1 Project: NETRA 2 - A Region-Based Image Retrieval System


Yining Deng, B.S.Manjunath


This is a prototype image retrieval system that allows users to search and retrieve images in the database based on color information. One of the distinctive aspect of the system is that it allows users to localize the information and select image regions of interest as queries, thus providing a more powerful search tool than other retrieval systems that use global image features. Compared to the original NETRA demo , this new version of NETRA emphasizes our latest work on color image segmentation and local color feature.


A new color image segmentation algorithm is developed and used to segment the images shown in the demo. This algorithm is fully automatic and fast. More information of the segmentation algorithm can be accessed from here.


A new representation for local color feature is used in the system. The main contributions of work are on color quantization, feature distance measure, and database indexing scheme. The color features are extracted from segmented image regions. First, colors in each region are quantized to a small number of representing colors using a perceptual quantization algorithm. The feature descriptor consists of these colors and their percentages in the region. A similarity measure is defined for the proposed color feature and is shown to be equivalent to the quadratic color histogram distance measure. The representing colors are indexed in the 3-D color space so that high-dimensional indexing can be avoided. During the search process, each quantized color in the query is used as a separate cue to find regions containing that color. The matches from all the query colors are then joined to obtain the final retrievals.


Image source: Corel photo CDs

total number of images: 2500

number of images in each set 100

total number of region regions: 20367

image size: 192x128 or 128x192

image format during processing: raw RGB

image format for displaying: jpg

JSEG Image Segmentation

fully automatic, a fixed set of parameter values are used for all 2500 images without any parameter tuning for individual images

Color Feature and Database Indexing

average number of quantized colors in each image region: 3.2 

average size of each color feature: 12.8 floating-point numbers


These materials are presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each authors copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

  1. Y. Deng and B.S.Manjunath, "An efficient low-dimensional color indexing scheme for region based image retrieval,"Proc. IEEE Intl. Conference on Acoustics, Speech and Signal Processing (ICASSP-99), Phoenix, Arizona, March 1999. [abstract]


Due to technical difficulties, Netra II demo will not be available for future use. 

Related works

NETRA systems                           Color image quantization               Image segmentation

Home | People | Research | Publications | Courses | Seminars | Links