Doctoral Student Presents Paper at the IEEE International Conference on Image Processing 2011
The paper “Fast Incremental Method for Matrix Completion: an Application to Trajectory Correction” was selected to be presented at the IEEE International Conference on Image Processing 2011. The paper was written by Ricardo Cabral, a Carnegie Mellon Portugal dual degree doctoral student in Electrical and Computer Engineering (ECE), with his two advisors João Paulo Costeira, from the Instituto Superior Técnico da Universidade Técnica de Lisboa (IST/UTL), and Fernando De la Torre, from Carnegie Mellon, with Alexandre Bernardino, from the IST/UTL. The Conference will be held on September 11-14, 2011, in Brussels.
In this paper the authors address the problem of incrementally recovering a matrix of tracked image points, based on partial observations of their trajectories. Besides partial observability, the authors “assume the existence of gross, but sparse, noise on the known entries,” explaining that “this problem has obvious applications in real-time tracking and structure from motion, where observations are plagued by self-occlusion and outliers.” Recently, research work in the optimization community has spun optimal methods for matrix completion when this matrix is known to be low rank by minimizing the nuclear norm, the sum of its singular values. Despite exhibiting several optimality properties, no available algorithms perform this minimization incrementally.
“Build upon the Nuclear Norm Robust PCA (Principal Component Analysis) method and SPectrally Optimal Completion to propose a fast and incremental algorithm which is able to cope with outliers,” the authors present “experiments showing the competitive speed of our method while maintaining performance comparable to the state-of-the-art.”
The International Conference on Image Processing, organized since 1994, is the first forum for the presentation of technological advances and research results in the fields of theoretical, experimental, applied image and video processing. This event will gather leading engineers and scientists in image processing from around the world.
Paper available at http://humansensing.cs.cmu.edu/projects/fimmc/fimmc.pdf