Iasonas Kokkinos

Post-Doctoral Researcher

Center for Image and Vision Sciences, Department of Statistics

University of Californa at Los Angeles

Email:       jkokkin[at]stat.ucla.edu

Address:   8145 Math. Sc. Bldg., CA 90095

Biosketch

I received the Diploma and the Ph.D. degrees in Electrical and Computer Engineering in 2001 and 2006 from NTUA, working in the Computer Vision and Signal Processing group. During part of 2004 I stayed with the Odyssee group, at INRIA Sophia Antipolis, and as of June 2006 I am with the Center for Image and Vision Sciences, at UCLA. (full CV)

Research and Selected Publications (complete publication list)

My research is on probabilstic approaches to vision. I am particularly interested in combining low- and high- level vision problems in a single probabilistic framework, as well as the individual low- and high- level problems.

Combination of Bottom-Up and Top-Down Processes for Computer Vision

There is a general consensus that the low- and high-level problems of vision are interrelated: low-level tasks are ill-posed without prior, high-level knowledge, while high-level models need bottom-up cues to activate them and drive their fitting. In my work I have been developping tools for detecting and segmenting deformable object models in this framework. (read more)

• I. Kokkinos and A. Yuille, Scale Invariance without Scale Selection , in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2008. [pdf]

  I. Kokkinos and A. Yuille, Unsupervised Learning of Object Deformation Models., in Proc. IEEE Int'l. Conf. on Computer Vision (ICCV), 2007 [pdf] [slides]

• I. Kokkinos, P. Maragos and A. Yuile, Bottom-Up and Top-Down Object Detection Using Primal Sketch Features and Graphical Models, in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2006. [pdf] [slides] [code]

• I. Kokkinos and P. Maragos, Synergy Between Image Segmentation and Object Recognition Using the Expectation Maximization Algorithm, in IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), to appear. [pdf]

• I. Kokkinos and P. Maragos, An Expectation Maximization Approach to the Synergy Between Image Segmentation and Object Categorization. Proc. IEEE Int'l. Conf. on Computer Vision (ICCV), 2005. [pdf]

Texture Analysis and Segmentation

I have worked on probabilistically interpreting image processing operations, like Gabor filtering or edge detection. This facilitates the association of likelihood values to filtering results, which in turn allows to adaptively weigh the influence of different cues in segmentation algorithms (read more).

• I. Kokkinos, G. Evangelopoulos and P. Maragos, Texture Analysis and Segmentation using Modulation Features, Generative Models and Weighted Curve Evolution, IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), to appear.[pdf] [slides]

Learning a Biologically Motivated Model of Low-Level Vision

During my stay at the Odyssee group at INRIA I worked on analyzing variationally and statistically a biologically motivated model of boundary detection. Apart from a better understanding of its many interacting modules, this led also to a learning algorithm for the network weights. (read more).

• I. Kokkinos, R. Deriche, O. Faugeras and P. Maragos, Computational Analysis and Learning for a Biologically Motivated Model of Boundary Detection, Neurocomputing, to appear. [pdf] [slides]

Speech Modelling using Models for Chaotic Systems

For my diploma degree I worked on modelling the nonlinear dynamical aspects of speech. As in chaotic time series analysis, the one-dimensional speech signal is embedded in a multi-dimensional space, on which nonlinear models are estimated using machine learning techniques. Geometric invariants, like Lyapunov exponents are then extracted and used as features for recognition. (read more).

• I. Kokkinos and P. Maragos, Nonlinear Speech Analysis Using Models for Chaotic Systems,IEEE Trans. on Speech and Audio Processing, Vol. 13(6), pp. 1098-1109, 2005. [pdf].