In this paper, real time algorithms for segmentation and detection of contrast objects in color images are proposed, which provide stable segmentation components for solving scene recognition problems. The selected contrast objects are approximately described using the graph of structural color bunches STG introduced by the author. A bipartite graph LRG of left and right contrast boundary curves (germs of global objects) in STG is attached to each image. Connected components in LRG are candidates for contrast objects in STG. Constructing algorithms are based on the geometrized histogram method developed by the author. The detection of stable segmentation components in images is applied to analysis of video sequences in order to find visual landmarks. Applications of the developed technique to autonomous navigation of robots are dis cussed.