Retrieval of Color Images using Statistical Feature Tree

Abstract

Efficient Image retrieval system with effective features is very essential. Digital images having optimal dimensions and high discriminative power is important requirement for such systems. Image Retrieval systems uses the image feature descriptors to retrieve similar images based on its similarity from the large database. Image Feature Vector generation is important step in any content based image retrieval (CBIR) system. Size and discrimination power of feature vectors affect performance of image retrieval system. These two parameters must be focused while designing the feature vector of any digital Image with its extracted inherited features. These two parameters are solely depend on the image feature extraction and feature storage approach followed. Most of the approaches of feature vector generation uses single level statistical summaries for extracting image features. Though these summaries are small in numbers hence useful to create compact feature vector. Generally these summaries carry low discriminative power. As per current trends it has been observed that most of the researchers have used Linear Data structure to store the extracted image features. Linear Data Structure faces the problem of handling parallel execution of algorithm. the This research work focuses on feature vector generation using extraction of cascaded statistical summaries of the image and store these level-wise summaries in Hierarchical Data Structure like Binary Tree. newlineLow-level Feature: Color Descriptors are used as Image Contents, to represents the image in feature vector form. Images in different color spaces are used for performance analysis of generated feature vector in Image retrieval. Images in Spatial Domain and Frequency Domain are considered for Feature Vector Generation. Energy compaction property of various Image Transforms is extensively used while creating the feature vectors of frequency domain images.