<title>Patch matching in stereo images based on shape</title>

M. Abbasi-Dezfouli, T. Graham Freeman

Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1994 · 15 citations · 0 references

Concepts

Abstract

Matching stereo images is an important method for determining an elevation model of terrain, but is very time-consuming when performed manually. Several groups have sought to develop an automated method of elevation determination. The approach we have adopted can be applied to pairs of SPOT images or images from other sources. We search for patches of uniform color of at least 3 X 3 pixels in the two images, and attempt to find patches that correspond in the other image. Shape, size, and relative geometry are used to distinguish the correctly matching patch. A chain code method for measuring the boundary shape of patches is described, with pleasing results in practical tests.