Low DOF Images Segmentation

A curve evolution approach for unsupervised segmetation of images with low depth of field

Firstly, saliency map is got from original image by using object detection method with a multi-scale reblurring model;
Secondly, the initial curve is got unsupervisedly;
Thirdly, an active contour model which is based on hybrid energy is used to segment the low DOF objects.