Scene segmentation in remotely sensed image by the use of a cost minimization approach
Fang Luo, Lu Liu · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1994
In remotely sensed image analysis, image segmentation is still a difficult task for some images with high noise level such as SAR images. For the analysis of land parcels, we present an object-based method for image segmentation. (1) A cost function, based on the radiometric characteristic of objects, is defined. (2) An iterative procedure drives the cost to the minimum, in which simulated annealing plays an important role. (3) To improve segmentation result, the geometric and topological properties of objects are introduced to constrain image segmentation. Experimental results show that this approach can produce a good segmentation for images with low or extremely low signal-to-noise ratio.