Research

Here is an outline of the things I've been doing:

Recent Work:

Publications

• Siddiqui, M. Sclaroff, S. Surface Reconstruction from Multiple Views using Rational B-Splines and Knot Insertion, 1st International Symposium on 3D Data Processing Visualization and Transmission 2002(3DPVT02)

• Siddiqui, M. Sclaroff, S. Surface Reconstruction from Multiple Views using Rational B-Splines and Knot Insertion, BU-TechReport[2002-010] (This tech report is a longer version of the 3DPVT paper)

• Siddiqui, K. I. Hero, A. O. Siddiqui, M. M. Mathematical Morphology applied for Spot Segmentation and Quantification of Gene Microarray Images, to appear in Proc. of 36th IEEE Asilomar Conference on Signals, Systems and Computers, Pacific Grove, California (2002)[pdf]

• Kashif I. Siddiqui, Alfred O. Hero, Matheen M. Siddiqui, Segmentation and Quantification of Microarray Images, to appear in IEEE Int. Conference on Image Processing, Barcelona, Spain, Sept. 2003.

Masters Work @ BU

Masters Thesis: Surface Reconstruction from Multiple Views

Publications

• Siddiqui, M. Sclaroff, S. Surface Reconstruction from Multiple Views using Rational B-Splines, Technical Sketch, CVPR'01 [pdf]

• Siddiqui, M. Sclaroff, S. Surface Reconstruction from Multiple Views using Rational B-Splines, BU-TechReport[2001-009] (This report is a longer version of the CVPR paper and includes a discussion on the perspective case and more results)
  

• Rosales, R. Siddiqui, M. Alon, J. Sclaroff, S. Estimating Body Pose using Uncalibrated Cameras, In Proc. CVPR'01[pdf]

Fall 01 (Class Project)

HMMs for Online Handwriting Recognition (handwritingrec.zip, web)

UGrad Days:

Fall 00 (Undergraduate Thesis)

Calibration and Its Application to Stereo

This was my undergrad thesis.

Summer 99 (Internship@ MIT: Linclon Laboratories)

Linear Discriminant Analysis fo Tank Classification

In this project I used the familiar eigen-everything approach on tanks. It actually seemed to have a lot of potential as the seperability of the classes was larger using the fisher vector.

Spring 99 (IVC)

Delaunay Triangulation of Images

In this project a triangular mesh (Delaunay) is generated that partitions images into regions of near constant color. Here we start with an initial mesh that is hand placed on a region of interest. We then proceeded to add points to the triangulation such that the resulting triangulation partitions the image into regions with smaller average variance per triangle. Points are added to triangles such that the larger triangles are partitioned before the smaller ones and the mesh is updated according to an incremental Delaunay algorithm.

The code for this algorithm is here

Fall 98 (IVC - with a UROP package )

Enhancements to "Active Blobs"

This was my first dive into vision. I extended the system in two ways by adding a simple lighting model which parameterized brightness and contrast over the blob by a (spatially) linear (latter quadratic) function in the parameterization of the blob. That way if light shined with a gradient on the blob, the brightness and contrast overlays would try and compensate. This was an extension to work that compensated for overall brightness and contrast.