compfund

This program computes the fundamental matrix using a variety of algorithms, including 8 point linear (Hartley) and 7 point non-linear (Torr). It randomly chooses from a set of input correspondences and computes a fundamental matrix from these correspondences. It repeats this process a number of times and returns the best fundamental matrix; which is the one with the largest number of supporting correspondences. Note that the fundamental matrix is refit to the support correspondences produced by the random sampling process, but only if this decreases the average error (and it almost always does).

**USAGE:**

compfund -f matchesfile [-n numberOfRandomSamples] [-count] [-l2] [-h] [-a] [-w][-m maxscalechange] [-q maxdisparitygrad] [-o <force order constraint>][-d distance from epipolar line] [-i imagefile] [-s success probability]

**Parameters:**

**-f: **the correspondence match file
used to compute the fundamental matrix.

*This is a mandatory input.*

**-n: **the number of random samples
to perform.

**-count:** use count scoring (score
by counting the number of supporting points).

**-l2:** use l2 scoring (score
supporting points based on their distance from the epipolar line).

**-h: **compute fundamental matrix
using 8 point Hartley algorithm.

**-a: **compute an affine fundamental
matrix.

**-w: **compute a planar warp (not
actually a fundamental matrix).

**-m: **maximum scale change allowed
for a point set when random sampling

**-q: **maximum disparity gradient
allowed for a point set when random

**-o: **enforce an order constraint
on the random point sets

**-d: **maximum allowable distance
from an epipolar line for a point to be considered an inlier

**-i: **the name of an image file for
saving the correspondence vectors

**-s: **success probability, stop
sampling when this probability of success has been reached

**Default Values:**

1000 samples

count style scoring

1.0 pixels from epipolar line

NO order constraint

max scale change: 1000

max disparity gradiant: 1000

Success probability: 99.9999 percent

Default algorithm: Non-linear with 7 samples

**Output:**

A file with the same name as the input file is produced with the extension FM. This holds the nine values of the fundamental matrix. E.g. X.matches.filtered à X.matches.filtered.FM