Abstract:Normalized neighborhood variance is defined, which is selected with pixels gray value, neighborhood mean as
the condition attributes of image granular space. A tolerance granular space of images is constructed by condition attributes.
The images are granulated by using tolerance relations. A tolerance granular distance measure function is defined. Mind
evolutionary algorithm(MEA) is used to optimize the thresholds, then decision granules are synthesized. Finally, target
regions are extracted and image segmentations are completed. The experimental results show that the proposed algorithm
has better denoise effects, strong stability and rapid convergence velocity.