Stella X. Yu : Research

The software is made publicly available for the purpose of advancing research in computer vision and related machine learning areas. It may be modified and redistributed under the terms of the

GNU General Public License.

Download now

User program

example.m : an example on ncuts for image segmentation

Graph cuts

imncut.m : master program on ncuts for image segmentation

cncut.m : constrained ncuts with attraction, repulsion and regularization

getbinsol.m : get a discrete partitioning from eigensolutions

pargrp.m : generate a constraint matrix from partial grouping cues

barqpqz.c : projection formula used in cncut

Image feature extraction

ic.c : compute intervening contours

quadeg.m : edge extractions through quadrature filters

make_filterbank_even2.m : make even-phase filters

make_filterbank_odd2.m : make odd-phase filters

doog1.m : 1D difference of Gaussian

doog2.m : 2D difference of Gaussian

fft_filt_2.m : filtering by FFT

gaussian.m : Gaussian function

Sampling and speed up

imnb.c : pixel pairs in a local window

csparse.c : generate a sparse matrix from C-index representation

spmd.c : sparse matrix times diagonal matrices

parmatV.c : binary partition matrix times a real matrix

Image

135069.jpg : from Berkeley image dataset, 160 x 240

Notes

These programs incorporate new developments on spectral graph partitioning as described in the following thesis:

Computational Models of Perceptual Organization

Stella X. Yu, May 2003, Carnegie Mellon University

except Chapter 4 where Hermitian matrices are used.

However, they can be extended by modifying related c functions to deal with imaginary number multiplications.