


Calculate optical flow using cross-correlation.
Calculate optical flow using correlation, followed by lucas & kanade on
aligned squares for subpixel accuracy. Locally, the closest patch within
some search radius is found. The distance measure used is the euclidean
distance between patches -- NOT normalized correlation since we assume
pixel brightness constancy. Once the closest matching patch is found,
the alignment between the two patches is further refined using lucas &
kanade to find the subpixel translation vector relating the two patches.
This code has been refined for speed, but since it is nonvectorized code
it can be fairly slow. Running time is linear in the number of pixels
but the constant is fairly large. Test on small image (150x150) before
running on anything bigger.
USAGE
[Vx,Vy,reliab] = optFlowCorr( I1, I2, patchR, searchR,
[sigma], [thr], [show] )
INPUTS
I1, I2 - input images to calculate flow between
patchR - determines correlation patch size around each pixel
searchR - search radius for corresponding patch
sigma - [1] amount to smooth by (may be 0)
thr - [.001] RELATIVE reliability threshold
show - [0] figure to use for display (no display if == 0)
OUTPUTS
Vx, Vy - x,y components of flow [Vx>0->right, Vy>0->down]
reliab - reliability of optical flow in given window (cornerness of
window)
EXAMPLE
See also OPTFLOWHORN, OPTFLOWLK
Piotr's Image&Video Toolbox Version 2.0
Copyright 2008 Piotr Dollar. [pdollar-at-caltech.edu]
Please email me if you find bugs, or have suggestions or questions!
Licensed under the Lesser GPL [see external/lgpl.txt]