Leaf Extraction from Complicated Background

Leaf extraction is an important step for automatic feature extraction, leaf classification and plant species identification, etc. But it is still a challenging problem especially from the images with complicated background such as with some interferents and overlaps between two adjacent leaves. In this paper, we have developed a leaf extraction algorithm to extract the target leaf from the images with complicated background. In this algorithm, the marker-controlled watershed segmentation method is applied on the gradient images of hue, intensity and saturation of the HSI color space, separately. The solidity (integrity) measure is then used to evaluate how well the segmented image is for extraction of the target leaf and determine the final leaf extraction result. Experimental results on some practical soybean leaf images demonstrate the effectiveness of the proposed algorithm.