19^th International CODATA Conference
Category: Poster, Data Visualization

A Content Based Image Retrieval System Based on Color Features

Irena Valova (Irena@ecs.ru.acad.bg), University of Rousse “Angel Kanchev,” Department of Computer Systems and Technologies, Bulgaria
Boris Rachev (Bob_Ra@acm.org), Technical University–Varna, Department of Computer Systems and Technologies, Bulgaria

Significant research has focused on determining efficient methodologies for retrieving images in large image databases. Most Content Based Image Retrieval systems use low-level visual features for representation and retrieval of images. This paper addresses the design and implementation of a new image abstraction technique based on two compact signatures bit-strings and an appropriate similarity metric. It focuses on a low-dimensional global color features and spatial color distribution based indexing technique for achieving efficient and effective retrieval performance. We propose a combined index structure using these color features. Images are indexed by dominant colors and similar images form an image database cluster stored in a hierarchical structure. The regions within an image are further representing by their dominant colors and this color distribution representation is invariant to translation, rotation and scaling. A query engine supporting tree type of queries (query by image example, query by user sketch and query by global color features) is build in the prototype system to retrieve images by global and local color features. The retrieval performance is studied with a prototype system for content base image organization and retrieval, developed in C++ for Windows and an example collection of 3000 heterogeneous images from www.freefoto.com.