Piotr Mirowski

Computer Vision, Spring 2006

Homework 3 - Stereo Vision



Matlab code:

HW3.m

LineGridIntersection.m

IntensityAccumulation.m

IntensityDerivative.m

CyclopeanMatch.m

EpipolarInteractionProbability.m

AdjacentCyclopeanProbability.m

DisplayImage.m

DisparityImage.m

WriteImage.m

BP_TwoGraphs.c

BP_TwoGraphs.h


1 - Image processing results on the Random Dots image


Best result (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

lambda=1.000
mu=1.000
tilt=1.000
occlusion=1.000

Most likely disparity (based on phi only)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

Most likely disparity (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

lambda=1.000
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.100
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.010
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.001
mu=1.000
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.100
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.010
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.001
tilt=1.000
occlusion=1.000

lambda=1.000
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.100
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.010
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.001
mu=1.000
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.100
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.010
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.001
tilt=0.100
occlusion=0.100

lambda=1.000
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.100
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.010
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.001
mu=1.000
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.100
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.010
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.001
tilt=0.010
occlusion=0.010

2 - Image processing results on the Square image


Best result (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 11

lambda=0.001
mu=0.001
tilt=0.010
occlusion=0.010

Most likely disparity (based on phi only)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

Most likely disparity (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

lambda=1.000
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.100
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.010
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.001
mu=1.000
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.100
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.010
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.001
tilt=1.000
occlusion=1.000

lambda=1.000
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.100
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.010
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.001
mu=1.000
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.100
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.010
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.001
tilt=0.100
occlusion=0.100

lambda=1.000
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.100
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.010
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.001
mu=1.000
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.100
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.010
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.001
tilt=0.010
occlusion=0.010

3 - Image processing results on the Corridor image


Best result (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 11

lambda=1.000
mu=0.001
tilt=0.010
occlusion=0.010

Most likely disparity (based on phi only)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

Most likely disparity (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

lambda=1.000
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.100
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.010
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.001
mu=1.000
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.100
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.010
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.001
tilt=1.000
occlusion=1.000

lambda=1.000
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.100
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.010
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.001
mu=1.000
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.100
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.010
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.001
tilt=0.100
occlusion=0.100

lambda=1.000
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.100
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.010
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.001
mu=1.000
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.100
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.010
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.001
tilt=0.010
occlusion=0.010

4 - Image processing results on the Pentagon image


Best result (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

lambda=1.000
mu=0.001
tilt=0.010
occlusion=0.010

3D reconstruction of the Pentagon
using the best disparity results above as depth

A simplistic model:
Left and right image averaged
even and odd columns of the disparity image averaged

Most likely disparity (based on phi only)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

Most likely disparity (using Belief Propagation)

White = no disparity, red = negative disparity, blue = positive disparity

Scale 3

Scale 5

Scale 11

lambda=1.000
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.100
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.010
mu=1.000
tilt=1.000
occlusion=1.000

lambda=0.001
mu=1.000
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.100
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.100
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.010
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.010
tilt=1.000
occlusion=1.000

lambda=1.000
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.100
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.010
mu=0.001
tilt=1.000
occlusion=1.000

lambda=0.001
mu=0.001
tilt=1.000
occlusion=1.000

lambda=1.000
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.100
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.010
mu=1.000
tilt=0.100
occlusion=0.100

lambda=0.001
mu=1.000
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.100
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.100
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.010
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.010
tilt=0.100
occlusion=0.100

lambda=1.000
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.100
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.010
mu=0.001
tilt=0.100
occlusion=0.100

lambda=0.001
mu=0.001
tilt=0.100
occlusion=0.100

lambda=1.000
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.100
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.010
mu=1.000
tilt=0.010
occlusion=0.010

lambda=0.001
mu=1.000
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.100
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.100
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.010
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.010
tilt=0.010
occlusion=0.010

lambda=1.000
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.100
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.010
mu=0.001
tilt=0.010
occlusion=0.010

lambda=0.001
mu=0.001
tilt=0.010
occlusion=0.010