Conference Paper

A variational framework for image segmentation combining motion estimation and shape regularization

Dept. of Comput. Sci., Univ. of California, Los Angeles, CA, USA
DOI: 10.1109/CVPR.2003.1211337 Conference: Computer Vision and Pattern Recognition, 2003. Proceedings. 2003 IEEE Computer Society Conference on, Volume: 1
Source: DBLP

ABSTRACT

Based on a geometric interpretation of the optic flow constraint equation, we propose a conditional probability on the spatio-temporal image gradient. We consistently derive a variational approach for the segmentation of the image domain into regions of homogeneous motion. The proposed energy functional extends the Mumford-Shah functional from gray value segmentation to motion segmentation. It depends on the spatio-temporal image gradient calculated from only two consecutive images of an image sequence. Moreover, it depends on motion vectors for a set of regions and a boundary separating these regions. In contrast to most alternative approaches, the problems of motion estimation and motion segmentation are jointly solved by minimizing a single functional. Numerical evaluation with both explicit and implicit (level set based) representations of the boundary shows the strengths and limitations of our approach.

Full-text preview

Available from: psu.edu
  • Source
    • "While edge-based methods are used to segment a feature that has a well-defined boundary, region-based methods can be used when a feature in an image has statistics that are different from the background. Examples of such statistics are inten- sity [7], texture [4], or motion [9]. Region-based level set methods can also be used to segment vector-valued images [6], or to segment an image into more than two regions using multi-phase level sets [4,32]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: We present a novel tree balancing constraint that is slightly stronger than the well-known 2-to-1 balancing constraint used in octree data structures (Tu and O’hallaron, Balanced refinement of massive linear octrees. Tech. Rep. CMU-CS-04-129. Carnegie Mellon School of Computer Science, Pennsylvania, 2004). The new balancing produces a limited number of local cell connectivity types (stencils): 5 for a quadtree and 21 for an octree. Using this constraint, we interpolate the data sampled at cell centers using weights pre-computed by interpolation or by generating interpolation codes for each stencil. In addition, we develop a parallel tree adjustment algorithm, and show that the imposed balancing constraint is satisfied even when the tree is adjusted in parallel. We also show that the adjustment has high parallelization performance. We finally apply the new balancing scheme to level set image segmentation and smoke simulation problems.
    Full-text · Article · Sep 2014 · The Visual Computer
  • Source
    • "Other methods use motion projection to link segments, i.e., the position of a segment in a future frame is estimated from its current position and motion features [6] [7] [8] [9]. Cremers (2003) tracked motion segments by jointly solving motion segmentation and motion estimation by minimizing a single energy functional [10]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Model-free tracking is important for solving tasks such as moving-object tracking and action recognition in cases where no prior object knowledge is available. For this purpose, we extend the concept of spatially synchronous dynamics in spin-lattice models to the spatiotemporal domain to track segments within an image sequence. The method is related to synchronization processes in neural networks and based on superparamagnetic clustering of data. Spin interactions result in the formation of clusters of correlated spins, providing an automatic labeling of corresponding image regions. The algorithm obeys detailed balance. This is an important property as it allows for consistent spin-transfer across subsequent frames, which can be used for segment tracking. Therefore, in the tracking process the correct equilibrium will always be found, which is an important advance as compared with other more heuristic tracking procedures. In the case of long image sequences, i.e., movies, the algorithm is augmented with a feedback mechanism, further stabilizing segment tracking.
    Full-text · Article · Nov 2009 · Sensors
  • Source
    • "With the advantage of being implicit, intrinsic and parameter-free, level sets track moving interfaces through either model-free (Paragios and Deriche, 2002) or either model-based (Cremers, 2003) methods. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we propose a novel mathematical model to address segmentation from an aerial and satellite imagery. Segmentation is performed through the consistent recovery of the zero iso-surfaces of a level set function towards image's foreground (desired object for extraction) and background discrimination. The level set method is embedded in a geodesic active contours variational formulation. Geodesic active contours technique is an advancement of the classical snakes and active contours that can handle the limitation to change their topology. The proposed mathematical model consists of a functional with two basic energies; one that forms a region-based energetic module for the evolving interface in the level set space and an another one that takes into account the a priori knowledge of the geometry of the desired for extraction objects. With such a functional shape information is embedded into the level set based segmentation scheme. Shape information can cope with missing or misleading information in the input images due to noise, clutter and occlusion. The shape priors were developed to segment objects of familiar shape in a given image. Promising results demonstrate the potentials of our approach, like in cases where the desired object for extraction were buildings.
    Preview · Article · Jan 2009
Show more