Among the challenging problems in computer vision and image processing, change detection plays an important rule. From two images of possibly different nature, the problem consists in identifying the areas where a real change has appeared (binary label change detection) and discriminating the nature of changed area (multi-label change detection). Applications of change detection techniques are numerous. A traditional way to face this problem relies on a thresholding of the image luminance differences. Such natural approach is efficient but is nevertheless strongly related to the value of the considered threshold and appears limited for textured objects or structures with high internal variability. This thesis is concerned with real change detection in pair of images. This is a challenging and open problem since the difficulties stemming from the confusion between real changes (depending on the objects/structures inside the images) and visual changes (observed through the difference in terms of image luminance) are numerous. Many applications are concerned with this crucial task, like video surveillance, event detection or remote sensing. We propose to solve this labeling problem as the minimization of a global cost-function using a min-cut/max-flow strategy. Because of the different nature of the input images (different sensor, shooting angle, ...) and of the variety of detailed information contained in an object, we propose to rely on several criteria, either able to detect abrupt or subtle changes. These criteria are computed on local patches whose size adaptively depends on the structure of the objects inside the images. In addition, bayesian model selection and metropolis-hasting sampling are introduced to efficiently deal with multi-label change detection. Experimental quantitative and qualitative results are shown on synthetic and real data in various applications.
修改评论