Home › Research - current
Can you believe your eyes (or algorithms) ?

Suppose your algorithm suggests some interpretation of a give image. Can we know how reliable these suggestions is ? Sometimes, using statistics extract from the image itself, we can.

Some examples on segmentation and categorization are:

R. Sandler and M. Lindenbaum, Unsupervised Estimation of Segmentation Quality using Nonnegative Factorization. Proc. IEEE Conf. Computer Vision and Pattern Recognition - CVPR08, 2008.

D. Peles and M. Lindenbaum, A segmentation quality measure based on rich descriptors and classification methods. In SSVM, 2011.

A. Miaskouvskey, Y. Gousseau, and M. Lindenbaum, Beyond independence: An extension of the a contrario decision procedure. In Int. J. Computer Vision, 101(1), pages 22–44, 2013..

Detecting change in high dimensional space

Characterizing distribution in high dimensional space and finding changes in them over time is known to be hard. We proposed to characterize distributions using a hierarchical set of minimum volume sets which approximate the levels sets of it density (NIPS-2013). This representation allows the usage of a generalized, nonparametric, Kolmogorov-Smirnov test for detecting distributional change in high-dimensional data. The advantage of this approach over competing method is shown empirically (NIPS-2012). The distribution representation may be used for effective hierarchical clustering as well (NIPS-2014).

Assaf Glazer, Michael Lindenbaum and Shaul Markovitch, Learning High-Density Regions for a Generalized Kolmogorov-Smirnov Test in High-Dimensional Data. NIPS-2012, Lake Tahoe, Nevada, 2012.

Assaf Glazer, Michael Lindenbaum, and Shaul Markovitch, q-ocsvm: A q-quantile estimator for high dimensional distributions. NIPS-2013, 2013.

Assaf Glazer, Omer Weissbrod, Michael Lindenbaum, and Shaul Markovitch, Hierarchical mv-sets for hierarchical clustering. NIPS-2014, 2014.


In surveillance and other applications it is desirable that computer vision systems will be able to keep track of people after they have left the field of view of one camera and entered the field of view of the next, even when these fields of view do not overlap. We propose a novel approach for pedestrian re-identification, which differs from previous algorithm by being implicit and of more general scope. The implicit camera transfer algorithm (ICT) achieves new state-of-the-art performance. Interestingly, the ICT advantage is maintained even when directly annotated training sets are not available for all camera-pairs at the site, using the Transitive Re-IDentification algorithm (TRID). See more details here.

Tamar Avraham, Ilya Gurvich, Michael Lindenbaum, and Shaul Markovitch,Learning Implicit Transfer for Person Re-identification. 1st International Workshop on Re-Identification (Re-Id 2012) In conjunction with ECCV 2012, LNCS 7583, pp. 381–390.

Yulia Brand, Tamar Avraham, and Michael Lindenbaum, Transitive Re-identification. BMVC (British Machine Vision Conference) 2013.

Probabilistic Local Variation Segmentation

To be added