基于相位一致的多尺度金字塔图像特征提取  

Image Feature Extraction Algorithm of Multi-scale Pyramid Based on Phase Congruency

作　　者：黄蕾[1] 邹海[1] 

机构地区：[1]安徽大学计算机科学与技术学院,安徽合肥230601

出　　处：《计算机技术与发展》2015年第3期36-39,共4页Computer Technology and Development

基　　金：国家科技重大专项基金资助项目(2009ZX05039-004)

摘　　要：图像特征提取是数字图像处理与模式识别领域中的关键问题,各种特征提取方法层出不穷。相位一致图像特征提取方法是基于局部相位信息进行图像特征提取,具有亮度和对比度不变性的优点,但是在轮廓特征提取方面存在缺陷。考虑到多分辨率、多尺度对图像特征提取的影响,提出一种基于相位一致的多尺度金字塔图像特征提取算法,新算法的关键在于拉普拉斯金字塔的分解和多尺度特征图像的融合。实验结果表明,该算法在提取图像轮廓特征方面要优于传统的相位一致图像特征提取算法。Image feature extraction is the key issue in the field of digital image processing and pattern recognition. The feature extraction methods are emerging in endlessly. Phase congruency image feature extraction method is based on local phase information for feature extraction,which has advantage of brightness and contrast invariance,but there are still insufficient for contour feature extraction. In order to fully consider the influence of multi-resolution,multi-scale image,present a feature extraction algorithm of multi-scale pyramid based on the phase congruency. The key is Laplacian pyramid decomposition and multi-scale feature fusion. The experimental results show that the new algorithm is superior to phase congruency image feature extraction algorithm in the contour of image feature extracting.

关 键 词：特征提取 相位一致 多分辨 多尺度金字塔 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]