F. Meyer, P. Bouthemy, Region-based tracking using affine motion models in long image sequences, CVGIP: Image Understanding 60 (1994) 119--140.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Turkey. 1 Introduction Object based video processing necessitates flexible representation of video objects that are robust to occlusions, including self occlusions due to out of plane rotations. Several methods for video object representation and tracking have been proposed in the literature [1, 2, 3]. The 2 D mesh based method in [4] has been demonstrated to be useful for tracking the local motion and intensity variations of video objects when there are no occlusions, and it has been used successfully for video object transfiguration. The motion representation utilized in that work is ....

F. G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60(2):119-140, Sept. 1994.

.... object appearance and dynamics, along with a relatively simple form of data association [16, 19] To circumvent the need for such specific prior knowledge, one could rely on bottomup, motion based approachs to segmenting moving objects from their backgrounds, prior to tracking and identification [10, 18]. Layered image representations provide one such approach [12, 20, 25, 28] With probabilistic mixture models and the EM (ExpectationMaximization) algorithm [6] efficient methods have been developed for determining the motion and the segmentation simultaneously. In particular, these methods give ....

F.G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60:119--140, 1994.

....and lines are commonly used as rigid object representations [35] but are limited when dealing with deformable objects. Different sorts of object templates have been widely employed for object tracking [32] Finally, representations based on 2 D regions have been used in the past for tracking [19], and more recently for SVO extraction [13] 17] A survey of the previous work in this area points toward a generic four step SVO extraction model, that comprises many of the existing methods: 1) SVO structure definition; 2) SVO computation; 3) SVO tracking, and (4) SVO postprocessing. In ....

....is sensitive to artifacts in object borders. On the other hand, the tracking process is responsible for discerning between past and present information in order to keep an accurate SVO representation along time. Considerable research on region based object tracking has been done [9] 11] [19], 34] Most of it can be generically labeled as partition prediction adjustment methods, in which the problem can be stated as follows: given the set of SVOs, the image sequence, and the last available partition, how does this partition need to be updated deformed to generate the current one The ....

F. Meyer and P. Bouthemy, "Region-based tracking using affine motion models in long image sequences," CVGIP: Image Understanding, vol. 60, pp. 119--140, 1994.

....and regionbased motion detection segmentation information. During the last years, a great variety of tracking algorithms have been proposed. These may be classified in two distinct categories: Motion based approaches rely on a robust method for grouping visual motion consistencies over time [10]. These methods are quite fast but they cannot ensure that the tracked regions have any meaning. Model based approaches impose high level semantic representation and knowledge [8] and are more reliable compared with the motion based approaches. On the other hand they suffer from being ....

F. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60:119--140, 1994.

....a closed form solution. The tracking is based on least squares fitting of the affine parameters relating sets of feature positions in adjacent frames, followed by smoothing of the parameters using a Kalman filter. In this respect, it resembles the affine region tracker described by Meyer et al. [8]. As well as improving stability, this also gives potential for over coming problems caused by missing features due to occlusions, etc. The pose estimation is achieved by an adaptation of the circle ellipse correspondence algorithm [7] if we use a fronto parallel view of the face as the ....

....7 5 (5) where 0 T = 0; 0] The resulting parameter estimates a from each frame are then fed into a Kalman filter (KF) The purpose of the filter is to smooth the variation of affine transformation over time. In effect, it is a feature based version of the affine region tracker described in [8]. The filter state is simply the true affine parameters and their derivatives, ie [a; a] T , and we adopt constant velocity state dynamics, ie a(k 1) a(k 1) a(k) a(k) a(k) ffl(k) 6) where the ffl(k) are normally distributed, zero mean, with covariance P. The filter ....

F. G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. Computer Vision, Graphics and Image Processing: Image Understanding, 60(2):119--140, 1994.

....a closed form solution. The tracking is based on least squares fitting of the affine parameters relating sets of feature positions in adjacent frames, followed by smoothing of the parameters using a Kalman filter. In this respect, it resembles the affine region tracker described by Meyer et al. [9]. As well as improving stability, this also gives potential for over coming problems caused by missing features due to occlusions, etc. The pose estimation is achieved by an adaptation of the circle ellipse correspondence algorithm [8] if we use a fronto parallel view of the face as the ....

