Incorporating a pre-attention mechanism in fuzzy attribute graphs for sequential image segmentation

Geoffroy Fouquier, Jamal Atif, Isabelle Bloch.

Sequential methods for knowledge-based recognition of structures require to define in which order the structures have to be recognized. We propose to address this problem by integrating pre-attention mechanisms, in the form of a saliency map, in the determination of the order. As pre-attention mechanisms extract knowledge from an image without object recognition in advance and do not require any a priori knowledge on the image, they provide useful knowledge for guiding object segmentation and recognition. Additionally, we make use of generic knowledge of the scene, expressed as spatial relations, since they play a crucial role in model-based image recognition and interpretation due to their stability compared to many other image appearance characteristics. Graphs are well adapted to represent this information, and finding an order then amounts to find a path in a graph. The proposed algorithms are applied on brain image understanding.

PDF full paper