We propose a new simple method of DC feature extraction thatenables to accelerate the speed and decrease storage needed in image retrieving process by aim of partial Joint Photographic Experts Group (JPEG) compressed images algorithm. Our feature extraction is carried out directly from JPEG compressed images. We extract one feature of DCT coefficients which is DC feature only. By applying DC feature as indexing key, this method only need 1/64times required to construct indexing keys compared to conventional Euclidean distance method whilststill maintain high precision and recall.Â We also introduce a DCT coding category approach to justify this proposed method.Â After all, even though this method is not revolutionary and perfect yet in improving speed for image indexing,itvery significantly reducedthe cost of image retrieval in real applications, primarily for huge image databases.
Copyrights © 2017