With the explosive advancement in imaging technologies and specially with proliferation of the world wide web, image retrieval has attracted the increasing interests of researches in the field of digital libraries, image processing and database system. Research in human perception of image content suggests that content-based image retrieval (CBIR) can follow a sequence of steps. The typical steps of CBIR system are: image query formation, image feature extraction, similarity measurement, indexing and retrieval, and user interaction. The correct choice and set up for each step will result in a well, efficient and suitable CBIR system.This work concentrates on one important and crucial step of the whole CBIR system: feature extraction (or feature formation). The image features used are all characterized as low – level features. These include: image luminance histogram, low – passed luminance histogram, luminance pyramid, color pyramid, and combined feature.he main contributions are: simplicity (i.e. easy to implement the feature extraction phase), suitability (i.e. provide acceptable retrieving results), efficiency, and economy. The CBIR with the presented feature extraction variants are tested on a selected database of a set of thirteen image classes. In general, the results indicate that the choice of image feature can greatly affect the performance of CBIR system. Experimental results showed that image features that utilize achromatic and chromatic information of the image can provide about 75% accurate results, while those depend on only intensity information can give accurate results in about 25% - 75%. Moreover, the combination of two features can give in better results.
