IJSR - International Journal of Scientific Research

Sangeetha Mary; Suvya P

Volume : IV, Issue : VI, June - 2015

Natural Image Statistics with Low–Level Hierarchical Segmentation in Content based Image Classification

Sangeetha Mary, Suvya P

Abstract :

?is paper aims at providing the statistics of the natural images. ?e properties being considered include the geometric, photometric and topological characteristics of natural images. ?e image is segmented using a lowlevel hierarchical multiscale segmentation algorithm. ?e algorithm provides different regions of the image which are arranged in a tree structure called segmentation tree. ?e geometric, photometric and topological characteristics of the regions of natural images are studied to represent the statistics of region properties. Along with the statistical properties some features like color, texture, edges and wavelet information are taken into consideration using a discrete wavelet transform. ?ese statistical features are used to model a dataset of natural images by a probability density function. ?e model then used to train the dataset and finally for image classification