Venation extraction of leaf image by bi-dimensional empirical mode decomposition and morphology

Leaf vein extraction is a key feature for plant recognition. An efficient leaf vein extraction method is proposed in this paper by Bi-dimensional Empirical Mode Decomposition with Gray-scale Morphology Processing (BEMD-GMP). The raw image transforms gray image firstly, the Intrinsic Mode Functions (IMF) component of the gray image is decomposed by Bi-dimensional Empirical Mode Decomposition (BEMD). We choose the first IMF component to segment venation by Gray-scale morphology operator, because the high frequency component and the noise has been removed the first IMF component. We analyzed the four different evaluation criteria for Gabor filter, Canny filter, Canny operator, Soble operator and BEMD-GMP. The experimental results show that the method of BEMD-GMP can obtain more satisfactory results.