Flower Region Extraction and Identification using k-Means Clustering

Abstract

Diversity in flower colors and different light conditions make the image segmentation task more
complex. Varity of flowers are available in nature. Flower play important role in species identification
due to its different shape and color. k-Means Clustering is the popular technique used for segmentation
of color images. Aim of this paper is to use k-Means clustering, L*a*b* and HSV colorspace to achieve
the segmentation and cluster labeling task. L*a*b* colorspace segmented flower regions from color
images. k-Means technique clustered the image into flower and background. Each cluster is mapped to
HSV colorspace. The clusters are labeled as flower and background by calculating mean of Value
channel. The OXFORD-102 challenging dataset is used for experimentation purpose. The proposed
technique shown promising results for segmentation and labeling. 75% images are correctly
segmented and 81% images are correctly labeled by the proposed techniques