With the rapid development of remote sensing, digital image processing has become an important tool for the quantitative and statistical analysis of remotely sensed images. These images most often contain complex natural scenes. The robust interpretation of such images requires the use of different sources of information about the scenes under consideration. This paper presents an integrated approach to robust analysis of SPOT images with the aid of map information as well as a priori knowledge about the contextual information of images. Markov random field theory and the Bayes formula are used to formulate the image analysis problem as a problem of optimization of an objective function, which in turn permits the application of various existing optimization algorithms to solve the problem. To increase the robustness of the result, several techniques are proposed to effectively use map information and image contextual information. The first one is concerned with the estimation of the parameters in the objective function with the help of these two sources of information. The second one is the integration of map information in Bayes image modeling using a Markov random field. The third one is a new optimization algorithm which takes into account map information and image contextual information by means of a feedback control scheme. The last technique proposed to increase the robustness of the result is concerned with the fusion of several (intermediate) analysis results by again using map knowledge and image contextual information for the estimation of the reliability of these results.

The scale-space representations and wavelet transform theory have provided us with good singularity detection frameworks. Thanks to such methods, image understanding processes have become more and more powerful. Some computer vision tasks also require a knowledge on the nature of detected singularities. We propose, in this paper, to take into account these requirements, and we present a singularity detection method based on fractional calculus arguments.>

This paper presents MESSIE, a multi-specialist scene analysis system. It is a centralized hierarchical blackboard architecture. The generic model of objects and the explicit description of sensors and materials allow the use of an application-independent interpretation strategy. Three applications of 2D and 3D scene analysis have been developed to validate MESSIE. The architecture of MESSIE, the objects model and the different interpretation and detection strategies are included. Current experiments on indoor scene interpretation and an interpretation running example using constrained low-level feature extraction mechanism to improve low-level processing results are described.>

Binocular stereo vision processes estimate 3D surfaces using a pair of images taken from different points of view. 3D surface characteristics are estimated by matching 3D image areas or features corresponding to the projections of same 3D points. The most classic area-based methods used cross correlation with a fixed window-size, but this technique presents a major draw-back: the computation of depth is generally prone to errors close to surface discontinuities. In this paper, we present our current results on aerial stereo images of urban areas from correlation-based algorithm, with an adaptive window-size constrained by an edge map extracted from the images. This technique is not used to refine an initial disparity map, but to create one with good discontinuities location.

Spatiotemporal segmentation in echocardiographic image sequences is discussed. Spatial properties and temporal properties are combined to compute segmentation and tracking in a single process. The Markov random field (MRF) framework is used for modeling the energy function. Starting from a reference image, where a manual segmentation is made, a method is developed to estimate the model parameters. An estimation is a crucial point in MRF models. Thus, given an initial segmentation of the sequence, this approach can segment and track a cardiac cavity during the cardiac cycle. Its performance is demonstrated on a real echocardiographic sequence.>

A regularity measure for discrete line geometry is presented. This quantitative measure based on a ratio between line lengths at different scales is analyzed in the framework of Brownian motion theory. The measure at a given scale is always computed from the maximum precision image, so that it does not introduce any subresolution assumption. A scale choice determines the quantity of global information vs. local information to be measured. Its statistical behavior is studied on two extremal models of curves: the Brownian motion and the digitized straight line. It is shown that this quantitative measure leads to relevant shape information. To illustrate this fact, an image segmentation application example is discussed based essentially on geometry criteria of region boundaries. Some experimental results performed on real-scene images are presented.>

A scene interpretation system is presented in a multisensor fusion context the application of which involves the interpretation of remote sensed images. The authors discuss how multisensor fusion is achieved, and derive the modeling problems for objects, scene, and strategy. The proposed multispecialist architecture takes into account the knowledge about sensors, the multiple viewing notion (shot), and the uncertainty and imprecision of models and data modeled with possibility theory. Generic models of objects are represented by concepts independent of sensors. Three kinds of specialists are present in the architecture: generic specialists (scene and conflict), semantic object specialists, and low level specialists. A blackboard structure with a centralized control is used. The interpreted scene is implemented as a matrix of pointers allowing easy conflict detection and spatial context verification. An interpretation system for SAR/SPOT sensor images is described, and an example of a scene interpretation is shown.>

