[ High dimensional joint radiometric-geometric statistical distance for tracking ]
.: Outline :.
Abstract : This work deals with region-of-interest (ROI) tracking in video sequences. The goal is to determine in successive frames the region which best matches, in terms of a similarity measure, an ROI defined in a reference frame. Two aspects of a similarity measure between a reference region and a candidate region can be distinguished: radiometry which checks if the regions have similar colors and geometry which checks if these colors appear at the same location in the regions. Measures based solely on radiometry include distances between probability density functions (PDF) of color. The absence of geometric constraint increases the number of potential matches. A soft geometric constraint can be added to a PDF-based measure by enriching the color information with location, thus increasing the dimension of the domain of definition of the PDFs. However, high-dimensional PDF estimation is not trivial. Instead, we propose to compute the Kullback-Leibler distance between high-dimensional PDFs without explicitly estimating the PDFs. The distance is expressed directly from the samples using the k-th nearest neighbor framework. Tracking experiments were performed on several standard sequences.
.: Demos :.
Tracking on different videos: green is our method, white cyan magenta and red are classical methods.
Video 1: Football sequence, complex motions
Video 2: Schnee sequence, noise
Video 3: Car sequence, partial occlusions
Video 4: Crew sequence, variations of illuminance
Video 5: Waterobject sequence, zoom experiment
.: References :.
C8
High-dimensional statistical distance for region-of-interest tracking: Application to combining a soft geometric constraint with radiometry S. Boltz, E. Debreuve, M. Barlaud in Proceedings of IEEE International Conference on Computer Vision and Pattern Recognition, (CVPR`07), Minneapolis USA 2007 PreprintPosterBibTeX
C7
kNN-based high-dimensional Kullback-Leibler distance for tracking S. Boltz, E. Debreuve, M. Barlaud in Proceedings of International Workshop on Image Analysis for Multimedia Interactive Services, (WIAMIS`07), Santorini Greece 2007 BibTeX
[ Entropy-based space-time segmentation ]
.: Outline :.
Abstract : This work deals with video segmentation based on motion and spatial
information. Classically, the motion term is based on a motion compensation
error (MCE) between two consecutive frames. Defining a motion-based energy as
the integral of a function of the MCE over the object domain implicitly results
in making an assumption on the MCE distribution: Gaussian for the square
function, Laplacian for the absolute value, and, more generally, parametric
distributions for functions used in robust estimation. However, these
assumptions are not necessarily appropriate. Instead, it is proposed to define
the energy as a function of (an estimation of) the MCE distribution. The chosen
function is a continuous version of the Ahmad-Lin entropy approximation, the
purpose being to be more robust to outliers of the MCE. Since a motion-only
approach can fail with homogeneous objects, the motion-based energy is enriched
with spatial information using a joint entropy formulation. The proposed energy
is minimized iteratively using active contours.
.: Demos :.
Segmentation on sequence "City" : comparison of motion criteria
Video 1: Minimizing entropy of the motion compensation error
Video 2: Using a robust M-estimator on the motion compensation error
Segmentation results, active contour evolution
Video 3: Segmentation on sequence Flower
Video 4: Segmentation on tracking of sequence Schnee
Tracking results on two sequences
Video 5: Soccer sequence, articulated motions
Video 6: Football sequence, complex motion and blur
.: References :.
C6
Space-time segmentation based on a joint entropy with estimation of nonparametric distributions A. Herbulot , S. Boltz , E. Debreuve , M. Barlaud , G. Aubert in Proceedings of International Conference on Scale Space Methods and Variational Methods in Computer Vision, (SSVM`07: joint edition of the 6th Scale Space and the 4th VLSM), Ischia Italy 2007 PreprintBibTeX
C5
Entropy-based space-time segmentation in video sequences S. Boltz, A. Herbulot, E. Debreuve, M. Barlaud in Proceedings of ECCV Workshop on Statistical Methods in Multi-Image and Video Processing, (SMVP`06 *received the best paper award*), Graz Austria 2006 PreprintBibTeX
C2
Robust motion-based segmentation in video sequences using entropy estimator A. Herbulot, S. Boltz, E. Debreuve, M. Barlaud in Proceedings of IEEE International Conference on Image Processing, (ICIP`06), Atlanta USA 2006 PreprintBibTeX
[ A minimum entropy procedure for robust motion estimation ]
.: Outline :.
Abstract : Motion estimation is a critical step in many image processing
problems, such as video compression or object tracking. Here we
focus on the block matching approach and suggest using a
minimum-entropy criterion. Many entro\-py-based estimation
procedures exist, such as plug-in estimators based on Parzen
windowing. We consider here an alternative that is applicable to
data of any dimension and that circumvents the critical issues
raised by kernel-based methods. The inherent robustness property
of entropy is expected to provide a robust and efficient
estimation of the motion vector of a block of a video sequence. In
particular, the minimum-entropy estimator should be robust to
occlusions and variations of luminance, for which standard
approaches like SSD usually meet their limitations.
.: Demos :.
Construction of a very noisy synthetic video, exact motion is known
Video 1: A synthetic sequence showing an head translating on a moving background
Video 2: Video 1 with several alterations : occlusion patches, flash, noise
Coding the first frame, and predicting the others using the motion field
Video 3: Motion field computed with our kNN method
Video 4: Motion field computed with the standard in most video coders : absolute differences
.: References :.
C3
A minimum-entropy procedure for robust motion estimation S. Boltz , E. Wolsztynski , E. Debreuve , E. Thierry , M. Barlaud , L. Pronzato in Proceedings of IEEE International Conference on Image Processing, (ICIP`06), Atlanta USA 2006 PreprintSlidesBibTeX
[ Puzzle : Motion-based segmentation for video coding ]
.: Outline :.
Abstract : Motion compensation is an essential problem in video coding. The
main drawback of the usual motion estimation methods is that they
divide the images into blocks or patches which do not correspond to
moving objects. In this paper, we propose a method to estimate the
motion in regions instead of blocks. We define a cost functional to
estimate simultaneously the segmentation and the motion of the
regions. We introduce a joint motion estimation and segmentation
algorithm based on the derivation of this cost functional. We show
some encouraging results for video compression.
.: References :.
C1
A joint motion computation and segmentation algorithm for video coding S. Boltz, E. Debreuve, M. Barlaud in Proceedings of European Signal Processing Conference, (EUSIPCO`05), Antalya Turkey 2005 PreprintPosterBibTeX
.: References :.
.: Demos :.