1988.
[8] B. HORN and B. SCHUNCK, Determining optical flow. Artificial Intelligence, 17:185-204, 1981.
[9] C. MORIMOTO, Electronic Digital Stabilization: Design and Evaluation, with Applications. Thèse de doctorat, University of Maryland, 1997.
[10] H. POURREZA, M. RAHMATI, and F. BEHAZIN, Weighted multiple bit-plane matching, a simple and efficient matching criterion for electronic digital image stabilizer application. In 6th International Conference on Signal Processing, volume 2, pages 957-960, 2002.
[11] J. SEBOT, Impact des extensions simd sur les performances d'applications multimedia. Technique et Science Informatiques, 21(2):185-204, Mar 2002.
[12] D. TSAI, C. LIN, and J. CHEN, The evaluation of normalized cross correlations for defect detection. Pattern Recognition Letters, 24:2525-2535, 2003.
[13] A. VERRI and T. POGGIO, Motion field and optical flow: Qualitative properties. In IEEE Trans. Pattern Analysis and Machine Intelligence, volume 11(8), 1989.
[14] Paul VIOLA and Michael JONES, Rapid object detection using a boosted cascade of simple features. In Proceedings IEEE Conf. on Computer Vision and Pattern Recognition, 2001.
[15] Z. ZHU, G. XU, Y. YANG, and J. JIN, Camera stabilisation based on 2.5d motion estimation and inertial motion filtering. In International Conference on Intelligent Vehicles, 1998.