Similarity between the features of the query image and that to the data base images is taken as the criteria for similar or relevant image retrieval. However, the features that falls into low level category may be enhanced to high level category by application of suitable weights to the features of reasonable rank. This may improve the similarity level between the query image and the data base images. In addition, in the presented work, the features are divided on the basis of domains like color, texture, statistical features (variance) and pattern specific. The feature extraction in different domain surely will enhance or improve the similarity index between the query image and data base images for faithful retrieval.
The feature set includes the following:
1 Color Domain color moments, mean of R-, G- and B-color components, standard
deviation of R-, G- and B-color component.
2. Statistical Features (Variance) Entropy, standard deviation, covariance
3. Texture Features Contrast, Homogeneity, Correlation, Power
4. Pattern Specific Features Figure Aspect, Symmetric Index, Max. And min. Radii, mean radii, standard deviation of radii etc.
The above domain features may classify completely the given query image and are sufficient to retrieve the relevant image from the data base. The performance evaluation may be done on the basis of precision and recall computation.