Region-Based Particle Filter for Video Object Segmentation

Abstract

We present a video object segmentation approach that extends the particle filter to a region-based image representation. Image partition is considered part of the particle filter measurement, which enriches the available information and leads to a re-formulation of the particle filter. The prediction step uses a co-clustering between the previous image object partition and a partition of the current one, which allows us to tackle the evolution of non-rigid structures. Particles are defined as unions of regions in the current image partition and their propagation is computed through a single co-clustering. The proposed technique is assessed on the SegTrack dataset, leading to satisfactory perceptual results and obtaining very competitive pixel error rates compared with the state-of-the-art methods.

Fully Connected Object Proposals for Video Segmentation

“…Significant progress has been achieved by methods designed to track keypoints over time and, more recently, over image regions [4,27,41]. These methods, however, only consider two consecutive frames of video and cannot handle sudden motion and appearance changes (i.e.…”

3D shape reconstruction from a humanoid generated video sequence

“…In our work, we propose to use the region-based particle filter presented in [14] to allow tracking and segmenting objects in sequences of different views. This is a suitable algorithm for this task as a geometrical shape is not considered to represent the object.…”

“…In this work, we adapt the region-based particle filter presented in [14] to extract the 2D shape of the object from partitions (See Figure 5(b)) associated with a set of multiple views of the object. Only images containing the entire object without blur are processed and a subset of these images is finally selected to reconstruct the object in accordance with their final segmentation quality.…”

“…Then, parts associated with both the object and background are put in correspondance for each pair of views. In [14], a region-based particle filter is presented in which Monte Carlo methods and a representation of the image in terms of regions are combined. This algorithm, does not only provide the position of the object as in the color-based approach.…”

Evolutionary Computing and Particle Filtering: A Hardware-Based Motion Estimation System

“…These applications include, but are not limited to, visual tracking [25] [26], object detection [27], image segmentation [28], contour detection [29], video stabilization [30], and even point set registration [31], i.e. finding a spatial transformation that aligns two given point sets.…”