Automatic segmentation of leaf images based on an improved geometric active contour model.

Measurement of leaf morphology is significant in automatic monitoring of seedling growth. The leaf images have to be extracted from the background prior to morphological measurement. We developed an automatic segmentation method of intact leaves based on the geometric active contour model according to the characteristics of color leaf images. The global information of images and C-V model were used for initial segmentation. When the curve moves close to the border of the object，the boundary of the object was located using the improved model based on the local information of the image. The proposed segmentation method combines the advantage of C-V model，i.e.，not likely to be affected by original position of the curve，and the strength of the improved model which overcomes the boundary leakage. Experimental results show that the proposed method can effectively segment leaf images.