The presented image sequence is in a two-layers transparency configuration implying a static background and a set
"heart and lungs" moving coherently. Surgical clips are moving with the lungs.
The estimated displacements are directed toward the past to respect the notations of the Transparent Motion Constraint.
As a result they are in the opposite direction compared to the usual represented velocity direction.
Estimation result at four interesting time instants
The anatomical motions are correctly estimated, but this example is particularly interesting for what happens at the
fourth presented time instant.
The last presented example shows the dependancy of the method to the global observability problem. Since we consider the whole image, we generally
have a gradient distribution varied enough to have motion information in every direction.
In this example though, the textures are too little contrasted to bring motion information perpendicular to the contrasted
limit of the heart.
Whole processed sequence
We also present the overall processed sequence. It shows the estimated affine motion fields, along with the corresponding
frames.
The results are satisfactory for about 70% of time instants, though little texture is present in the image.
|
|