Segmentation of leaf veins using principal curvature and morphological methods

Extracting the network of leaf veins is a crucial step in characterizing and classifying the species of plants. Researchers have categorized the plants for many years by understanding several vein features including vein density, length per unit area, vein topology, etc. To extract leaf veins, this paper presents a simple and effective approach based on image processing methods that utilized information about the curvature at each pixel of a leaf image. In the proposed system, Gaussian filtering is performed as a pre-processing step to remove the noise and area opening as post-processing for removing small undesired segments from the segmented vein network. The maximum principal curvature image is computed followed by contrast enhancement that resulted in the segmented veins image.