The examples shown here demonstrate the improvement in the disparity
maps comparing to the plain Confidently
Stable Matching algorithm (which is used for the final matching in
the Stratified Dense Matching algorithm). The
accuracy of the results remains comparable, while the density increases
about five-fold. The left column shows input left images of the
selected scenes, the middle comlumn results computed by Confidently
Stable Matching and the right column the results computed by Stratified
Dense Matching.
In all cased 5x5 matching window and normalized cross-correlation was used to compute the matching similarity. The confidence interval parameters were set equally for all the experiments and also both methods to alpha=20, beta=0.05.
Disparity is color-coded: large
disparities (foreground) are red, small disparities (background) are
dark blue. Holes are gray.
The Birch Image Pair
image size 484 x 640, disparity range 0 .. 55 (ordering
violated)
|
|
|
The Shrub Image Pair (shrub-3 and shrub-21)
image size 480 x 512, disparity range 10 .. 30
|
|
|
The Parking Meter Image Pair (pm-2 and pm-14)
image size 480 x 512, disparity range 0 .. 30
|
|
|
The Pentagon Image Pair
image size 512 x 512, disparity range -10 .. 10
|
|
|
The EPI Tree Image Pair (epi-32 and epi-16)
image size 233 x 256, disparity range 0 .. 35 (ordering
violated)
|
|
|
The Lab Scene (Tsukuba) Image Pair
image size 288 x 384, disparity range 4 .. 15
|
|
|
The Venus Image Pair
image size 380 x 434, disparity range 0 .. 19
|
|
|
The Sawtooth Image Pair
image size 380 x 434, disparity range 0 .. 19
|
|
|
The Map Image Pair
image size 216 x 284, disparity range 0 .. 60
|
|
|
Images courtesy of Carnegie Mellon University, SRI International,
University of Tsukuba and Middlebury College.
Jana Kostkova |
Last Modified: 30/01/2003
|