IMALBUM: Content Based Image Retreival System
|
The developments achieved
during this work aimed to propose a color-based system for image
retrieval. Those systems integrate many techniques as image
segmentation and description, similarity evaluation, search processing.
A new color image segmentation method based on the minimization of the
cross entropy is proposed, dealing with an efficient and compact color
descriptor. Two others color descriptors are proposed to compensate the
lake of the spatial information related to the histogram approach. The
developed system is compared to the one suggested by MPEG-7. An
objective evaluation method for color image similarity is proposed,
with the use of synthetic image database.
When the color and the semantic of an image are not correlated, the use
of the color as unique attribute is not relevant. Therefore, other
attributes, as texture and shape, have to be considered for image
description. We propose a local approach for image retrieval, where the
user’s query is given by an objet of interest defined by the user. ART
descriptors and Haralik and Gabor descriptors are respectively
used for shape and texture.
Finally, a new method for images database access is proposed, allowing
the user to visually navigate in the database, using a semantic
criteria. In some cases, the tool allows objects identification with
the use of a reference list.
The proposed system, namely IMALBUM, offer a friendly user interface
for indexing, similarity search and visual navigating through image
databases.
original
image
segmented image with 30 colors IMALBUM User
Interface 
|
2-K. Idrissi, J. Ricard, G. Lavoué, A. Baskurt, “Object of interest based visual navigation retrieval and semantic content identification system”, Computer Vision and Image Understanding CVIU , Vol. 94, n° 1-3, pp. 271-294, April-June 2004.
3- K. Idrissi, J. Ricard, A. Baskurt, “An objective performance evaluation tool for color based image retrieval systems”, In Proc. of IEEE Int. Conf. on Image Processing, ICIP'0, Vol. 2, pp. 389-392, Rochester, USA, Sept. 2002.
4- K. Idrissi, J. Ricard; and A. Baskurt ,"Multi-component Cross Entropy segmentation for Color Image Retrieval", In Proc. of the 2nd International Symposium on Image and Signal Processing and Analysis, Pula, Croatia, pp. 132-137, June 2001. pdf
5- K. Idrissi, J. Ricard, A. Anwander and A. Baskurt, "An Image Retrieval System Based on Local and Global Color Descriptors", In Proc. Of the 2nd IEEE Pacific-Rim Conference on Multimédia, Beijing, China, October 2001. pdf