Segmentation of mango region from mango tree image

In this paper we propose a novel framework for segmentation of mango regions from its tree image. The proposed framework consists of mango localization followed by mapping of boundary information to the located region for segmentation. Initially thresholding is applied to each individual color band R,G and B by adaptive thresholding and later they are combined back. Application of smoothing and binarization to the combined image gives the location of mangoes along with noise. The texture features are extracted from each location then matched with template stored in the database to eliminate the noisy regions. Finally, locations of the mangoes are obtained and edge information is superimposed on to those locations for segmentation. An experiment is performed on our own dataset and efficiency is evaluated by computing the precision, recall and F-measure with respect to the human segmented images considering as a ground truth