....7 5 (6) where 0 T = 0; 0] The resulting parameter estimates a from each frame are then fed into a Kalman filter (KF) The purpose of the filter is to smooth the variation of affine transformation over time. In effect, it is a feature based version of the affine region tracker described in [9]. The filter state is simply the true affine parameters and their derivatives, ie [a; a] T , and we adopt constant velocity state dynamics, ie a(k 1) a(k 1) a(k) a(k) a(k) ffl(k) 7) where the ffl(k) are normally distributed, zero mean, with covariance P. The filter ....

F. G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. Computer Vision, Graphics and Image Processing: Image Understanding, 60(2):119--140, 1994.

....[6] A good initialization may be provided by a temporal tracking of the model which enables to resort to fast local (deterministic) optimization procedures. Kalman filtering techniques have been widely used in computer vision for tracking various image features (points [1] edges [5] regions [11]) The case of deformable structures has been handled recently [2, 3, 13] by considering contours, characteristic points or regions as features for the Kalman filter governing equations. In this paper, we develop a Kalman filter which performs prediction and filtering of the global deformation ....

F. MEYER and P. BOUTHEMY. -- Region-based tracking using affine motion models in long image sequences. -- CVGIP: Image Understanding, to appear, Sept. 1994.

....for instance be used for interpretation purposes. A review of existing methods in curve tracking may be found in [7] Contrary to standard approaches, based on the tracking of discrete low level or intermediate level image primitives such as characteristic points [2] segments [19] or regions [42], the tracking procedure proposed here considers the structure of 182 KERVRANN AND HEITZ a b c d Figure 5: Segmentations with missing data and partial occlusions. interest and its possible deformations as a whole by tracking the hyperparameters Theta = M(k; T;b ) describing the ....

....and partial occlusions. interest and its possible deformations as a whole by tracking the hyperparameters Theta = M(k; T;b ) describing the configuration of the deformable template. The tracking of rigid motion parameters, describing the movement of homogeneous regions, was proposed in [42]. This approach has recently been extended to rigid motions associated to active contour models in [3] In [57] Terzopolous and Szelisky use a Kalman filter to guide the dynamic evolution of an active contour model. In their approach (called Kalman snakes ) the deformable model is governed by ....

[Article contains additional citation context not shown here]

F. Meyer and P. Bouthemy, Region-based tracking using affine motion models in long image sequences, CVGIP: Image Understanding 60, 1994, 119-140.

....sites entail large numbers of surveillance cameras. In practice, since one operator cannot monitor all scenes simultaneously, it is useful to single out the scenes containing potentially interesting events. A key success factor for security applications is to maintain a low false alarm rate (see [18] [21] 36] ffl Traffic surveillance. Simulation of traffic flow using computer models provides powerful and cost effective tools for the design of roading projects. These models entail a number of parameters that must be calibrated using data measured from real traffic flows. Active projects in ....

F.G.Meyer and P.Bouthemy. Region based tracking using affine motion models in long image sequences. CVGIP - Image Understanding, 60(2), 1994.

....like 3 D reconstruction and 3 D representation. During the last years, a great variety of tracking algorithms have been proposed. They may be classified in two distinct categories: 1. The Motion based approaches that rely on a robust method for grouping visual motion consistencies over time [12]. 2. And the Model based approaches that impose highlevel semantic representation and knowledge [5, 9] In both cases the tracking is performed using measurements provided by geometrical or region based properties of the # This work was funded in part under the VIRGO research network (EC ....

.... In this direction there are two main approaches: the boundary based (they usually referred to as edge based approaches) rely on the information provided by the object boundaries (shape properties) 6, 8, 11] while the region based approaches rely on information provided by the entire region [1, 12] (texture and motion based properties) Besides, there are some efforts to combine the boundary and region based modules under a common framework [2] The objective of the proposed work is to provide a general framework that incorporates boundary based and region based tracking modules under the ....

F. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60:119--140, 1994.

....model that integrates boundary and region based information. During the last years, a great variety of tracking algorithms have been proposed. These may be classified in two distinct categories: 1. Motion based approaches rely on a robust method for grouping visual motion consistencies over time [14]. # This work was funded in part under the VIRGO research network (EC Contract No ERBFMRX CT96 0049) of the TMR Program. 2. Model based approaches impose high level semantic representation and knowledge [12] but they suffer from being computationally expensive. In both cases the tracking is ....

