Alain Jacot-Descombes

Video segmentation and camera motion characterization using compressed data

We address the problem of automatically extracting visual indexes from videos, in order to provide sophisticated access methods to the contents of a video server. We focus on tow tasks, namely the decomposition of a video clip into uniform segments, and the characterization of each shot by camera motion parameters. For the first task we use a Bayesian classification approach to detecting scene cuts by analyzing motion vectors. For the second task a least- squares fitting procedure determines the pan/tilt/zoom camera parameters. In order to guarantee the highest processing speed, all techniques process and analyze directly MPEG-1 motion vectors, without need for video decompression. Experimental results are reported for a database of news video clips.

Asynchronous perceptual grouping: from contours to relevant 2-D structures

This paper addresses the issue of bottom-up recovery of structured 2-D geometric primitives from real, complex images. A novel approach for perceptual grouping is presented, using asynchrony whose fundamental principle is to process some primitives before some others, because they carry more information. The main contribution of this work lies in the reduction of the complexity of the grouping process to a problem linear with respect to the number of image contours. Another advantageous characteristic of the proposed method is that only two parameters need to be specified. The asynchronous perceptual grouping processes a data flow of ordered contour primitives, with a rank in the flow determined according to a measure of saliency. It sequentially attempts to group each contour primitive coming in the data flow with other neighboring ones having arrived earlier. Acceptance or rejection of the current contour depends on the geometrical configuration of its neighborhood in the image; properties tested include proximity, colinearity, curvilinearity, and parallelism. The topology as well as the morphology of the relevant structures emerging throughout the whole image are captured in a rigid graph. Junction and surface characteristics are easily extractable, which can help to choose pertinent primitives for an indexing or a recognition system.

Efficient Segmentation and Camera Motion Indexing of Compressed Video

In order to provide sophisticated access methods to the contents of video servers, it is necessary to automatically process and represent each video through a number of visual indexes. We focus on two tasks, namely the hierarchical representation of a video as a sequence of uniform segments (shots), and the characterization of each shot by a vector describing the camera motion parameters. For the first task we use a Bayesian classification approach to detecting scene cuts by analysing motion vectors. Adaptability to different compression qualities is achieved by learning different classification masks. For the second task, the optical flow is processed in order to distinguish between stationary and moving shots. A least-squares fitting procedure determines the pan/tilt/zoom camera parameters within shots that present regular motion. Each shot is then indexed by a vector representing the dominant motion components and the type of motion. In order to maximize processing speed, all techniques directly process and analyse MPEG-1 motion vectors, without the need for video decompression. An overall processing rate of 59 frames/s is achieved on software. The successful classification performance, evaluated on various news video clips for a total of 61 023 frames, attains 97.7% for the shot segmentation, 88.4% for the stationary vs. moving shot classification, and 94.7% for the detailed camera motion characterization.

Probabilistic approach to 3-D inference of geons from a 2-D view

A new, probabilistic approach for inferring 3-D volumetric primitives from a single 2-D view is presented. This recognition relies on the assumption that every object can be decomposed into component parts that belong to a finite set or alphabet of volumetric primitives (geons). For each possible primitive from the permissible set, a conditional probability function is computed. This law specifies the probability of obtaining the primitive given an observable 2- D measure or feature. The distribution functions are determined by simulation, on the basis of a representative number of random projections of the primitives. The measures themselves are chosen in such a way that they can easily be extracted from real images and their discriminative power for the volumetric primitive inference is high. Examples illustrate the proposed approach.

LaboImage: portable window-based environment for research in image processing and analysis

LaboImage provides scientists with general purpose as well as specific processing families and tools in a highly interactive environment. The current software results from an evolution reflecting several years of development and experiences. This paper first presents the new X Window / OSF Motif version of LaboImage, as it is seen by the user. It also describes how an image is manipulated in the system, how processing methods are applied and results are visualized. Multiple types of interaction between the user and the system are addressed. The implementation aspects are then detailed. They concern data structures as well as algorithms and interfaces. The data file and descriptor file formate used for strong images is described. The organization in memory of multiple data such as images, vectors, and macros is presented. The source code organization is also discussed. A clear separation between algorithmic and interface parts in the code appears to be very important, in order to allow easy further developments of the system.

LABOIMAGE: AN EXPERIMENTAL WINDOW-BASED ENVIRONMENT FOR TEACHING AND RESEARCH IN IMAGE ANALYSIS

LaboImage provides teachers and scientists with general purpose as well as specific processing families and tools in a highly interactive and experimental environment. The current software reflects half a decade of development and experiences. Throughout this paper, the initial need for such a software is presented as well as its evolution. The current features are listed; also described is the LaboImage interface design and how users interact with the software. The implementation aspects are then detailed, concerning data structures as well as algorithms and interfaces. The data file and descriptor file format used for storing images is described. The organization in memory of multiple data such as images, vectors and macros is presented, as well as the source code organization. A clear separation between algorithmic and interface code has proven to be very important, in order to allow easy extensions of the basic system. Finally, some aspects that distinguish LaboImage from other imaging softwares are discussed.