Watch this space for evaluation guidelines,
as well as how I usually decide on students on seminars and projects. Previous projects appear here.
(Also, you could look at the projects mentioned here.)

The following projects are for B.Tech students (work to be done in the academic year 2002-2003) in order of preference

  1. Fast Multipole Algorithms (FMM) for Graphics. FMM has been declared to be one of the top 10 influential algorithms in the 20th century. (Yes, the same century which saw the explosion of computers). The algorithm was invented in 1989 and won the best ACM Computer Science dissertation award. Many applications of the algorithm has been snapped up. We will be uncovering applications of this algorithm for radiosity.

    As a side note, my doctoral dissertation was also completed in 1989 :-)

  2. Video Indexing. A fast region relationship algorithm has been discovered for querying images by content. This seems to be competitive with current methods available to the general lay person (e.g. Google Image Search, Yahoo Image Search). A demo of this is available at http://www.cse.iitb.ernet.in/~graphics/naga/old/naga/demo.html

    Video gives us more challenges and that is the endeavor in this effort.

  3. The Metropolis Algorithm for Computer Graphics. This also figured in the top 10. (By now you might be wondering what else figures in the top 10. Here they are Metropolis Algorithm for Monte Carlo, Simplex Method for Linear Programming, Krylov Subspace Iteration Methods The Decompositional Approach to Matrix Computations, QR Algorithm for Computing Eigenvalues, The Fortran Optimizing Compiler, Quicksort Algorithm for Sorting, Fast Fourier Transform, Integer Relation Detection, Fast Multipole Method)

    It turns out that Monte Carlo is currently fancied, and is a competitor to the methods of FMM. This project is open.

  4. Automatic tracking in video images.
  5. Registering Terrains in Satellite Images:

    Driving a car with one eye is virtually impossible; with only one image of the surroundings, judging depth is difficult. With normal vision (both eyes), two images are formed (one with the left eye and another with the right eye - Stereo Vision) and you are now able to estimate how far objects are from you.

    One of the first steps a computer vision algorithm takes is to find the positions that correspond to the same point in the scene in the two images. This is the classic stereo correspondence problem. The depth of the scene point can now be calculated.

    In this project, you will work with a (stereo) pair of satellite images taken from different angles. For a level terrain (with little or no height variation), the positions of the same scene point in the two images are related by an affine transformation. You will work with the more challenging problem of registering UNDULATING TERRAINs.

  6. Octree based modeling and rendering. See my paper in ICVGIP for more details.
  7. Surface generation. Mixture of graphics and vision. Details in person.
Some other project ideas.
  1. P2: Web-Enabled Solid Modeling with Manufacturing Applications . Solid Modelers such as ParaSoft, BRep and others have become an integral part of the manufacturing process in virtually all industries. However, most solid modeling systems run on standalone workstations, which as a consequence has to be fairly powerful. With the increased influence of Internet, and the availability of world wide talent, it is becoming increasingly evident that a web-enabled solid modeler would be useful.

    In this project, you will use the ACIS kernel to create a client-server based solid modeler system. You will also create a prototype application tha will allow dynamic sharing of runtime design data.

    Related work: