A New Approach for Choice of Optimal Spatial Scale in Image Classification Based on Entropy

The existing methods for choice of optimal spatial scale are evaluated.It is pointed out that these are some methods based on statistics.However,these methods have not taken into account the spatial distribution.A new approach based on information entropy is introduced to select an optimal spatial scale in image classification.An experimental evaluation is also conducted.Results show that the new approach is more meaningful than traditional one.The proposed method will be useful for a variety of scale-related land cover classification tasks.