Segmentation of Leaf Images Based on the Active Contours

Abstract

Leaves contain important genetic information which can be used as a basis for the identification of plants. As a first step of modeling virtual three-dimensional plant, how to extract visual characteristic information form leaf images has great significance. We propose an optimized C-V model in this paper, which can detect objects in homogeneous regions of given leaf images and speed up running time. The new method combines local information with global information and optimizes the defect that SDF needs to be reconstructed partially so that the energy function is improved. Experimental results show that our algorithm can stop active contours on the correct boundary, get accurate image segmentation, and the speed is more than 1.5 times faster to C-V model.