In this paper, a novel method is proposed to extract a stable feature set representa-tive of image content. Each image is represented by a linear combination of fractal or-thonormal basis vectors. The mapping coefficients of an image projected onto each or-thonormal basis constitute a feature vector. The distance remains consistent, i.e., isomet-ric embedded, between any image pairs before and after the projection onto orthonormal axes. Not only similar images generate points close to each other in the feature space, but also dissimilar ones produce feature points far apart. Therefore, utilizing coefficients de-rived from the proposed linear combination of fractal orthonormal basis as a key to search an image database will retrieve similar images; while at the same time exclude dissimilar ones. The coefficients associated with each image can be later used to recon-struct the original. The content-based query is performed in the compressed domain. This approach is efficient for content-based query.