Masters Theses

Date of Award

5-1995

Degree Type

Thesis

Degree Name

Master of Science

Major

Electrical Engineering

Major Professor

M. A. Abidi

Committee Members

Marshall Pace, Daniel B. Koch

Abstract

Range images are important in robotic environments as they provide us with three-dimensional depth information about a scene. However, this information is too low-level for performing scene interpretation directly. Several computational steps are needed in order to extract and characterize meaningful features from raw range data. One of these computational steps is segmentation, the process of partitioning the image into groups of pixels that reveal the inner borders between three-dimensional scene elements and surfaces in an image.

In this research, a segmentation technique for range images is proposed using the wavelet transform (WT) and multiresolution analysis. The wavelet transform is used as a smoothing multiscale differentiation operator. A model for range image features is developed based on the differential properties of step and roof edges in range images. By applying one-dimensional oriented band pass filters at multiple scales, use of the wavelet transform avoids the execution delays associated with area-based segmentation algorithms for range image analysis. The result is a segmented image with labeled regions indicating different surface patches.

The developed algorithm is reasonably fast, taking under three seconds for the complete segmentation of a 128 x 128 range image on a Sun SPARCstation. The computations are implemented efficiently and may be performed in parallel. In computing the transform, one automatically has access to image features of various sizes that are enhanced and available for three-dimensional reconstruction and recognition tasks.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS