仿射不变的非负函数积分变换

摘要

The extraction of affine invariant feature plays an important role in pattern recognition, computer vision and so on. The contour-based method cannot deal with objects with several separable components. The region-based method is usually at the expense of high computational complexity and sensitive to noise in the background of the image. In order to overcome these defects, non-negative function integral transform (NFIT) is put forward. By using NFIT and stationary wavelet transform, a novel algorithm is also con-structed to extract affine invariant feature. Firstly, a general contour is constructed from the object by NFIT. Furthermore, the derived general contour is parameterized by equal area normalization method. Finally, sta-tionary wavelet transform is applied to the obtained general contour. Simulation results show that the method not only can deal with objects with several separable components, but also have some properties of low computational complexity and robustness to noise.%仿射不变特征提取在模式识别、计算机视觉等众多领域起着重要的作用. 为了克服轮廓类算法不能处理由多部分组成的目标, 区域类算法往往具有计算复杂度高和对背景噪声敏感特点的缺陷, 提出一种非负函数积分变换;并结合非负函数积分变换和平稳小波变换, 构造一种新的仿射不变特征提取算法. 首先利用非负函数积分变换将目标转化为一条广义轮廓, 然后利用等面积参数化方法对所提取的广义轮廓进行参数化, 最后对参数化后的广义轮廓进行平稳小波变换. 仿真实验结果表明, 该算法不仅适用于由多部分组成的目标, 而且具有计算量小和对噪声不敏感的优点.