Tracking of Objects in Video Scenes with Time Varying Content

  • PDF / 7,259,274 Bytes
  • 13 Pages / 600 x 792 pts Page_size
  • 88 Downloads / 256 Views

DOWNLOAD

REPORT


Tracking of Objects in Video Scenes with Time Varying Content Amal Mahboubi IRCCyN UMR 6597 CNRS EPUN, rue Christian Pauc La chantrerie, BP 60601, 44306 Nantes, France Email: [email protected]

Jenny Benois-Pineau LABRI, CNRS UMR 5800, Universit´e Bordeaux-1, 33405 Talence, France Email: [email protected]

Dominique Barba IRCCyN UMR 6597 CNRS EPUN, rue Christian Pauc La chantrerie, BP 60601, 44306 Nantes, France Email: [email protected] Received 27 July 2001 and in revised form 5 February 2002 We propose a method for tracking of objects contained in video sequences. Each video object is represented by a set of polygonal regions. A bottom up approach (spatial segmentation/motion estimation) is applied for the initialisation of the method, a limited human interaction is used to build the semantic map of the first frame in video sequence. The tracking of this model along a video sequence is based on detecting and indexing new objects in a video scene. Semantic rules are used to label new objects and, the current state of segmentation is validated by forward projection of the background. Keywords and phrases: new region extraction, labelling, object indexing, forward projection.

1.

INTRODUCTION

The new ongoing standard of video representation and coding MPEG4 [1] gives tremendous possibilities for the composition of heterogeneous video scenes combining video objects of various nature. The main challenge behind MPEG4 technology is the development of efficient and truly automatic methods for extracting and tracking of objects in video. Once video objects are known at each time instant, they can be manipulated, put into another scene and so on. Numerous research works, developed recently [2, 3, 4, 5, 6], are devoted to the problem of automatic tracking of a selected video object planes (VOP) in a scene. The focus of study [2] is an object tracking which combines motion and spatial information in order to be able to track objects which do not present either homogeneous texture or motion. The method has been applied to the problem of the generation of video objects for content-based functionalities in objectbased coding schemes. To extend this work, [3] presents an interesting technique for generic object tracking. This method perfects the first tracking scheme [2] by projecting region-based partition. This projection accommodates the previous partition information in the current image. Then

the object partition is re-segmented and it is projected on the following image using motion information. Nevertheless, this method assumes that objects have been defined in the first image and the process is not able to correctly detect a new object in the scene. To mitigate those consequences, it introduces the concept of user interaction in the algorithm to handle variation of objects. The works [4, 5] surveyed the region-based active contours approaches. The first study introduces the active contour criteria and the second presents the use of a B-spline parametric contours for object tra