Multiresolution Object-of-Interest Detection for Images
with Low Depth of Field

Jia Li, James Ze Wang, Robert M. Gray and Gio Wiederhold
Stanford University, Stanford, CA 94305

This paper describes a novel multiresolution image segmentation algorithm for separating sharply focused objects-of-interest from other foreground or background objects in low depth of field (DOF) images, such as sports, telephoto, macro, and microscopic images. The algorithm takes a multiscale context-dependent approach to segment images based on features extracted from wavelet coefficients in high frequency bands. The algorithm is fully automatic in that all parameters are image independent. Experiments with the algorithm on more than 100 low DOF images have shown results close to the human segmentation of these images. Besides high accuracy, the algorithm also provides high speed. A 768 x 512 pixel image can be segmented within two seconds on a Pentium Pro 300MHz PC.

Full Paper in Color
(PDF, 0.2MB)

Full Paper in Color
(PostScript, 1.5MB)

1999 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

Last Modified: Fri Oct 2 00:35:07 PDT 1998
1998, James Z. Wang