<s>
In	O
computer	B-Application
vision	I-Application
,	O
the	O
Birchfield	B-General_Concept
–	I-General_Concept
Tomasi	I-General_Concept
dissimilarity	I-General_Concept
is	O
a	O
pixelwise	O
image	O
dissimilarity	O
measure	O
that	O
is	O
robust	O
with	O
respect	O
to	O
sampling	B-Algorithm
effects	O
.	O
</s>
<s>
In	O
the	O
comparison	O
of	O
two	O
image	O
elements	O
,	O
it	O
fits	O
the	O
intensity	O
of	O
one	O
pixel	B-Algorithm
to	O
the	O
linearly	O
interpolated	O
intensity	O
around	O
a	O
corresponding	O
pixel	B-Algorithm
on	O
the	O
other	O
image	O
.	O
</s>
<s>
It	O
is	O
used	O
as	O
a	O
dissimilarity	O
measure	O
in	O
stereo	O
matching	O
,	O
where	O
one-dimensional	O
search	O
for	O
correspondences	O
is	O
performed	O
to	O
recover	O
a	O
dense	O
disparity	O
map	O
from	O
a	O
stereo	B-Algorithm
image	I-Algorithm
pair	I-Algorithm
.	O
</s>
<s>
When	O
performing	O
pixelwise	O
image	B-Algorithm
matching	I-Algorithm
,	O
the	O
measure	O
of	O
dissimilarity	O
between	O
pairs	O
of	O
pixels	B-Algorithm
from	O
different	O
images	O
is	O
affected	O
by	O
differences	O
in	O
image	O
acquisition	O
such	O
as	O
illumination	O
bias	O
and	O
noise	B-Algorithm
.	O
</s>
<s>
Even	O
when	O
assuming	O
no	O
difference	O
in	O
these	O
aspects	O
between	O
an	O
image	O
pair	O
,	O
additional	O
inconsistencies	O
are	O
introduced	O
by	O
the	O
pixel	B-Algorithm
sampling	B-Algorithm
process	O
,	O
because	O
each	O
pixel	B-Algorithm
is	O
a	O
sample	O
obtained	O
integrating	O
the	O
continuous	O
light	O
signal	O
over	O
a	O
finite	O
region	O
of	O
space	O
,	O
and	O
two	O
pixels	B-Algorithm
matching	O
the	O
same	O
feature	O
of	O
the	O
image	O
content	O
may	O
correspond	O
to	O
slightly	O
different	O
regions	O
of	O
the	O
real	O
object	O
that	O
can	O
reflect	O
light	O
differently	O
and	O
can	O
be	O
subject	O
to	O
partial	O
occlusion	O
,	O
depth	O
discontinuity	O
,	O
or	O
different	O
lens	O
defocus	O
,	O
thus	O
generating	O
different	O
intensity	O
signals	O
.	O
</s>
<s>
The	O
Birchfield	O
–	O
Tomasi	O
measure	O
compensates	O
for	O
the	O
sampling	B-Algorithm
effect	O
by	O
considering	O
the	O
linear	O
interpolation	O
of	O
the	O
samples	O
.	O
</s>
<s>
Pixel	B-Algorithm
similarity	O
is	O
then	O
determined	O
by	O
finding	O
the	O
best	O
match	O
between	O
the	O
intensity	O
of	O
a	O
pixel	B-Algorithm
sample	O
in	O
one	O
image	O
and	O
the	O
interpolated	O
function	O
in	O
an	O
interval	O
around	O
a	O
location	O
in	O
the	O
other	O
image	O
.	O
</s>
<s>
Given	O
that	O
the	O
interpolant	O
is	O
linear	O
within	O
each	O
unit	O
interval	O
centred	O
around	O
a	O
pixel	B-Algorithm
,	O
its	O
minimum	O
is	O
located	O
in	O
one	O
of	O
its	O
extremities	O
.	O
</s>
