摘要
为了解决视差图像的快速、鲁棒配准问题,提出了一种基于小波近似图像和垂直细节图像由粗到精的加权全局配准策略.首先利用小波变换的多分辨率特性,对图像进行小波变换生成一个基于灰度的金字塔,从金字塔顶层到底层对每一层的近似分量进行由粗到精的配准.为了进一步提高配准可靠性,在金字塔次底层分别对近似图像和垂直细节图像进行加权相关配准,得到稳定的最优配准.
To solve the quick and robust registration of parallax images, a coarse-to-fine global registration strategy using the coarse images weighted with detail image was proposed. Taking the advantage of multi- resolution characteristic of wavelet transform, a gray pyramid was built by the wavelet transform of images, and the approximate components of each layer was matched layer by layer from top to bottom. To improve the reliability of registration ulteriorly, the optimized registration was gotten from the weighed correlation registration of approximate components with vertical details in the penultimate layer of the pyramid.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2007年第3期574-576,共3页
Acta Photonica Sinica
关键词
小波变换
视差图像
几何全局配准
加权相关配准
Wavelet transform
Parallax images
Geometric global registration
Weighted correlation registration
