A Hybrid Approach to Indexing Image Databases
Abstract
Image database storage and retrieval continues to be a challenging problem. There are two categories of images: still images and motion pictures. The images in both these categories have specific requirements for archival and retrieval. In still images, each image is separately processed for storage and retrieval, whereas in motion pictures the frames are processed in groups(GOP). Image retrieval may be specified in terms of image features or image intensity color content. In the past, several techniques have been used for indexing images. Some of them are holistic wavelet based which lack invariance to rotation, scaling and translation. Others use a combination of image segmentation, Fourier transform and histogramming. We present a new hybrid technique, which uses the compression power of wavelet transform and rotation, scaling & translation (RST) invariance of Fourier Transform Spectrum. This hybrid technique is faster, and robust in the construction of image signatures. The experimental results on signature algorithm, accuracy, and speedup are presented. This indexing technique is useful for reliable content-based image query resolution.
Recommended Citation
C. Sabharwal, "A Hybrid Approach to Indexing Image Databases," Intelligent Engineering Systems Through Artificial Neural Networks, American Society of Mechanical Engineers (ASME), Nov 1998.
Department(s)
Computer Science
Keywords and Phrases
Compression; Discrete Fourier Transform; Image Query; Signature Algorithm; Wavelet Transform; Indexing
Document Type
Book - Chapter
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 1998 American Society of Mechanical Engineers (ASME), All rights reserved.
Publication Date
01 Nov 1998