Classification of Flower Images using Clustering Algorithms

This paper presents a flower classification and identification system that takes a flower image as input and identifies it to be belonging to a particular category present in the database. It begins by performing pre-processing operations on the input image. A set of digital images are segmented using k-means clustering algorithm from which texture and color features are extracted. Texture features are extracted using Log Gabor filters and Color features are extracted by calculating Mean and Standard deviation of color distribution from R, G, and B color channels. Texture and color features are concatenated to form a feature matrix. Features are clustered using K-Means clustering algorithm and Prim’s minimum spanning tree algorithm to obtain classes. Test image will undergo the same segmentation and feature extraction process. Finally test image is identified as belonging to particular class based on similarity measure. The algorithms are implemented in MATLAB using Image Processing toolbox.