关键词: 主成分分析, 多尺度分解, 局部方向估计, 边缘提取

Abstract: Most of the existing image orientation estimation algorithms are very sensitive to noise. Therefore, this paper presented a new local orientation estimation method of image, which was based on Principal Component Analysis (PCA) and multi-scale gradient pyramid decomposition. The PCA was applied to search for the Maximum Likelihood (ML) estimation of the local orientation of block in image. Through presenting the local orientation estimation both of simulated images and real images, the experimental results show that the proposed algorithm has excellent robustness against noise, its speed is very fast and the accuracy is very high.

Key words: Principal Component Analysis (PCA), multi-scale decomposition, local orientation estimation, edge extraction