....(they are usually referred as edge based approaches) rely on the information provided by the object boundaries (shape properties) 4, 8, 11] and are usually implemented using active contour models. On the other hand, the region based approaches rely on information provided by the entire region [14, 15] (texture and motion based properties) Finally, there are some approaches that combine intensity, boundary and motionbased information [3, 7] The objective of this work is to deal simultaneously with the motion estimation and the tracking problem under a front propagation framework. Towards ....

F. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60:119--140, 1994.

....objects in the scene, denoted by N , is considered as known. We assume that the 2 D apparent motion field induced by them can be approximated by 6 parameter affine models. Affine motion models have been proved useful for motion estimation [27] 28] motion segmentation [9] 8] 29] and tracking [30] [31] and are a good compromise between efficiency and complexity. Denoting with = ff 1 : ff 6 ) the parameters of the model and v i the model generated motion vector at image location i = x; y) v i ( u i v i = ff 1 x ff 2 y ff 3 ff 4 x ff 5 y ff 6 (2) The motion ....

F.G. Meyer and P. Bouthemy, "Region-based tracking using affine motion models in long image sequences," CVGIP: Image Understanding, vol. 60, no. 2, pp. 119--140, Sep. 1994.

....requires tracking of such objects. Object tracking has been an active field of research in computer vision for more than a decade. Existing methods for object tracking can be broadly classified as feature point (or token) tracking [4] boundary (and thus shape) tracking [5] and region tracking [6]. These methods address tracking the shape and location of the object, if the object boundary is provided only in one frame, the starting frame, or not provided in any frame. In the latter, motion segmentation methods are employed to find the objects in a scene. Yet, most existing approaches ....

F.G. Meyer and P. Bouthemy, "Region-based tracking using affine motion models in long image sequences," CVGIP: Image Understanding, vol. 60, pp. 119-140, Sept. 1994.

....2 D motion fields generated by surface patches moving in the scene. The advantage of using them in preference to dense optical flow measurements is twofold: the number of affine patches will be relatively small in a typical sequence; and they provide an effective way of tracking motion fields [16, 10, 17]. Despite this, direct use of affine flow parameters for recovering 3 D information has received surprisingly little attention. The notable exceptions are Negahdaripour and Lee [13] and Meyer [9] who describe closed form solutions using the formulations of Longuet Higgins and Pradzny [7] and ....

F. G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long range sequences. CVGIP: Image Understanding, 60(2):119--140, 1994.

No context found.

F. Meyer, P. Bouthemy, Region-based tracking using affine motion models in long image sequences, CVGIP: Image Understanding 60 (1994) 119--140.

No context found.

F. G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long range sequences. CVGIP: Image Understanding, 60(2):119--140, 1994.

No context found.

F. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60:119--140, 1994.

No context found.

F.G Meyer and P. Bouthemy, "Region-based tracking using affine motion models in long image sequences", CVGIP: Image Understanding, 60(2), pp. 119-140, 1994.

No context found.

F. G. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. Computer Vision, Graphics and Image Processing: Image Understanding, 60(2):119--140, 1994.

No context found.

F. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP: Image Understanding, 60:119--140, 1994.

No context found.

F. Meyer and P. Bouthemy, "Region-Based Tracking Using Affine Motion Models in Long Image Sequences," in CVGIP:Image Understanding, (1994) pp. 119-140.

No context found.

Meyer, F.G. and Bouthemy, P. Regionbased tracking using Affine motion models in long image sequences. CVGIP, Image Understanding, Vol. 60, No. 2, pp. 119-140, 1994.

No context found.

F.Meyer, P.Bouthemy. Region-based tracking using affine motion models in long image sequence. Computer Vision Graphics Image processing, 60 (2), 1994, 119-149.

No context found.

F. Meyer and P. Bouthemy. Region-based tracking using affine motion models in long image sequences. CVGIP:IU, 60(2):119--140, 1994.

No context found.

F. G. Meyer and P. Bouthemy, Region-Based Tracking Using Affine Motion Models in Long Image Sequences, CVGIP:Image Understanding, Sep 1994.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC