Implicit Free-Form-Deformations for Multi-frame Segmentation and Tracking

Abstract

Abstract. In this paper, we propose a novel technique to address motion estimation and tracking. Such technique represents the motion field using a regular grid of thin-plate splines, and the moving objects using an implicit function on the image plane that is a cubic interpolation of a "level set function" defined on this grid. Optical flow is determined through the deformation of the grid and consequently of the underlying image structures towards satisfying the constant brightness constraint. Tracking is performed in similar fashion through the consistent recovery in the temporal domain of the zero iso-surfaces of a level set that is the projection of the Free Form Deformation (FFD) implicit function according to the cubic spline formulation. Such an approach is a compromise between dense motion estimation and parametric motion models, introduces smoothness in an implicit fashion, is intrinsic, and can cope with important object deformations. Promising results demonstrate the potentials of our approach.

Multitemporal Very High Resolution From Space: Outcome of the 2016 IEEE GRSS Data Fusion Contest

“…A number of studies were focused on jointly tackling the problem of registration and semantic segmentation for mainly video sequences or medical images [48]- [51]. Similar research efforts were focused on jointly addressing the tasks of segmentation and tracking in image video sequences [52]- [54]. However, such formulations cannot exploit sparse multitemporal datasets with changes in-between the various acquisition dates.…”

Integrated Segmentation and Non-linear Registration for Organ Segmentation and Motion Field Estimation in 4D CT Data

“…In [16] the authors extend [12] to include a B-spline-based nonrigid transformation. Variational models using a level set segmentation and a non-linear registration where proposed in [17][18][19][20][21]. However, in these approaches the deformation field is defined only close to object surfaces [18,20] or no shape prior information is incorporated in the model [19][20][21].…”

A Variational Approach for Combined Segmentation and Estimation of Respiratory Motion in Temporal Image Sequences

“…Variational approaches based on level set segmentation and non-linear registration where proposed in [7,1,13,25,12]. However, in these approaches the deformation field is defined only near object surfaces [1,25] or no shape-prior information is incorporated in the model [25,13,12]. In [7] image sequences are segmented into regions of homogeneous motion.…”

“…In [2,8] the joint segmentation/registration of atlas and study images is formulated by a mixture of Gaussians model to describe image intensities and by a non-linear registration of the study image with tissue probability maps provided by the atlas. Variational approaches based on level set segmentation and non-linear registration where proposed in [7,1,13,25,12]. However, in these approaches the deformation field is defined only near object surfaces [1,25] or no shape-prior information is incorporated in the model [25,13,12].…”