Hazem El-Alfy            -                  Research Page


Multi-Scale Video Cropping


We consider the problem of “cropping” surveillance videos. This process chooses a trajectory that a small sub-window can take through the video, selecting the most important parts of the video for display on a smaller monitor. The result is a meaningful video with a lower resolution that can fit on smaller displays and save bandwidth. The globally optimal trajectory for a cropping window is found by using a shortest path algorithm.  The method is applied on real surveillance videos.


Hazem El-Alfy, David Jacobs and Larry Davis, “Multi-Scale Video Cropping,” ACM Multimedia 07, Sep. 2007

Paper: pdf (446 KB)

Presentation: ppt (3.95 MB) [20% acceptance: 18 out of 90 submissions to the Applications Track.]

Videos: (download with presentation in same folder. Do NOT modify names!) video1 (1.7MB), video2 (2.4MB), video3 (6.9MB), video4 (5.2MB), video5 (4.1MB), video6 (1.6MB), video7 (4.6MB), video8 (4.3MB), video9 (4.2MB).

Additional videos (not on presentation): [AVI format – DivX codec compression] video10 (14.8MB)

Matlab code: zip (19.4KB)


Assigning Cameras to Subjects in Surveillance Systems

Given an environment with obstacles, and many people moving through it, we construct a separate video for each person, by stitching together video segments from multiple cameras over time. We employ a novel approach, using bipartite matching, to assign a camera to each person as a function of time, with camera switches when needed. When the number of people is large, we cluster as many people as possible into small groups, then assign cameras to groups using a minimum cost matching algorithm. The method is tested using numerous runs from different simulators.


Hazem El-Alfy, David Jacobs and Larry Davis, “Assigning Cameras to Subjects in Video Surveillance Systems,” IEEE ICRA ’09, May 2009.

Paper: pdf (338 KB).

Presentation: zip (1.5 MB).


Stochastic Multiple Scattering

We solve the problem of wave scattering by multiple spheres, subject to uncertain boundary conditions. Uncertainty is modeled through a Karhunen-Loève expansion of the right hand side. Useful properties of spheres are exploited, by discretizing the problem in a basis of spherical harmonics, and speed-up is achieved through multipole reexpansion.


Supervised by: Howard Elman and Ramani Duraiswami.

Hazem El-Alfy, “Multiple Scattering from N Spheres with Uncertain Source Location Using Stochastic Multipoles,” Scholarly Paper, M.Sc., Computer Science, Univ. of Maryland, College Park, USA.

Paper: pdf

Matlab code: zip


Implicitization Problem


f(x, y) = 2x4 - 3x2y + y4 - 2y3 + y2 = 0

Implicitization is the process of converting equations of curves and surfaces from parametric form into implicit form. We implement algorithms for currently available methods. In addition, we devise a new method for problems for which no direct method is available. The method relies on producing an approximation of the input problem. Several variants of this new method try to offer a compromise between its accuracy and versatility.


Supervisors: Abdel-Karim Aboul-Hassan and Mohamed Sayed.

Hazem El-Alfy, “Computer Algebra and its Applications,” M. Sc. Thesis, Engineering Mathematics Department, Faculty of Engineering, Alexandria University, Egypt.

Thesis: pdf