Michal Irani

Professor.

 

Dept. of  Computer Science and Applied Math
(Ziskind building, room 224)
The Weizmann Institute of Science
Rehovot, 76100 ISRAEL

E-mail:  michal.irani weizmann.ac.il
Phone: +972-8-9344297

 
                                        

My research area is computer vision and video information analysis. In the past few years it has focused on two main themes: 

1. Space-Time Analysis of Video:

Although space and time are very different in nature, they are closely interrelated. This leads to inherent visual trade-offs between time and space. This is also what makes video much more than just a plain collection of images of the same scene taken from different view-points. Therefore, in contrast to the traditional way of analyzing video on a frame-by-frame basis, my work over the past few years has focused on analyzing simultaneously all available data in entire space-time volumes. I have shown that such a space-time approach is essential in order to perform tasks that are very difficult and often impossible to perform otherwise, when only “slices” of this information are used (such as discrete image frames or discrete feature points). The space-time approach gives rise to new powerful ways of analyzing and exploiting recorded visual information from single and multiple visual sources.

My research work in this area is aimed toward:   (a) developing theories and tools for analysis and interpretation of space-time visual information,   (b) developing methods to exploit this rich visual information for useful real-world applications, and   (c) develop new and improved visual capabilities that exceed optical bounds of today’s visual sensors (including the human eye).

 

2. Visual Inference by Composition:

I have recently begun developing an "Inference-by-Composition" approach, which gives rise to analysis and prediction of complex visual information (both in images and in video data) without resorting to any pre-defined parametric models, nor requiring an exhaustive set of prior visual examples (which are often relied upon). Visual "pieces of evidence" from a small number of available visual examples are composed and integrated into new global visual configurations that were never seen before. This allows to make inferences about the likelihood of a combinatorially large set of complex scenes and events that were never observed. This "Inference by Composition" approach opens the door to analysis and prediction of very complex static and dynamic information, which could not have been previously handled. Moreover, the applicability of this approach extends beyond the field of Computer Vision to multiple disciplines and research areas. I am currently working towards developing it into a general approach to the representation and analysis of visual as well as non-visual digital data.

 

More details can be found in my papers below and in their corresponding demo webpages.

To learn more about the Vision Lab in the Weizmann Institute (people, projects, demos) --  click the following link.

Selected Publications:
 
 

  • L. Zelnik-Manor,  M. Machline,  and   M. Irani,  Multi-body Factorization With Uncertainty: Revisiting Motion Consistency.   International Journal of Computer Vision (IJCV),   68(1): 27-41  (special issue on Vision and Modeling of Dynamics Scenes),    June 2006. 
    (A shorter version of this paper appeared in VMODS Workshop -- Vision and Modelling of Dynamic Scenes -- June 2002.)

  • Y. Caspi,  D. Simakov,  and   M. Irani,  Feature-Based Sequence-to-Sequence Matching.   International Journal of Computer Vision (IJCV),   68(1): 53-64  (special issue on Vision and Modeling of Dynamics Scenes),    June 2006.    *See webpage with example sequences and results.
    (A shorter version of this paper appeared in VMODS Workshop -- Vision and Modelling of Dynamic Scenes -- June 2002.)

  • L. Zelnik-Manor  and  M. Irani,  Statistical Analysis of Dynamic Actions. IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), 28(9): 1530--1535, September 2006. 
    (A preliminary version appeared in CVPR’2001: “Event-Based video Analysis”).
  • Y. Caspi  and  M. Irani,  Aligning Non-Overlapping Sequences.    International Journal of Computer Vision (IJCV), Vol. 48, No. 1, pp. 39-51, 2002. 
    *See webpage with example sequences and results. 
    (A shorter version appeared in ICCV'2001). 
    * Received the Honorable Mention for the 2001 Marr Prize.

  •  Y. Caspi  and  M. Irani,  Spatio-Temporal Alignment of Sequences.   IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), Vol. 24, No. 11, pp. 1409-1424, November 2002.   
    (A shorter version appeared in CVPR'2000: "A step Towards Sequence-to-Sequence Alignment".)
    *See webpage with     example sequences and results .


  • M. Irani and  P. Anandan,  Factorization with Uncertainty . European Conference on Computer Vision (ECCV), June 2000.  
    * Received the best-paper prize at ECCV'2000.
    Longer journal version: 
    P. Anandan and M. Irani,      Factorization with Uncertainty .  International Journal of Computer Vision (IJCV),   49(2-3): 101-116,    September   2002. 

  • L. Zelnik-Manor  and  M. Irani,  Event-Based Video Analysis . IEEE Conference on Computer Vision and  Pattern Recognition (CVPR), December 2001.

  • M. Irani and P. Anandan,  About Direct Methods.   ICCV workshop on Vision Algorithms, pp. 267-277, Corfu, September 1999.
  • L. Zelnik-Manor  and  M. Irani,  Multi-View Constraints on Homographies. IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), Vol. 24, No. 2, pp. 214 223, February 2002.
    (A shorter version appeared in ICCV'1999.)
  • M. Irani, P. Anandan, and Meir Cohen, Direct Recovery of Planar-Parallax from Multiple Frames . IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), Vol. 24, No. 11, pp. 1528 1534, November 2002.
    (A shorter version appeared in ICCV'99 Workshop: Vision Algorithms 99, Corfu, September 1999.)
  • L. Zelnik-Manor  and  M. Irani,  Multi-Frame Estimation of Planar Motion .  IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), Vol. 22, No. 10, pp. 1105-1116, October 2000.
    (A shorter version appeared  in CVPR'99: "Multi-frame alignment of planes").