• Khaled Mahar, “JPEG Compressed Image Retrieval Based On Statistical Features And Blocks Co-Occurrence Matrix,” ”Proceedings of 4th international conference of informatics and systems, INFOS 2006, Cairo University, March 2006.

Abstract:Current trends in computer and internet use show a significant increase in the amount of images being distributed and stored in a compressed format. The most popular compression format defined by joint picture expert group (JPEG) becomes the current de facto standard for image compression. This paper introduces a framework for image retrieval system that operates in the compressed domain of JPEG. Compressed domain retrieval allows the calculation of image features, and hence the image retrieval, to be performed without full decompression. The proposed algorithm extracts a feature vector generated from the Discrete Cosine Transform (DCT) coefficient of the compressed images. This feature vector not only includes statistical information of a block color, but also carries coherence information of neighboring blocks. Due to the high dimensionality of the feature vector, principal component analysis (PCA) is used to reduce the dimensionality by eliminating redundant dimensions. Experimental results support the idea of the proposed algorithm in achieving good performance in terms of retrieval efficiency and effectiveness with comparison to other common methods.

• Mohand Dawod, Ameen Shoukry and Khaled Mahar, “Combining Features Using Principle Component Analysis and Independent Component Analysis for Image Retrieval,” Alexandria Engineering Journal, Vol. 48. No. 3, pp.273:278, May 2009.

Abstract: Research work shows that Independent Component Analysis (ICA) and Principle Component Analysis (PCA) are good variants of projection pursuit. In the present paper, a comparison between combined using PCA and ICA and uncombined features for Content Based Image retrieval (CBIR) is performed. The paper benefits from ICA and PCA and uses a Hierarchical Self-Organizing Map (HSOM) for better images clustering. To assess the performance of the proposed technique, two features provided by MPEG-7 are used, Color Structure (CS), and Edge Histogram (EH) are used. Research work showed that these two features give better performance among other MPEG-7 features such as color and texture features. The proposed technique is applied to two datasets. The results showed that using ICA in the combining process gives better performance that PCA, and PCA gives better performance than uncombined features.