Level Set Evolution Based Logic Fusion: A Novel Man-made Objects Segmentation from Radar Image

Our purpose is to explore an application of level set evolution theory to seek the solution of objects (as bridge or dam) segmentation above river from radar imagery with unwelcome effects. We present a novel level set evolution formulation for improving accuracy of segmentation. Critically, a logic fusion term is constructed using logic operators defining in multi-source satellite imagery with high precision of registration by logical operations, such as union, intersection and negation, then incorporated into the general evolution equation. Under the logic framework, we combine diverse features both from radar and optical image into evolution equation so as to overcome disadvantages, such as missing information, image artifacts and weak boundary and help objects segmentation to radar image. Results of experiments on pairs of ERS and LANDSAT imagery and pairs of RADARSAT and SPOT imagery, demonstrate effectiveness and robust of our method to objects segmentation in complex cases