Document Type

Research Paper

Abstract

In the "content based" image query, also referred to as "query by example," "similarity retrieval" or "sketch retrieval," the query image is provided by the user either as a sketch of the object, as the output of a scanner or a video camera. Some of the difficulties associated with content based image query are described in [1], e.g. significant difference between the "query image" and the "target image," artifacts and poor resolution of the query image make a straightforward comparison of images using L1 and l2 metrics not effective.

In [2] a new strategy is suggested based on wavelet decomposition of the query image and the database images combined with a metric which is designed to be insensitive to small differences in the query process. This approach is found to be fast and overcomes the above mentioned problems.

Wavelet coefficients of an image may be very strongly when the image is displaced or rotated (unlike color histogram of an image which is invariant under displacement and rotation). Although the metric suggested in [2] is more robust to these errors when compared to the L1 and L2 metric (but worse when compared to the metric based on color histograms), still the error is significant.

In this paper we provide some experimental data on the sensitivity of the wavelet coefficients to displacement and rotation in the context of the standard characters and suggest an integration of the Hotelling transform to improve on this sensitivity. We also provide some experimental data on the distribution of the largest wavelet coefficients at different levels of the wavelet decomposition and discuss some questions relevant to the approach of using a set of the largest wavelet coefficients for image query.

 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.