Semester

Summer

Date of Graduation

2006

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Donald A Adjeroh

Abstract

In the development of an automatic image mapping system for mouse brain images, we focus on obtaining the best match in the two dimensional atlas images for the query image of a mouse brain section. The matching issue has to deal with the tissue distortions and tears, which are routinely encountered and possible scale, rotation and shear changes. With the goal to achieve high performance for image matching, a dual-stage approach is proposed which is based on radial distributions and region-based template matching.;Our main challenge is identifying shape features and corresponding distance metrics that produce effective characterization of 2D sections for similarity comparison. The mapping algorithm analyses the shape characteristics of target models and perform a similarity measurement against database templates. In this technique, the first stage makes use of the contour of the brain section and intrinsic local geometric features to estimate the approximate location of the given section along the brain axis. Here, we considered three approaches: morphometric-based features, Fourier based descriptors and radial distributions. In the second stage, a content-based approach is adopted in which we use texture and localization features to make a final decision about the mapping. More specifically, we studied morphometric-based methods and the use of Gabor filters for this fine-grained analysis. The statistical distribution of these features constitutes the essential concept for achieving a match between reference and subject images.;With the efficient integration of the stages, our proposed approach for histological brain mapping gives the benefits of both speed and accuracy. Comparative results with other proposed methods are studied.

Share

COinS