A Subpixel Matching Method for Stereovision of Narrow Baseline Remotely Sensed Imagery

Abstract

In this paper, an accurate and efficient image matching method based on phase correlation is proposed to estimate disparity with subpixel precision, which is used for the stereovision of narrow baseline remotely sensed images. The multistep strategy is adopted in our technical frame; thus the disparity estimation is divided into two steps: integer-pixel prematching and subpixel matching. Firstly, integer-pixel disparity is estimated by employing a cross-based local matching method. Then the relationship of corresponding points is established under the guidance of integer-pixel disparity. The subimages are extracted through selecting the corresponding points as the center. Finally, the subpixel disparity is obtained by matching the subimages utilizing a novel variant of phase correlation approach. The experiment results show that the proposed method can match different kinds of large-sized narrow baseline remotely sensed images and estimate disparity with subpixel precision automatically.