A region segmentation algorithm is presented, using a model for joint probability density. Joint probability density can be defined as an N*N cooccurrence matrix in which each coordinate (i, j) gives the probability for the gray-level transition i, j between two neighbor pixels. The approach consists in modeling the energy distribution within a cooccurrence matrix of a region. Regions are assumed to be stationary. A region-growing scheme that proceeds in two steps is used. The first step consists of learning the parameters of the model. The second step is the segmentation process. Starting with a seed pixel, new pixels are incorporated in the region if their neighborhoods fit the model.>

Presents a scene interpretation architecture in the context of multi-sensor fusion. This architecture is based on blackboard and multi-specialist concepts, and offers knowledge representation capabilities to express knowledge about an application in an explicit way. Knowledge description includes the characteristics of the sensors, and a semantic object description independent of the sensor characteristics. The detection strategy is based on the fundamental notion of spatial context linking the objects of the scene, and the notion of salient object. An interpretation is a semantic model of the real observed scene, including object location, object characteristics and relations between objects. This architecture has a highly modular structure allowing easy incorporation of new knowledge, and new specialists. To demonstrate the reliability of the approach, an application of SAR/SPOT data fusion developed with this architecture is briefly presented, and the results of an interpretation session are shown.>

Multiscale Analysis of surfaces allows a hierarchical representation of their composing features. To represent a surface at a given scale, structures that are insignificant at that scale have to be eliminated. A typical example for this approach is cartography. However, the aims of cartographers reach beyond simply gradually eliminating the structures; in the majority of cases, the nature of geomorphological structures which compose the surface have to be preserved across all scales. Thus a global smoothing of the surface is not suitable to solve the present problem, since that would cause inevitably morphological modifications of certain important structures. In fact, the points to be preserved across scale variations are to be chosen interactively by the user. The authors present a surface model which allows them to perform a Multiscale Analysis which takes the importance of local structures into consideration, i.e. structures which are inherent to the relief morphology. From that discrete model we extract a context-dependent Multiscale Analysis Operator which can be isotropic or anisotropic and can be expressed in different forms.>

A formal representation of corner and vertex detection is presented. In particular, an analytical study is presented which allows one to know exactly what the behavior is of a detector around trihedral vertices. It is shown that near three surfaces, two elliptic maxima of DET exist, and their location is inside extremal contrast surface. The intermediate surface always shows a hyperbolic minima. It is shown that the detector allows to find the exact position of vertex. The approach proposed has been tested on many noisy synthetic data and real images and its robustness seems promising.>

Consideration is given to a corner model, and its behavior in the scale-space is studied. The authors derive results that clarify the behavior of some well known approaches used to detect corners. In particular, they show that some of the approaches are inadequate for an exact localization of the corner. A novel approach is then proposed in order to correct the displacement effect and detect exactly the corner position. Some promising experimental results obtained on real data are shown.>

The architecture of an image analysis system called MESSIE (Multi Expert System for Scene Interpretation and Evaluation) is presented. This system reasons from geometric models which are represented by four concepts (geometry, radiometry, context, and function). The aim is to find the class instances from the generic models which are present in the scene. The necessity of having a hierarchic and opportunistic method of solving the problem of interpretation is shown. In a first step, MESSIE tries to detect salient objects. Then, using characteristics of salient objects and knowledge about the context of objects, MESSIE tries to confirm the object hypothesis and to infer new objects in the scene. The domain used to develop MESSIE is aerial imagery interpretation. Results on the detection of roads and buildings in suburban images are given.>

OPSILA, is a description given of a general-purpose parallel architecture with two different forms of parallelism: the well-known SIMD (single-instruction, multiple data shown), a synchronous form of parallelism; and SPMD (single program, multiple data stream), which is an asynchronous mode. It is shown that OPSILA is efficient for a wide variety of image algorithms including low and high level processing. The efficiency of OPSILA is demonstrated for the low-level algorithms, through the implementation of a set of typical operations: local (convolution), global (histogram), and geometric corrections.>