基于主成份分析的仿射不变特征图像匹配方法

Affine Invariant Features Image Matching Approach Based on Principal Components Analysis

从理论和实验上分析比较了尺度不变特征图像匹配方法和仿射不变特征图像匹配方法的异同,为提高图像匹配的速度和图像特征描述的准确性,提出了一种利用主成份分析对仿射不变特征匹配特征描述向量降维的方法。实验结果表明该方法能加速仿射不变特征图像匹配方法,在不降低匹配效果的前提下,将128维特征向量降低至15维。

A comparison of the Scale-Invariant Features Transform approach and Harris-Affine Invariant features approach was studied theoretically and experimentally. To improve the speed of image matching and accuracy of image features descriptors, an approach of reducing the dimensions of image matching features descriptors was proposed, which was applied to Harris-Affine region detectors based on Principal Components Analysis. Experimental results show that the method can accelerate the Harris-Affine Invariant features approach. On the premise of not degrading the matching effect, the method can reduce the dimensions of features vectors from 128 to 15.