This paper describes a method for detecting salient regions in remote-sensed images, based on scale and contrast interaction. We consider the focus on salient structures as the first stage of an object detection/recognition algorithm, where the salient regions are those likely to contain objects of interest. Salient objects are modeled as spatially localized and contrasted structures with any kind of shape or size. Their detection exploits a probabilistic mixture model that takes two series of multiscale features as input, one that is more sensitive to contrast information, and one that is able to select scale. The model combines them to classify each pixel in salient/nonsalient class, giving a binary segmentation of the image. The few parameters are learned with an EM-type algorithm.

This paper explores a generic approach to predict the output accuracy of an algorithm without running it, by a careful examination of the local context. Such a performance prediction will allow one to qualify the appropriateness of an algorithm to treat images with given properties (contrast, resolution, noise, richness in details, contours or textures, etc.) resulting either from experimental acquisition conditions or from a specific type of scene. We have to answer the following question: a context c being given at any site, what will be the performance? In our experiments, c is described by three contextual variables: Gabor components, entropy and signal noise ratio. As initially proposed in the related work, the prediction function is determined from training using a logistic regression model. This technique is illustrated on aerial infrared images for two types of algorithm: edge detection and displacement estimation.

We are given a set of points in a space of high dimension. For instance, this set may represent many visual appearances of an object, a face, or a hand. We address the problem of approximating this set by a manifold in order to have a compact representation of the object appearance. When the scattering of this set is approximately an ellipsoid, then the problem has a well-known solution given by principal components analysis (PCA). However, in some situations like object displacement learning or face learning, this linear technique may be ill-adapted and nonlinear approximation has to be introduced. The method we propose can be seen as a nonlinear PCA (NLPCA), the main difficulty being that the data are not ordered. We propose an index which favors the choice of axes preserving the closest point neighborhoods. These axes determine an order for visiting all the points when smoothing. Finally, a new criterion, called "generalization error", is introduced to determine the smoothing rate, that is, the knot number for the spline fitting. Experimental results conclude this paper: The method is tested on artificial data and on two databases used in visual learning.

The paper is related to a nondestructive control industrial task: the detection of defects in gamma radiographic images. The images are very noisy and have a strong luminosity gradient. The authors adopt a Bayes-Markov model in order to estimate the noise, the gradient and the defects. The proposed model is general and can be used in other situations for detecting valley bottoms in noisy images.>

To develop a method of automatic hip prothesis design, it is necessary to achieve a detection of the edges of the medullary canal of the femur, from X-ray images. The procedure of the pattern recognition of the medullary canal is described. It uses a statistical detector of change in level in a white noise and a bayesian estimation of a step signal in an autoregressive noise.