In image large-scale overlap region of the effective segmentation is always a difficult problem

the traditional Log operator

Sobel operator

Canny operator and gradient operator and algorithms to solve large scale pixel overlap problem

model will in no convergence situation

causing segmentation effect is bad

in order to solve this problem

the paper proposes a kind of enhanced EM model is used to solve the problem of overlapping area image segmentation

use curvelet transform in curvelet domain of images edge features

and positioning characteristic curvelet coefficient.Through the enhancement features curvelet coefficient

enhance the edge feature of kriging

multi-scale segmentation more structural elements morphological detection edge image

and eliminate the interference of overlapping brings.The simulation results show that the edge of the segmentation is more complete and accurate

and satisfactory results have been obtained.